1,: 241,. 1.

.3

“ma L

an: .-vsu

n! .

:m".

‘11:“.
1,)"
L.'I‘ 1 av .. c 1..

1.. Z‘ti"_'L:1.
1i

9—: ‘
‘SrEVL A“;
v.14.

1
. ‘ yd11'?'11>.“.

11.. J.

1"“ 1.,
. a. u‘ 4"-
- 2.51," I“. 1“

-
"‘3..- .
”-1?me

, l. .1
.r VA
\ ..-
., .
1‘ u'

«$.11

1,. u afi-

.1

:5111 117.13“. 1

.1‘

1m.
wi .m

1r.
{1%.} “31.121
1‘ “o4

'.."J::.

L‘

.Aty-j' 1- u
-1 1 mm ~11- v

\m
*1“. .. u» \

’..._:. 1.9;"
‘3.

n 5::11
\:| n;

.- 3..
13?.11‘19

1: “.1

. v
T-nxi-

V 1
3.1:": _.\,_.
it}, \uufJ \
21:21“
Il‘m‘d'.‘ .WJ"
l“:" It:

.
1 . .1 “1.
' gruuga

"" ' .41
“($31.qu
\1‘.‘ ., ,1:

. l" fi’;~';1\~.rm' ~. )

yum {‘jq‘w.
-.\"" u ‘31."

11:11; 111;“

1-
1.7 in: - 24a

.11..
.

'1

”-13%:

'1893174' 'f-‘J

fsfjfni

-"x.~.. -',-|
9:

,
~1

A..
l

7',
--w.4‘4‘6
'15 >

4'1: 1-1
Av- V

v .
" at .11...
.

t
53;" ‘ "~ up ‘15.: 1'11 H.” 1'
“ «11:1
1"

I‘d;

fi’x‘

1.3:;

“‘1 A -. _J‘
1 553:”: .j flak: ‘
.1 3:11;.qu

291.1113:

.31 . ,
. 39.1.9 .
““23”"...
‘ {Artistry}

W11
3%:
1).”. '1“. *5; 1"".1’ ..:
1:1.” 1,.“ «in: 511;}

u:

u\-,

rm
“ Q

{in 5:: ..
’F ! L511
-
w A?
um. ..
m ”“11": 3111:. #111513
«In. ‘
h“”‘.‘-"" “'2: ‘ "‘1’ L1 .111 a:
* .11; a - ,.
0. “w fit
“(‘0

1 ”‘1'

n m
~1

.
‘gm 3

 

:LVJL‘:
:vw
”1.. "_"*'1"

1:13"? 01 1‘: _ . a}

 

 

 

 

 

31293

This is to certify that the

dissertation entitled

A RECREATION LIABILITY KNOWLEDGE
BASED SYSTEM

presented by

PETER K. FORSBERG

has been accepted towards fulfillment
of the requirements for

_Eh‘.D_-__degreein Park and Recreation
Resources

Major professor

fwd/MW
/

Date December 1 9 9 O

 

MS U is an Affirmative Action/Eq ual Opportunity Immune» 0-12771

ll‘lllllllull\\ \\\"i\l\lllll " Z “’ 4 v

 

 

LIBRARY
Michigan State
; University

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.

77 ='
DATE DUE DATE DUE DATE DUE

. “‘ 5
8 g g .‘ may \
9

.113 i: 51 97

.- p
i ‘~‘\< L r
A jib, I

 

 

 

 

 

 

 

 

 

       
   

 

 

 

MAY2419 9

 

 

 

 

 

 

—l
J

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

LT—‘T

MSU Is An Affirmative Action/Equal Opportunity Institution
czlclrcmma-pj

 

 

 

 

 

———————-’——'

A RECREATION LIABILITY KNOWLEDGE BASED SYSTEM

BY

Peter K. Forsberg

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of

Park and Recreation Resources

1990

,.
nL'.

6045~ 7

ABSTRACT

A RECREATION LIABILITY KNOWLEDGE BASED SYSTEM

BY
Peter K. Forsberg

Providers of recreation opportunity operate within a
legal environment that requires them to consider the
safety of participants. Economic and ethical factors
also provide incentive for providers to maintain safe
land conditions for the recreating public. Negligence
law as it pertains to premises liability is the formal
body of knowledge that forms the basis for legal decision
making therein.

Risk management is a proactive means to limit the
potential for injury and subsequent civil remedy. The
determination of legal negligence, however, is the final
arbitrator in assessing the conduct of a defendant who
.manages .land. conditions for recreation. activities.
Unfortunately, this decision-making process is
encountered only when one is named as a defendant in a
suit. Thus it is difficult to understand the decision-
making process determining negligence ‘without entering

the civil law arena.

Peter K. Forsberg

A knowledge based computer program was constructed
to model this decision-making process. Sources of the
domain expertise include statutory and civil law, legal
publications, and experts within the field of recreation
law. The research prototype system strives to assess the
existence of landholder negligence as the plaintiff's
burden of proof prior to a civil case. Defenses to the
assertion of negligence are also assessed. The goal of
the research is not the ultimate determination of
negligence, but the identification of key issues
associated with a case.

Hypothetical case facts were supplied by the domain
experts to test the system's ability to identify key
issues and estimate the type and degree of negligence
present. The outcome was then compared to the expert's
assessment of the case.

These tests indicated that the system performed
adequately for the stage of research prototype yet
remained fragile in the manipulation of legal concepts
that have contextual meaning.

The results of this research indicate that knowledge
based systems show promise for modeling recreation law.
The research also highlighted avenues for improving

future systems.

ACKNOWLEDGMENTS

I would like to thank Dr. Lewis Moncrief’ whose
patience and guidance in the role of my major professor
has been invaluable. I would also like to acknowledge
Dr. James Bristor, Dr. Stephen Harsh, and Dr. Leighton
Leighty for their efforts in keeping this project on
track.

Dr. Don Holecek of the Michigan Travel, Tourism, and
Recreation Resource Center generously provided assistance
in the form of use of computer resources. Thanks are
also due to Dr. Edward Mahoney for serving as mentor, and
funding source, and friend.

Special thanks to my wife Jane who has given me
love, sustenance, and a daughter during my graduate

studies.

iv

Table of Contents

Page

LIST OF FIGURES .................................... vii
LIST OF APPENDICES ................................ viii
PROBLEM STATEMENT .................................... 1
Introduction........... .......................... 1

The Problem......... ............. .. .............. 2
Potential Uses of the Study ...................... 8
Objectives ....................................... 9

The Research Document .......................... 9
Definition of Terms ....... ....... ........... .10
Artificial Intelligence Glossary. ............ 10
LITERATURE REVIEW. . . ................................ l 3
The Expert System Approach ..................... .26
History of Expert Systems ..................... 26
Expert System Components ...................... 28
Neural Networks... ..... .. ..................... 32

THE STUDY............ ............................... 37
Expert System Applicability ..................... 37
Expert System Suitability ....................... 37
Expert System Justification ..................... 40
Expert System Appropriateness ................... 40
Tool Selection....................... ..... ....43
System Building ............................... 44
Knowledge Acquisition.. ....................... 47

IMPLEMENTATION ...................................... 51

The Knowledge Base.......... .................... 51
Civil Law Issues ..... . .......................... 54
Public Entities ............................... 54
Rural and Urban Lands. ................ . ...... 57
Excerpted Legal Concepts as Defined by
Black (1986) .................................. 58
Knowledge Base Structure ....................... 64
Rule Structure ............................... 69
Basic Legal Concepts as Defined by
Black (1986) .................................. 66
Search Strategy. .............................. 74
Model Structure........ ...... . ................ .75
Open Texture Predicates ......................... 80
Analysis and Results ............................ 83
Test Cases. ...... . ........ . ............ .......84
Discussion..... .............................. ..104
Summary and Conclusions ........................ 110
Summary ...................................... 110
Conclusions....... ..... .............. ....... .113
Limitations.. ................................ 115
Recommendations for Further Study .............. 115
APPENDICES... ...... .... ............................ 117
BIBLIOGRAPHY ....................................... 163

vi

LIST OF TABLES

TABLE Page

1. Bugs Arising from Knowledge Programming ..... 106

vii

LIST OF FIGURES

Figure Page
1. Risk Management Decision Matrix ............... 6
2. Generalized Decision Tree in RANGER .......... 17
3. An Illustration of Object Inheritance ........ 33
4. Architecture of a Hidden Layer Neural

Network ...................................... 35
5. Domain Characteristics Necessary for

Expert System Development .................... 38
6. Factors Justifying Expert System

Development .................................. 41
7. Characteristics Making Expert System

Development Appropriate ...................... 42
8. The Hays-Roth et al. Life Cycle Model

for Expert System Development ................ 45
9. Modules in TOTO's Knowledge Base ............. 71
10. Evolution of Expert Systems .................. 85
11. Firing of the Permission Rule and

Subsequent Pruning ........ .... ............... 90
12. Traditional Logic v. Fuzzy Logic ............. 95

viii

LIST OF APPENDICIES

Appendix Page
A. Test Case 1 ..................................... 118
B. Test Case 2 ...................................... 124
C. Test Case 3 ...................................... 129
D. Edited Knowledge Base in TOTO ...................... 134
E. Portion of TOTO's Knowledge Tree ................... 154
F. Sample Session Reports ............................. 160

viii

PROBLEM STATEMENT

Introduction

Liability and insurance issues have impacted many
businesses and governments by threatening their
feasibility of operation. This threat is often realized
in the form of costs of civil litigation and insurance
premiums. The recreation industry has become a
participant in the increase of liability costs due to the
risky nature of many recreation activities that occur
within private and public facilities. When a recreation
provider becomes a defendant in a liability case, the
cost of litigation and potential remedies are incurred.
If an operation avoids litigation, costs are still
realized in the form of insurance premiums that are based
upon the risk of litigation (demonstrated by operators
who have been sued). Much concern has been expressed on
how these symptoms may impact the recreation in Michigan.
The outward costs of liability issues, however, are only

manifestations of a complex decision—making process

 

carried out in the formalized arena of civil law.
Normally an individual's first exposure to this process
is when they are named in a suit. Unfortunately, this

exposure comes too late to enable the correction of

situations where negligence may exist. The realm of
artificial intelligence, and specifically knowledge based
systems, provide a means to model this process and
enabling participation in legal decision making without
going to court. It is proposed that a knowledge based
system be constructed that embodies the domain of
negligence law as it pertains to private landholders,
facility operations, as well as local, state, and federal

levels of government within Michigan.

The Problem
Recreation and tourism has been promoted as a
potential element in creating a stable, diversified
economy in Michigan. This industry is made up of many
independent private and public operators. As liability
costs escalate, these operators incur the burden on
1. Direct costs of insurance premiums
2. The threat of civil action by invitees
(customers)
3. The potential costs of civil remedies in torts
brought against operators (Hronek, 1986).
These elements impact the profitability of individual
operations and can even force them out of business
(Headrick, 1985). These costs also can prohibit new
developments altogether. For the recreation consumer,

these elements are manifested as:

1. Limited diversity of recreation opportunity

2. Increase on-site expenses as liability costs are

passed on to consumers

3. The creation of an adversarial relationship

between recreation providers and consumers

These elements also impact public providers of
recreation. The threat of legal liability can also
influence the formulation of management objectives and
the evolution of recreation and policy (McAvoy, Dustin,
Rankin, & Frakt, 1989). For example, in developing a
site, a manager must consider long-range maintenance
plans as an incurred duty to protect the public. Another
element is the duty imposed by implementing admission
charges. Inn sum, the threat of legal liability impacts
all levels of operation in both private and public
operations.

Another dimension crucial to this problem has to do
with organizational ethics. Both common and statutory
laws are grounded in social values which are subject to
change through time (van der Smissen, 1990). In
negligence law, the ethics of relationships is manifest
in the concept of duty. The duty one owes to another to
protect them from unreasonable risk of injury is
ultimately an ethical consideration. The formal rules of
law attempt to codify this ethic in a structured format.

For purposes of this problem, the legal dimension of

ethics will be isolated from the sociocultural ethic
determining acceptable behavior.

The economic, legal, and political dimensions of
recreation negligence are based on what society considers
ethical. Thus the root of a recreation provider's duty
toward the public is primarily an ethical consideration.
Quite simply, providing' safe .1and conditions for
recreation is the right thing to do.

Much concern is directed toward settlement decisions
as the causal element of the so-called "liability
crisis." It is the decision-making process, however,
that determined these settlements, and thus is the
essential element in understanding recreation liability
problems. Until recently there has been insufficient
methodological basis for conducting research of this
decision-making process. Specifically the process that
needs evaluation is the step-by-step processes in which
legal components are sequentially examined by plaintiffs
and defendants' lawyers, as well as judges in the
settlement process.

In recreation liability, the realms of .nuisance,
strict liability in tort, trespass and negligence are
applicable. The most common of these torts is negligence
(Bronstein, 1985). In the broadest sense, negligence
needs to be evaluated in terms of the determination of

duty, breach of duty, proximate cause, and demonstration

of damages. Each of these elements are confounded by
myriad variations and rules.

The area of risk management has emerged as a
proactive means to address negligence issues. To
identify and remedy potentially hazardous situations not
only aids in reducing injury, but also provides evidence
that the organization may use to demonstrate due care
exists. Records of maintenance, inspections, and staff
credentials demonstrate active concern for the publics
served. Risk management not only has legal and economic
impact, but also moral and ethical impact (van der
Smissen, 1990).

Van der Smissen has extensively detailed the
principles of risk management breaking out topics as risk
management planning, operations control, and management
of financial risks. One common element in risk
management is the decision matrix that aids in assessing
the frequency and severity of risk potential as
illustrated in Figure 1. In risk management planning and
operations control, this tool is a valuable means of
suggesting the approach to control. The decision matrix
is a practical tool based on a qualitative assessment of
how the legal system will react to injuries. Thus it is
a reflection of probable legal outcomes with little

consideration to the legal decision—making process.

 

 

 

 

 

 

 

 

 

High Medium Low
S
E High Avoid or Transfer Transfer
V transfer
E
R Medium Transfer Transfer Transfer
I or Retain or Retain
T
Y Low Retain Retain Retain

FREQUENCY

Fig. 1. Risk management decision matrix.

Source: van der Smissen, 1990.

There is currently no reliable means for
individuals, operators, and agencies to evaluate real and
hypothetical potential for legal liability without
entering the decision-making process in the civil law
arena. The construction of a knowledge based system in
this domain would provide a means to simulate this
process and, therefore, assist in limiting liability and
associated costs.

The domain of recreation liability decision making
is both complex and narrow. Thus it is well suited for

the imposition of a knowledge based system to evaluate

the elements and factors impacting the decision-making
process. It is currently unclear to recreation scholars,
administrators, and operators as to which elements of
liability law and litigation have, and will impact the
operation of a recreation enterprise. The basic modeling
of this system would provide a laboratory for:
1. evaluating causal relationships in tort law
2. determining responsibilities of both consumers
and providers of recreation
3. predicting how hypothetical changes in law may
impact the role of operators, individuals, and
agencies
Certainly the discussion of "costs" of recreation
liability is important to assess the current operating
environment of recreation operators. The underlying
problem, however, has not been researched in part due to
its being shrouded in nebulous legal idioms and
processes. This lack of understanding has resulted in
the promotion of "tort reform" to artificially restrain
the limits of damages and responsibilities via
legislative mandate. This is a top-down approach which
ignores the process in which plaintiffs and defendants
arrive in court and determine settlements. Basic
research is needed to model this process so that it may
be understood in empirical terms. Without this empirical

knowledge, recreation leaders and operators are forced to

deal with the problem on a symptomatic level without
understanding the ontogeny of the liability tort process.
Many issues of fact remain unresolved by lack of research
(Twardizik, 1985), and finding a means of controlling
liability costs is a challenge that must be addressed
(Moline, 1985).

A basic model for the legal elements in recreation
liability is a first step in understanding the nature of
the problem. Specifically, the problem to be addressed
is the lack of a formalized methodology to evaluate the

 

dynamic elements of recreation liability within the legal

 

arena .

Potential Uses of the Study
The development of a knowledge based system based on
the domain of recreation liability would provide a means
to:

1. Identify and structure legal rules and doctrines
that determine the inferential processes of
litigators

2. Identify elements in the recreation environment
that may be controlled to reduce potential
liability

3. Model how changes in legal doctrines may change
responsibilities of both providers and consumers

of recreation

4. Simulate legal decision making enabling out-of-
court experience in portions of the litigation

process

Objectives

The objectives of this study are as follows:

1. To provide a systematic means of evaluating
legal decision making in recreation liability
cases. To understand the problem-solving task

2. To construct and evaluate a knowledge based
system that is theoretically and practically
valid

3. to create extensive explanation facilities in
the system to enhance understanding of the

problem—solving task by the end users

The Research Document

Chapter II is a literature review that begins with a
survey of expert systems research, in generals Here
pioneering systems from ‘various domains are discussed
with respect to the problem at hand. The review then
focuses upon the application of expert system methods in
the field of natural resources and then recreation.
Finally applications and research in the legal domain are
reviewed to narrow the focus to areas of specific

contribution to this study.

10

Chapter III provides a broad overview of the expert
systems approach. to problem-solving tasks. Here the
history and basic elements of expert systems are
discussed to assist the reader in understanding the
vocabulary and conceptual framework of the method. A
brief survey of various knowledge representations is
included to familiarize the reader with concepts of
knowledge engineering.

Chapter IV is a description of' methods used to
assess a problem for the expert system approach, as well
as methods for building the specific system for this
research. This is the methods chapter.

The next chapter deals with the evolution for the
expert system built for this study (called TOTO). There
is a description of the model structure and how the
search strategy brings the system to life. There is also
a discussion on how difficult problems were approached.

Chapter VI describes the results of TOTO's
performance on test cases. The final chapter summarizes
the study and discusses the results. Study limitations

and avenues for further research follow.

Definition of Terms
Artificial Intelligence Glossary
The following glossary is based in part on Waterman

(1986).

ll

Antecedent/Consequent: An if—then rule form where

 

the antecedent is a list of preconditions to the meeting
of a conclusion.

Artificial intelligence: A field of computer

 

science focusing on creating intelligent computer
programs.

Confidence factor: A numerical value assigned to a

 

fact by the system user to rank confidence to a query.

Domain expert: A person who embodies problem—

 

solving expertise in a domain.

Expert system: A subfield of artificial

 

intelligence that strives to create computer programs
that emulate expertise of highly skilled humans.

Fuzzy logic: A means to approximately quantity

 

concepts that are of imprecise meaning.

Heuristic: A rule of thumb used by domain experts

 

in fields where strict procedures are absent or
unreliable for problem solving.

Inference engine: The component of an expert system
that accesses the problem-solving knowledge and processes
the knowledge base to solve problems.

Knowledge base: The component of an expert system

 

that contains domain knowledge.
Lisp: A symbolic programming language favored in
the United States.

Metarule: A rule that controls other rules.

 

12

Object inheritance. The means in which object

 

oriented programs utilize to describe an object in
relationship to attributes shared with others.

Production rule: Rules in the knowledge base that

 

are made up of IF, AND, OD, OR conditions followed by a
THEN action.

Proloq: A symbolic programming language favored in
Asia and Europe but gaining popularity in the U.S.

Pruning: The expert system process of narrowing the
problem—solving task by eliminating branches of
knowledge.

Search: Looking through the range of possible
solutions to a problem in the attempt to solve it.

Search space: All possible solutions to a problem

 

often narrowed by pruning.

LITERATURE REVIEW

The development of knowledge based systems are a
means to extract thorough and detailed information on
complex expert decision-making processes with a means to
systematically explore and organize that information
(Waterman & Peterson, 1985). Current "expert systems"
include implementation in the domains of medical
diagnosis (Buchanan & Shortliffle, 1984; Clancey &
Shortliffe, 1984), mineral exploration (Duda et a1.,
1979), tactical warfare (Klahr et a1., 1985). Other
applications include the areas of agriculture, chemistry,
computer systems, electronics, engineering, as well as
manufacturing and space technology (Waterman, 1986).

Expert system methods have found wide application in
the area of natural resource management with a fbcus on
agricultural-related and forestry-related problems.
Davis and Clark (1989) and Rauscher and Hacker (1989)
have compiled extensive bibliographies of expert system
applications for natural resource management problems.
The survey of Rauscher and Hacker (1989) demonstrates the
breadth of applications in natural resource management on

breaks down domain types as follows:

13

l4

Silviculture/Growth Yield 27%
Fire Management 18%
Pest Management 16%
Soil/Site/Environment 14%
Land Management Planning 12%
Harvesting/Products 7%
Others 6%
(N - 203)

The category of "others" includes recreation where
one system under development is reported. This system
developed at Virginia Polytechnical Institute and State
University is being developed to predict scenic quality
impacts by comparing the relative effectiveness of
statistical models in combination with expert system
methods (Rauscher & Hacker, 1989).

Harris and Swanson (1990) used knowledge base
programming for a resource economics problem involving
the contingent valuation method (CVM) for estimating
benefits of resource use. The goal of this research is
to "increase the validity and reliability of value
estimates obtained with the CVM by providing people with
the information and structure (i.e., a policy referendum
context) necessary for sound CV decision making" (Harris
& Swanson, 1990). Essentially, the program acts as a
counselor to aid the user in completing a CVM, on screen

questionnaire. Cognitive aids (such as graphics) provide

15

immediate feedback to the user on the consequences of
their value judgments. The user may then revise their
judgments based on this feedback, thus increasing 'the
validity of estimating benefits where there is cognitive
complexity in a decision-making process. The value
judgments of the user are subsequently recorded for later
analysis.

RANGER is a prototype expert system developed for
the United States Department of Agriculture Forest
Service intended to aid in marketing efforts by giving
expert site selection advice and providing a means of
monitoring client characteristics (Forsberg, 1990). In
the Forest Service setting, recreational resources are
both extensive and diverse. Meeting the specific mix of
customer needs to improve satisfaction is problematical
because the expertise needed to direct clients to sites
is needed in many locations and is scarce. Forest and
district level personal have this expertise yet their
skills are focused upon operational elements of resource
management. Also the peak recreational season coincides
with the fire season so experts are often unavailable for
consultation. The Recreational Opportunity Guides (ROG)
inventories are extensive, yet inefficient or impossible,
to use by the public. Most recreation inquiries are

handled by receptionists who cannot efficiently sort

16

through ROG to arrive at an optimal site to meet the
individual's needs.

RANGER is designed to be operated by untrained staff
members or the public to match client needs to forest
resources. The problem-solving task is diagnostic in
nature. The consultation session consists of a series of
queries that refine the profile of the user to reflect
scenic preferences, facility needs, desired activities,
and specific needs such as target fish species. The
system subsequently searches the knowledge base
(partially based on the ROG catalogue) to match client
needs and recommend a site. Figure 2 exemplifies a
portion of the search space implemented in RANGER.

The knowledge base is not entirely passive to client
needs. The system can be modified to act in a management
model to place users in sites that met management
objectives. For example, certain sites can be "marketed"
while others may be "demarketed." Thus use can be
spatially concentrated or dispersed based on management
objectives. User profiles also may be used to shift use
type. For example, many wilderness users may be more
satisfied with simiprimitive resources. RANGER may be
modified to identify "fence sitters" and direct them into
underutilized sites.

The record of customer profiles provide a market

research data base that can be evaluated to assist in

17

Figure 2. Generalized Decision Tree in RANGER.

18

       
    
    
   
    
 

Experience um
I y pl P uquv n

Fleimlne

  

Command
Selection

Activitie-

Location

19

marketing and planning decision making. Profiles are
written into a database which then may be assessed with
conventional research methodologies.

One of the early attempts to apply computer
technology to the legal domain was in the area of
jurimetrics (Gardner, 1987). The basis of jurimetrics
was the statistical prediction of judicial behavior
without consideration of the decision-making process.
Thus it is based on behaviorism rather than cognitive
science.

As artificial intelligence: methods emerged, legal
knowledge based systems arose as research tools in both
theoretical and practical realms. Systems developed for
practical tasks include drafting legal documents (Sprowl,
1979), drafting statutes (Allen, 1980), drafting estate
plans (Michaelson, 1984) and information. retrieval
(Hafner, 1981).

JUDITH (Popp & Schlink, 1975) was an early attempt
to explore legal reasoning in civil law cases. JUDITH
was patterned after MYCIN, a pioneering system that
attempts to diagnose and treat, meningitis infections.
JUDITH, however, embodies a knowledge base on the
systematic taxonomy of statutory and civil codes. Case
law was incorporated into the knowledge base of TAXMAN
(McCarty, 1977) to augment general rules of law. A case

under scrutiny is described in terms of its divergence

20

from prototypical cases in memory, via a prototype and
deformation model. This model is also capable of
representing amorphous concepts or open texture
predicates. The plethora of amorphous concepts in legal
idioms has been consistently troublesome in representing
legal knowledge.

Since civil law is based on case precedents, case
based inference has been explored as a methodology. The
use of case based inference exclusively has been
implemented for predicting judicial behavior without
consideration of legal rules and decision—making
processes. The information processing task is
determining a mathematical or statistical function that
"fits" a case to those in memory. It is unclear if this
approach can be considered in the realm of artificial
intelligence (Gardner, 1987).

The approach of HYPO (Rissland, 1985) avoids the
pitfalls of predictive case based programs. HYPO
utilizes hypothetical cases used in law teaching to
evaluate comparatively a problem case to uncover legal
arguments and case strength. A case under consideration
is not static; the program modifies it to strengthen
elements that may be evaluated in a trail.

Theoretically, case based inference could expand a

HYPO type system by evaluating consultations and learning

from them. Dolodner (1987) suggests that problem-solving

 

21

capabilities may be enhanced by case based inference as
in the following:
1. recalling previous mistakes and avoiding them

2. precious decisions can be recalled to limit
search space

3. cases can reveal abstract scheme or that can
augment general knowledge (Dolodner, 1987).

HYPO has subsequently been modified and refined to
utilize actual cases in evaluating trade secret disputes
(Ashley & Rissland, 1988).

Meldman's LEGAL ANALYSIS SYSTEM (1975) utilizes both
general rules and case information. These two elements
are considered secondary and primary sources of authority
in decision making. General doctrines are represented as
general rules. Rather than matching a case against a
body of similar cases, Meldman represented significant
cases as specific rules. Key facts of these cases are
utilized for explanation of the rules and are not used
for inference. By utilizing the holding of a case for
the basis of a rule, oversimplification occurs. This
method ignores the secondary characteristics of a case
that are important justifications for a holding (Gardner,
1987).

LDS (Waterman & Peterson, 1984) evaluates product
liability by representing the domain as a body of rules.
The weakness of the system is common to many in this

domain, representing open texture predicates or amorphous

22

concepts, such as "ordinary care" or "foreseeability."
The authors suggest this problem can be reduced by the
following:

1. Provide rules that describes how an imprecise
word was used previously in particular contexts

2. Display a brief description of prior use of the
term and let the user decide whether or not the
term applies in the current instance

3. Ask a series of questions to elicit information
about the specific case in which the imprecise
term is at issue, compare the answers with prior
cases in which the term applied, and provide a
numeric rating that indicates the degree of
certainty that the rule refining the term
applies presently.

4. Use a system of gradual refinement by query to
determine whether or not a term applied
(Waterman & Peterson, 1984).

Gardner (1983, 1987) has developed a program that
evaluates contract law via four levels of processing.
The first level embodies knowledge of the ordering of
basic legal categories. The second level contains
knowledge of the definitions of the major concepts (as
if/then rules). The third level of processing contains
knowledge about undefined predicates (open texture) in
the form of rules and examples. The examples are real
and hypothetical cases which are matched to the problem
case to fill in gaps discovered by their processing
levels. They act as prototype cases to provide meaning
individually, or in combination, to an open texture

predicate in the run time situation. The fourth, and yet

23

undeveloped, level of processing will be an attempt to
allow and resolve disagreement among experts. Competing
examples will be utilized as the basis for this
information processing task. 0

Of particular relevance to this project is the
knowledge representation of The Latent Damage System
(LDS) by Richard E. Susskind of Ernst and Whinney, UK.
The Latent Damage System deals with assessing temporal
issues associated with tort, contract, and product
liability law. The system was implemented in the rule
based Crystal expert system development tool. The
Boolean logic and knowledge structure of the Latent
Damage System is similar to that proposed in the
development of TOTO. This structure consists of rules
that are "reverse compound conditional statements"
(Susskind, 1989) that have antecedent and consequent
clauses. For example, in the rule (IF A and B OR C THEN
D), the antecedents are A, B, and C, whereas the
consequent is D.

This simple rule structure becomes increasingly
complex as consequents become antecedents of other rules.
The structure can be represented as an inverted decision
tree where conditions must be met to follow a particular
branch. The search space is gradually narrowed to focus
upon branches that contain specialized knowledge

applicable to a particular case.

24

The major problem associated with LDS is that human
judgment is required by the user during the consultation
process. Although domain expertise is not required by
the user, a degree of logic and common sense is required
to interpret queries. This problem is not exclusive to
LDS. Within the legal domain concepts gain. meaning
within case context so that terms such as "reasonable"
have variable meaning (these are called "open texture
predicates").

Rissland and Shalak (1989) are developing a system
that combines traditional rule based reasoning with case
based reasoning (CBR) to interpret the meaning of open
texture predicates. CABARET initially assesses a case
with a rule based reasoner to collect information and
narrow the scope of cases to be considered. A control
rule activates the case based reasoner and relevant cases
are brought under consideration. Various case facts are
compared with the problem under consideration to support
or reject antecedents. The CBR attempts to match facts
of cases in memory to those in the current case . A
match may or may not be found. A near miss initiates a
search for other cases to satisfy the unsubstantiated
conditions. Thus open texture predicates are assigned
meaning based on actual case conditions in memory.

Another potential method of dealing with open

texture predicates in the use of fuzzy numbers. Coughlan

25

and Running (1989) utilized fuzzy numbers to quantify
symbolic knowledge for a geographical information expert
system (GIS). A GIS data base consists of interval data,
whereas rules are comprised of symbols. In order to link
symbolic meaning to mathematical meaning, fuzzy numbers
are assigned to concepts. The problem is summarized as
follows:

If a rule antecedent contains IF ELEVATION IS LOW,

how do we numerically define LOW so that we can link

it to the data base? Also, what are the boundaries
between HIGH and LOW and how are they determined?

First, definitions area context sensitive. .Assume

that HIGH and LOW are contained within the elevation

ranges of our study area, 1200 to 3000 m, with 3000

m being a perfect HIGH and 1200 M a perfect LOW.

Secondly, inexactness between rule and GIS data can

be expressed by a fuzzy function which translates

the degree to which a particular GIS variable is
represented by a discrete symbolic variable

(Coughlan & Running, 1989).

Mapping fuzzy numbers to concepts is a subjective
process which posses a problem for application in the
legal domain. Cases are not decided on a degree of truth
and thus the concept of partial, membership is often
inapplicable (Gardner, 1987). Fuzzy numbers, however,
may be utilized to map the degree of membership of an
open texture predicate to the meaning present in the
knowledge base. The process remains subjective, yet can
be successful if guided by domain expertise. For
example, the term "reasonable" can gain specific meaning
by balancing the elements of the magnitude of risk with

those of the burden of alternative conduct. The set of

26

variables associated with this balance could be assigned
fuzzy numbers. The balancing fuzzy function would then
map the domain definition of "reasonable" to that derived
in the session context to reveal a degree of membership.
The evaluation of current knowledge based systems
reveals several common themes. They include:
1. Problems associated with bringing meaning to
open texture predicates
2. The potential for processing problems at various
levels of knowledge representation
3. The potential for utilizing examples (i.e.,
cases, hypothetical or real) at deeper
processing levels
4. The allowance for disagreement among experts by
utilizing competing examples
5. The lack of explanation facilities as a system

component

The Expert System Approach
History of Expert Systems
Since the development of computing machines, there
has been an expectation that someday they will be able to
reason and make judgments like humans. Hardware and
software advanced in the early 1970's created a means to
approach the problem of making computers "think." The

area of artificial intelligence (AI) emerged in

27

universities and corporations as a discipline including
computer scientists, engineers, and psychologists. By
the 1980's expert systems emerged to handle practical
problem-solving tasks. Initially work focused on medical
diagnostic tasks where extensive bodies of expertise were
encased in systems, such as MYCIN. MYCIN was
specifically designed to diagnose meningitis infections
and recommend microbial therapy (Buchanan s. Shortliffe,
1984). Soon expert system applications began to cross

disciplinary lines to approach, problems classified as

follows:

Interpretation Inferring situation descriptions
from sensor data

Prediction Inferring likely consequences of
given situations

Diagnosis Inferring system malfunctions from
observable data

Design Configuring objects under con-
straints

Planning Designing actions

Monitoring Comparing observations to expected
outcomes

Debugging Prescribing remedies for mal-
functions

Repair Executing plans to administer
prescribed remedies

Instruction Diagnosing, debugging, and repair-
ing student behavior

Control Governing overall system behavior

(Hays-Roth et al., 1985).

28

Expert systems were initially developed to handle
narrowly defined problem-solving tasks. The knowledge
bases of these programs were isolated from the inference
mechanism resulting in generic system building tools
called shells. Expert system shells enable the
programmer to focus on developing knowledge structure
without actually programming inference strategies. Thus
expert systems can more readily be applied to problem-
solving tasks within the realities of time and fiscal

constraints.

Expert System Components

 

In their simplest form, expert systems are comprised
of a knowledge base and an inference engine. The
knowledge base is the symbolic representation of
expertise in a given area, and the inference engine is
the control or search strategy that brings the knowledge
base to life.

A knowledge base is the symbolic representation of
expertise gleaned from domain experts. It often contains
rules, facts, attributes, and rules of thumb (heuristics)
that represent the proficiency of expert problem—solving
behavior. The symbolic representation of this expertise
may be modeled as semantic networks, frames, and

production rules or objects.

29

Semantic nets represent knowledge as a network of
nodes (conceptS) linked to each other by relationship
describing archs. Frame representations are collections
of concepts described by another collection of attributes
called slots. Both semantic net and frame
representations are .hierarchical structures with. lower
nodes inherenting attributes of higher level nodes. Both
of these methods are, therefore, especially suited for
representing taxonomies in natural systems.

The production rule is a simple knowledge
representation where a decision tree structure is
comprised of condition and action statements (Waterman,
1986). Typically, rules are based on IF, AND, OR, THEN,
and ELSE statements, followed by clauses. Fer example,
the following production rules determine an appropriate
procedure of statistical analysis:

Rule 1

1. IF the criterion variable is scaled intervally

2. OR the criterion variable is scaled ratio

3. AND there is one criterion variable

4. AND there is more than one predictor variables

5. AND the predictor variable is scaled nominally

6. THEN implement ANOVA

7. AND procedure found

This simple structure contains conditions that are

possibly unclear. To make the knowledge base usable,

 

30

these conditions may be represented in more rules.
Condition (1) may be satisfied by firing another rule.

Rule 2

1. IF numbers used to rank items are numerically

equidistant

2. AND the zero point and measurement ends are

arbitrary

3. THEN the criterion variable is scaled intervally

In order to satisfy the first condition of Rule 1,
Rule 2 must first be satisfied. Each condition in each
rule may require assessment of additional rules for the
rule to "fire." A rule fires when all conditions are met
according to Boolean logic. Thus the simple production
rule representation becomes an increasingly complex set
of interdependent nodes. Typically a knowledge base will
contain from 50 to 500 rules.

The inference engine is the control scheme to
manipulate the knowledge into producing conclusions.
Essentially, it is the problem-solving knowledge used to
search the knowledge base. A backward changing inference
strategy is initiated by setting a goal for the engine to
meet by manipulating the knowledge base. For example if
the GOAL "procedure justified" is stated prior to the
example knowledge base, a backward search is implemented
by the inference engine and the user is queried to meet

conditions until the goal is reached. In many bodies of

31

knowledge, facts cannot be represented with. a simple
"yes" or "no" queries. In other words, facts are not
always true or false in expert problem-solving behavior.
Thus certainly factors are often assigned to conditions
of rules. Certainty factors are more or less arbitrary
values associated with facts. In the preceding rule
base, for example, the first conditions of Rule 2 (IF
numbers used to rank items are numerical equidistant) may
be interpreted differently by a social scientist and a
mathematician. The definition of "equidistant" is to
some degree viable. To handle this variance, a certainty
factor could be assigned by the user to assign a degree
of truth to the condition. The user could be asked by
the system to rank his/her confidence in the truth of the
condition on scale of 0 to 100 or -100 to 100. The
addition of certainty factors to rules adds a basis of
"fuzzy" thinking often utilized by experts. It is also
clear that when a knowledge base has hundreds of rules,
these certainty factors add a confounding element in the
accurate representation of domain expertise.

A knowledge representation that is gaining
prominence is object oriented programming. _Objects are
entities that are descriptions of chunks of knowledge
that contain data, attributes, values, and procedures.
Unlike conventional programming' structures, objects

contain both data and procedures making the approach

32

suitable for knowledge based programming (Richer, 1889)
.A group of similar objects comprises a class, which is,
in turn, a member of a metaclass. Class variables and
methods are inherited to the superclass forming a latus
knowledge structure. Object inheritance simplifies the
definition of concepts and is illustrated in Figure 3.
Object oriented programs are active taxonomies of domain
knowledge where objects send messages between one another
to perform the problem—solving task. Restructuring our
simple rule based representation into classes and objects

results in the following (truncated) example.

Metaclass statistical methods

Class multivariate methods

variables: measurement scales
distribution
variance

predictor variables
criterion variables

Method perform procedure
Sub class ANOVA
variables: criterion variables—-1

predictor variables—-1+
measurement scales criterion--interval+
measurement scales predictor-—nominal

Method perform procedure ANOVA
display results

Object interval measurement scale

Obj. variables zero and end—-arbitrary

number ranks--equidistant

Neural Networks

 

A method that has been gaining prominence in the

realm of artificial intelligence is that of neural

Luxury goods

Fragile goods \

' eggs
\ hotdogs
’

‘ marshmallows

I O \
grocenes A’\\

equipment

camcorder

lantern
rv

tent

. rain gear
clothing < warm clothes

Camping goods

 

gasoline

Figure 3. An Illustration of Object Inheritance.

34

networks. Neural networks are hardware or software
simulations of neurological models made up of many highly
interconnected processing elements called neurodes
(Trelease, 1988) Neural networks do not execute like
conventional programs, but rather, react to inputs.
Neural networks learn by self-organizations Training
paradigms are specified by the programmer to handle
various learning tasks.

The network in Figure 4 illustrates the architecture
of a simplified "hidden layer" configuration. The input
layer represents the data which the hidden layer will
respond to. The output layer is the function or vector
that is gleaned from the data set. The network responds
to the data set stimulus and self-organizes into weighted
neurodes (Caudill, 1987). An activation function then
determined the level of excitement associated with the
neurode. In the basic "backpropogation" training
paradigm, the error between the desired and actual output
(in the least squares sense) is cycled back to adjust the
connections between neurodes (Josin, 1988). The data set
is iterated between layers of the network eventually
reaching equilibrium. This state is then frozen to
represent a state of "knowledge." Learning paradigms are
inherently statistically based (White, 1989). As a new
statistical modeling technique it is unclear how neural

networks will impact recreation research. Potential

35

 

Output Layer

Hidden Layer

 

Input Layer

Architecture of a Hidden Layer Neural

Network.

Figure 4.

36

applications include networks that. provide alternative
means of approaching forecasting and modeling problems.
Neural networks also show promise in adding in the
development of adaptive expert systems that respond to
large dynamic data sets.

The expert system approach has been utilized for a
variety of problem—solving tasks. There are a variety of
ways to represent knowledge and it is the job of the
system developer to select the one which best suits
characteristics of the domain. As this technology
progresses, traditional knowledge representations will be
refined while new ones are developed to handle domain-

specific tasks.

THE STUDY

Expert System Applicability

The question of the applicability of a system's
approach is similar to that of assessing the range of
traditional research methods. Traditional methods
include experimental survey and qualitative designs. The
expert system's approach to a problem embodies elements
of each of these designs and can be basic or applied in
nature. Basic research in expert systems involves the
area of cognitive psychology where the focus is on
discovering elements of expert cognition in problem
solving. Fraponents of this approach insist that until
we understand human cognition, one cannot successfully
develop valid expert systems. The applied approach is
not directly concerned with cognitive processes, but
rather, it is results-oriented and pursues developing
working systems by utilizing existing technology. This
study is more applied than theoretical in nature.

Expert System Suitability
Not all problem-solving tasks are suitable for
application of the expert system approach. The domain
characteristics necessary for system development are

illustrated in Figure 5 (based on Waterman, 1986).

37

38

 

Task does not
require
common sense

 

 

 

Task requires
only cognitive
skills

 

 

Experts can
articulate their
methods

      
    

 

(
Expert System

 

 

Genuine
expertise
exists

Development

 

 

 

 

  
 
 

Possible

.4
'i. .‘513511554' 56 h'. .1 . ...-.¢‘- mesmer- $fi$$$$$2¢$$55fi -"-'-" '

 

Experts
a ree on
so uiions

 

 

 

Task is
not too
difficult

 

 

 

Task is not
poorly
understood

 

 

Figure 5. Domain Characteristics Necessary for
Expert System DevelOpment.

39

The domain of legal decision making is constructed
of a body of common sense reasoning. Civil law is a
dynamic, precedent-based process that follows legal
principles developed over many years. The domain is not
too difficult, nor is it poorly understood and thus is
moderately well structured. Expertise in recreation
negligence law exists and experts can articulate the
methods of legal decision making. A problem arises,
however, in that experts do not always articulate their
actual cognitive processes and rely upon expected or
learned methods when reporting them. The process of
knowledge engineering is the methodology of extracting
heuristics from experts and other reliable sources.

The fifth characteristic (in Figure 5) of experts
agreeing upon solutions is also problematical. In the
civil law arena, legal decisions are ultimately decided
by jury. There is great variability on how any jury
would respond to any case. This final decision, however,
is not the focus of the current problem-solving task.
The focus here is upon the decision-making process to
determine negligence as a prerequisite to a civil case.
Here experts in general agree upon solutions where the
decision-making process is formalized.

In summary, at this preliminary stage, this domain
meets the basic requirements for expert system

development as stated by Waterman (1986).

40

Expert System Justification

Figure 6 lists five characteristics which Waterman
(1986) suggests for justifying expert system development.
Providing a. means for recreation operators to assess
potential legal liability outside of court has a high
payoff for both operators and users. Operators with such
a system may identify elements in their environments that
contribute or diminish liability and thus offering a
means to reduce the potential for injury. Recreators
would receive a payoff in safe conditions in the
recreational environment (ethical economic and political
benefits). In both cases, the payoff is great in terms
of reducing personal injury and subsequent settlement and
insurance costs.

This domain satisfies both characteristics three and
four as well. Michigan has a great diversity of
recreation opportunity provided by both the public and
private sectors. If each operator had a legal consultant
to assess liability issues, there could be no
justification to construct an expert system. This is not
the case, however, and it is clear that expertise is

scarce and needed in many locations.

Expert System Appropriateness
The expert system approach is appropriate if it

fulfills certain intrinsic qualities (Figure 7)

41

   
 
   

Task solution
has 0 hi h

payo

 

 

..................
............................................

   

Human
expertise
being lost

 

 

   

 

expertise
is scarce

 

. ._ Development

 

J

 

 

 

 

 

Figure 6. Factors Justifying Expert System Development.

42

  

Task Requires
Symbol
Manipulation

    
   
 
 

Task Requires
heuristic
solutions

    
   
 
 
  
 

Expert

    
  

System

T k' t

93059383, AND Approach
Appropriate

 
 

Task has
practical
value

    
      

Task is of
manageable
size

    

Figure 7. Characteristics Making Expert System
Development Appropriate.

 

43

(Waterman, 1986). Legal knowledge and decision making is
based on manipulation of symbols. Concepts are
represented by a string of characters of words and are
combined with rules of thumb or heuristics to form the
problem-solving process.

Characteristics three and four are complementary in
that if the task were easy, there would be little
practicality in developing an expert system to handle it.
It is clear that legal decision making in the recreation
environment is not an easy task. The practicality of the
task is readily demonstrated by reviewing the costs of
physical injury and settlement processes.

In order for an expert system to be developed and
remain valid, it must be of manageable size. In this
case, the problem-solving task is limited to assessing
recreation liability for any landholder in the State of
Michigan, excluding liability incurred by recreation
programming. It is also limited in depth to the
decision-making processes prior to actual trial. These
constraints are necessary to delimit the site of the
problem so that it may be addressed with an expert

system's approach.

Tool Selection

 

There are many tools available to assist one in
developing expert systems. High level programming

languages, such as Lisp and Prolog are used to develop

44

systems from the ground up. Currently, however, there
are packaged tools available for knowledge engineering
called "shells." Shells offer a variety of knowledge
representations and.jproblem-solving‘ control strategies.
Other components of shells include various types of user
interfaces, explanation facilities, ability to access
other programs, and certainty factors.

The most important element in tool selection is the
matching of domain characteristics to a particular
knowledge representation. Here the system developer
should have a firm grasp of the range of representations
and a conceptual model of domain fundamentals. The
knowledge engineering environment of the tool should be
able to provide explanations of queries. The shell also
should embody a means to process fuzzy knowledge or
handle degrees of certainty in answering individual

queries to the user.

System Building

 

There are a variety of methodologies for developing
expert systems. Weilinga and Bredeweg (1989) classify
these methodologies into those that involve rapid
prototyping, software engineering, or life cycle models.
The life cycle modeling approach that is most broadly
recognized is that of Hayes-Roth et a1. (1983). This

approach is illustrated in Figure 8.

45

IDENTIFICATION

qui
Identify Problem 9‘0 ,9

Characteristics (8 T3
53’

CONCIJ’TUALIZAIION

Find Concepts To 0000‘,
Represent Knowledge (3 -%

 

 

FORMALIZATION

Design Structure tru
To Organize S 0"

Knowledge Q :3

IMPLEMENTATION

 

 

 

 

’90, Formulate Rules R
04,) To Embody We
(’49,,- Knowledge Q) m
o
05. ,9
90'
e
e. .
0s Validate Rules
,9 That Or anize
04' Know edge
08 .
4%
6

Figure 8. The Hays-Roth et a1. Life Cycle Model for
Expert System Development.

46

The identification phase can be summarized by the
steps of identifying the participants, the problem, the
resources, and the goals of the system. The concepts
discovered in the identification phase will be refined
and embellished to provide a means to diagram the task
with relationships made explicit. This is the
conceptualization phase which may involve domain experts,
written materials, and other reliable sources of
knowledge.

Generally, the formalization phase involves imposing
the conceptualizations and relationships discovered in
the conceptualization phase onto the specific knowledge
representation and control structure provided by the
development tool shell. Specifically, this step involves
determining the hypothesis space including developing
specific hypothesis for the problem-solving task, and
' determining the granularity of concepts and structure.
Granularity refers to the size or level of detail of
elements to form meaningful aggregates or "chunks" of
knowledge.

Another element in the formalization process
includes determining the underlying behavioral model that
will impose logic upon concepts and relationships. Also,
one must determine a means to deal with uncertainty in

the model, and identify hard and soft data. Hard data

47

include reliable prima facie elements, whereas soft data
refer to less reliable, nebulous concepts.

The implementation phase is the actual programming
of the system in the programming environment. Here a
prototype of the system is developed based on information
gained from the previous phases.

Testing of the prototype consists of consulting the
system to discover weaknesses in its problem-solving
behavior. This not only includes testing the
accurateness of diagnostics, but also includes reviewing
the representatives and clarity of queries to the user to
evaluate if questions are answered in the intended way.
The testing process again involves domain experts to aid
in evaluating the validity of conclusions drawn from
specific case elements. Testing leads to revision of the
system to improve its performance.

It must be emphasized that each phase in
constructing the expert system creates a feedback loop to
earlier phases to refine the model. This evolutionary
process is essential in. maintaining proper focus and

direction in approaching the problem-solving task.

Knowledge Acquisition

 

The sources for domain knowledge for this project
include legal textbooks, legislation, recreation

negligence cases, and domain experts. The framework of

48

the model is based upon rules of law that pertain to
negligence torts in general. Pertinent public acts were
utilized to bring the general rules of law into the
reality of the legal environment of Michigan. Case law
was utilized to bring the model into the current status
in the civil arena. Domain experts were used as
consultants to validify and reject model components.

The knowledge acquisition and system building
methodology generally follows the life cycle model of
Hays-Roth et a1. (1983) as in Figure 8. The problem of
recreation negligence is broad and initially had to be
narrowed into a manageable size. The problem was
narrowed by focusing upon the legal environment of
Michigan , and specifically , premises liability .
Eliminating other types of negligence, such as personal
and programming liability made the problem of manageable
size . In the identification phase , problem
characteristics were identified which, in turn, provided
a basis for disclosing requirements for the
conceptualization phase. Problem conceptualization
included finding the key concepts to represent knowledge.
Here the basic legal and recreation concepts arose
primarily from secondary authoritative written materials.

These concepts were then organized into a formalized
structure and the basic model called TOTO was created.

This process involved the creation of extensive graphic

“use

49

charts displaying relationships among the concepts.
Charting disclosed that the initial problem
conceptualization was again too broad. For example, it
was hoped that the system could embody extensive
knowledge of safety standards for various recreation
activities. Incorporating this knowledge would broaden
the problem to unmanageable size and also such standards
are not clearly defined for many activities. Thus a
generic standards model was proposed to assess this
element in the more general sense. The knowledge base
was also focused on issues pertaining to a limited period
in the legal decision-making process.

TOTO focuses on the evaluation process conducted by
attorneys to demonstrate that negligence exists. In
order for a case to go to trial, a judge must weigh
evidence to determine that there is grounds for
negligence. The elements of duty, causal connection, and
injury must be present for negligence to exist. TOTO is
based on these elements as well as an assessment. of
defenses available to the defendant. By limiting the
scope of the problem to pretrial phases of decision
making the knowledge base is narrowed to a manageable
size.

Formalization provided the basic structure the rule
base should follow. During implementation the rules were

written to embody the concepts and relationships in

50

recreation negligence decision making. The rules serve
many functions beyond representing the actual knowledge.
Rules also provide a means to gather factual information
from the user to initially classify the case at hand and
narrow the range of problems to be considered. Some
rules are classified as METARULES which provide guidance
over the problem-solving task. METARULES also
compartmentalize subproblems into individually executable
tasks. This structure simplifies problem solving by
searching' only relevant rule. sets 'thus increasing 'the
speed of computation during system consultation.

The formalization phase disclosed potential problems
with the model as conceptualized. A simple rule
structure was found adequate to handle straightforward
legal problems. Many processes in legal decision making,
however, are not readily structured into rule sets.
Specifically, the determination of "reasonable” and
"willful and wanton" conduct in assessing a case emerged
as a major difficulty. These open texture predicates
warranted a more sophisticated scheme to make them
workable in the knowledge base. The approach to this
problem will be discussed in detail in the next section.

IMPLEMENTATION

The Knowledge Base
Liability in recreation is based on civil law (as
opposed to criminal law), and specifically the law of
torts. Negligence torts are common law remedies in which
the plaintiff pursues a claim based on injuries sustained
by the lack of care of the defendant (Prosser, 1984).
The burden of proof in asserting negligence falls on the
plaintiff to demonstrate liability on the part of an
individual or public agency landowner. In order for a
suit to proceed, the plaintiff must allege that the basic
elements of negligence are present. Generally, the
elements in demonstrating negligence include:
1. There is duty of the. defendant owed to the
plaintiff to provide a standard of care
2. There is a breach of duty in applying the
standard of care
3. There is proximate cause linking the breach of
duty to an injury
4. There are actual damages to person or property
Assuming these elements are demonstrated, the

plaintiff incurs the burden of proof in the civil suit to

51

52

provide the claim based on evidence. The preponderance
of evidence falls in favor of the plaintiff if the facts,

more likely than not, support the assertion. of

 

negligence. This preponderance of evidence is contrasted
with the burden of proof required in criminal cases where
one must demonstrate fault beyond a reasonable doubt
(Kaiser, 1986).

Returning to the element of duty, there are various
levels of care owned to the relationships between
plaintiff and defendant. Historically, those

relationships have been defined as:

1. Invitee: A person is expressly invited onto a
premises and pays a fee to enter, or for an
activity therein. This may be considered a

business relationship.

2. Licensee: A person has access to a premises yet
does not pay to enter. This could be a person
hunting or skiing on one's property or simply an
individual entering a business to use a
telephone.

3. Trespasser: The direct physical, unauthorized
invasion of the exclusivity of one's property
rights (Shulman, 1984).

For each relationship, the landowner is liable for gross
negligence, and willfu1 wanton misconduct leading to
injury. Generally, this is the only liability a
trespasser can claim. An invitee is owed the highest
level of care where liability reaches beyond willful
wanton misconduct to the ordinary care of premises. The
status of licensee falls between the two extremes where

ordinary and gross negligence may be considered.

53

The person pursuing recreation activities on public
or private lands who has not reached the status of the
invitee has historically been able to sue for ordinary
negligence. This has had the effect of discouraging
recreation use of private lands. This effect has become
troublesome since there is considerable demand for
recreation resources beyond which, public entities can
supply (Holecek & Westfall, 1977). Recreation use
statutes were subsequently initiated to limit a
landowners' liability in allowing the public to use
private lands free of charge. In 1965 the Council of
State governments published a model recreation use
statute which. subsequently formed the basis for
legislation in forty-nine state jurisdictions (Kozlowski,
1986). The Michigan Statute reads as follows:

No cause of action shall arise for injuries to any

person who is on the lands of another without paying

to such other person a valuable consideration for
the purpose of fishing, hunting, trapping, camping,
hiking, sightseeing, motorcycling, snowmobiling or
any other outdoor recreational use, with or without
permission, against the owner, tenant, and lessee of
said premises unless injuries were caused by the
gross negligence or willful and wanton misconduct of
the owner, tenant, or leases (Mich. Stat. Ann. at

13.1485, 1987).

Recreation use statutes are intended to change the
level of negligence liability that a private landowner is
responsible for. This legislation changes the

landowner's standard of care from ordinary negligence to

gross negligence or willful and wanton misconduct. Thus

54

the landowner owes no duty to warn recreational users of
known or discovered hazards on one's premises The
desired effect of this action is to discourage negligence
suits by increasing the criteria in the burden of proof
in proceeding with a claim The enactment of recreation
use statutes are substantial means of limiting liability.
The Michigan recreation user statute creates a new user
category that adds to the categories of licence, invitee,
and trespasser. Since the statute indicates that it may
be invoked only if there has been no valuable
consideration paid for the recreation activity, the
category of invitee may still be considered. The
Michigan Recreational Trespass Act (Mich. Stat. Ann. at
13.1485) defines the recreational trespasser yet the
recreational use statute may be invoked whether or not
the plaintiff has permission to enter a premises. The
trespasser category may still be considered in the
assessment of gross negligence or willfu1 and wanton
misconduct. .As courts interpret these status, however,
diverse elements are introduced which create new issues

in the civil law arena.

Civil Law Issues
Public Entities
The concept of the recreation use statute has spread

in many jurisdictions to include public entities.

55

Currently 119 jurisdictions have found that recreation
use statutes extend to public entities (Kozlowski, 1987).
In Michigan this extension has added another defense to
negligence actions against the state. Traditionally, the
State Department of Natural Resources has sought
dismissal of claims by:

1. statute of limitations

2. general governmental immunity

3. public employee immunity (replaced by Tort

Reform Act, 1885)
4 lack of possession or control (leasing lands or
operations to concession) (Hughes, 1987)

Since the enactment of a recreation use statute (Mich.
Stat. Ann. at 13.1485) it has been held that state—owned
land is included under this legislation (McNeal v.
Department of Natural Resources, 140 Mich App 625).

Municipalities in Michigan are currently covered by
the recreation use statute. Williams v. City of Cadillac
(1985) held that municipalities were only liable for
willful and wanton misconduct which reversed the holding
of Anderson v. Brown Bros., Inc., (1975).

It is clear that at present, Michigan courts have
given liberal interpretation of the state's recreation
use statutes in extending coverage to a wide range of

individuals and vertical arrays of government.

56

In several states the protection offered by
recreation use statutes has extended to include federal
lands as well. In Michigan, this concept was tested in
Miller v. United States Department of Interior (1986).
The issues surrounding this case exhibit the major
elements in determining the scope of the recreational use
statute.

The plaintiff, Jerry Miller, was injured on land
within the Sleeping Bear National Lakeshore when he fell
or jumped from a rope swing on the banks of the Platte
River. There are no fees charged by the National Park
Service to enter the park. Miller and a friend were
canoeing on the Platte River on August 5, 1980, when they
stapped at an area often used for picnicking and
sunbathing. Across from this area was a rope swing that
had existed for several years. It was frequently used by
the public to swing out and drop into the river. Miller
had used this swing on several occasions, and on this day
made 15 to 20 jumps before the injury occurred. On this
last swing, Miller became unbalanced and entered the
water head first in a shallow area resulting in a
fractured neck and partial paralysis.

Miller argued that the state recreation use statute
did not apply in this case because it applies only to
licensees and not to invites such as park visitors.

Miller also contended that the statute was intended to

57

open up land for recreational use that would otherwise be
unavailable (unlike a National Park).

The National Park Service contended that under the
Federal Tort Claims Act, the United States is to be
treated as a individual citizen in such cases. The court
acknowledged this claim, and therefore, the recreation
use statute did apply. Under the statute the defendant
must demonstrate willful and wanton misconduct (or gross
negligence) as determined by the preponderance of
evidence. In Michigan the courts have determined that
willful and wanton misconduct is present when there is an
attempt to harm or indifference resulting in willingness
for harm to occur. The court in this case found no such
intent or willingness on the part of the National Park
Service and thus gross negligence was not found.

In Michigan we see the general extension of
recreation use statute to encompass public entities. In
other jurisdictions, such as California and Florida,
however, recreational use statutes are currently limited
to private landholders (Nelson v. City of Gridly s McPhee

v. Data County) (Kozlowski, 1987).

Rural and Urban Lands
The general rhetoric of recreation use statutes has
lead to cases surrounding the applicability of the

statutes to rural vs. urban lands. The initial intent of

58

many such statutes ‘was to jpromote activities such. as
hunting, hiking, and sight-seeing on private lands. Many
courts have determined that these statutes are applicable
only to undeveloped or in unimproved lands. In cases
involving this issue, focus is on determining the intent
of the specific verbage in the given statute. In many
jurisdictions recreation uses are enumerated and focus on
outdoor, rural activities. Courts in. Wisconsin, New
Jersey, and Louisiana have interpreted these enumerated
uses in the literal sense and have rejected other
nonlisted uses as inapplicable (Kozlowski, 1986).

In Michigan the statute delineates recreation
activities as "fishing, hunting, trapping, camping,
hiking, sightseeing, motorcycling, snowmobiling or any
other outdoor recreational ‘uses" (Michs Stat” .Anna at
13.1485). The courts initially interpreted this to
include both rural and urban lands (Yahrling v. Belle
Lake Ass'n., Inc., 1985). The Yarling case was
subsequentlyly reversed and consolidated with Wymer v.
Holmes (1987). Thus the current interpretation of the
recreation user statute does not include urban, suburban,
and subdivided lands or social invitees.

Excerpted Legal Concepts as Defined by
Black (1986)
Assumption of risk: The doctrine of assumption of

risk, also known as volenti non fit injuria, means

59

legally that a plaintiff may not recover for an injury to
which he assents, i.e., that a person may not recover for
an injury received when he voluntarily exposes himself to
a known and appreciated danger. The requirements for the
defense of volenti non fit injuria are that: (1) the
plaintiff has knowledge of facts constituting a dangerous
condition, (2) he knows the condition is dangerous, (3)
he appreciates the nature or extent of the danger, and
(4) he voluntarily exposes himself to the danger. An
exception may be applicable even though the above factors
have entered into a plaintiff's conduct if his actions
come within the rescue or humanitarian doctrine.

Burden of proof: (Lat. onus probandi.) In the law

 

of evidence, the necessity or duty of affirmatively
proving a fact or facts in dispute on an issue raised
between the parties in a cause. The obligation of a
party to establish by evidence a requisite degree of
belief concerning a fact in the mid of the trier of fact
or the court.

EEBEEJ rn (Lat. EEEEE-I Each separate antecedent
of an event. Something that precedes and brings about an
effect or a result. A reason for an action or condition.
A ground of a legal action. An agent that brings
something about. That which in some manner is
accountable for condition that brings about an effect or

that produces a cause for the resultant action or state.

60

Comparative negligence . Under comparative

 

negligence statutes or doctrines, negligence is measured
in terms of percentage, and any damages allowed shall be
diminished in proportion to amount of negligence
attributable to the person for whose injury, damage or
death recovery is sought. Many states have replaced
contributory negligence statutes or doctrines with
comparative negligence. Where negligence by both parties
is concurrent and contributes to injury, recovery is not
barred under such doctrine, but plaintiff's damages are
diminished proportionately provided his fault is less
than defendant's, and that, by exercise of ordinary care,
he could not have avoided consequences of defendant's
negligence after it was or should have been apparent.

9313:. A human action which is exactly conformable
to the laws which require us to obey them. Legal or
moral obligation. Obligatory conduct or service.
Mandatory obligation to perform.

In negligence cases term may be defined as
obligation, to which law will gives recognition and
effect, to conform to particular standard of conduct
toward another. The word "duty" is used throughout the
Restatement of Torts to denote the fact that the actor is
required to conduct himself in a particular manner at the
risk that if he does not do so he becomes subject to

liability to another to whom the duty is owed for any

61

injury sustained by such other, of which that actor's
conduct is a legal case.

Due care. Just, proper, and sufficient care, so far
as the circumstances demand it; the absence of
negligence. That care which an ordinarily prudent person
would have exercised under the same or similar
circumstances. "Due care" is care proportioned to any
given situation, its surroundings, peculiarities, and
hazards. It may and often does require extraordinary
care. "Due care," "reasonable care," and "ordinary care"
are often used as convertible terms.

This term, as usually understood in cases where the
gist of the action is the defendant's negligence, implies
not only that a party has not been negligent or careless,
but that he has been guilty of no violation of law in
relation to the subject-matter or transaction which

constitutes the cause of action.

Injury. Any wrong or damage done to another, either
in his person, rights, reputation, or' property. 'Ehe

invasion of any legally protected interest of another.
Gross negligence. The intentional failure to
perfornr a manifest. duty’ in reckless disregard of’ the
consequences as affecting the life or property of
another.
Invitee. A person is an "invitee" on land of

another if (1) he enters by invitation, express or

62

implied, (2) his entry is connected with the owner's
business or with an activity the owner conducts or
permits to be conducted on his land and (3) there is
mutuality of benefit or benefit to the owner.

Licensee. Person to whom a license is granted. One
who comes on to the premises for his own purpose but with
the occupier's consent. For merely, the duty owed to a
licensee was that of refraining from wilful, wanton and
reckless conduct This rule has been changed and now, in
most jurisdictions, the occupier of land owes the
licenses the duty of reasonable or due care.

Ordinary negligence. The omission of that care
which a man of common prudence usually takes of his own
concerns. Failure to exercise care of an ordinarily
prudent person in same situation. A want of that care
and prudence that the great majority of mankind exercise
under the same or similar circumstances. Wherever
distinctions between gross, ordinary and slight
negligence are observed, "ordinary negligence" is said to
be the want of ordinary care.

Negligence. The omission to do something which a
reasonable man, guided by those ordinary considerations
which ordinarily regulate human affairs, would do, or the
doing of something which a reasonable and prudent man

would not do.

63

The failure to use such care as a reasonably prudent
and careful person would use under similar circumstances;
it is the doing of some act whidh a person of ordinary
prudence would not have done under similar circumstances
or failure to do what a person of ordinary prudence would
have done under similar circumstances. Conduct which
falls below the standard established by law for the
protection of others against unreasonable risk of harm;
it is a departure from the conduct expectable of a
reasonably prudent person under like circumstances.

The term refers only to that legal delinquency which
results whenever a man fails to exhibit the care which he
ought to exhibit, whether it be slight, ordinary, or
great. It is characterized chiefly by inadvertence,
thoughtlessness, inattention, and the like, while
"wantonness" or "recklessness" is characterized by
willfulness. The law of negligence is founded on
reasonable conduct or reasonable care ‘under all
circumstances of particular case. Doctrine of negligence
rests on duty of every person to exercise due care in his
conduct toward others from which injury may result.

Proximate cause. That which, in a natural and
continuous sequence, unbroken by any efficient
intervening cause, produces injury, and without which the
result would not have occurred. That which is nearest in

the order of responsible causation. That which stands

64

next in causation to the effect, not necessarily in time
or space but in causal relation. The proximate cause of
an injury is the primary or moving cause, or that which,
in a natural and continuous sequence, unbroken by any
efficient intervening cause, produces the injury and
without which the accident would not have happened, if
the injury be one which might be reasonably anticipated
or foreseen as a natural consequence of the wrongful act.
An injury or damage is proximately caused by an act, or a
failure to act, whenever it appears from the evidence in
the case, that the act or omission played a substantial
part in bringing about or actually causing the injury or
damage; and that the injury or damage was either a direct
result or a reasonably probable consequence of the act or
omission.

Wilful, wanton or reckless negligence. These terms
are customarily treated as meaning essentially the same
thing. The usual meaning assigned to "willful," "wanton"
or "reckless," according to taste as ix) the word used,
is that the actor has intentionally done an act of an
unreasonable character in disregard of a risk known to
him or so obvious that he must be taken to have been
aware of it, and so great as to make it highly probable
that harm would follow. It usually is accompanied by a

conscious indifference to the consequences, amounting

65

almost to willingness that they shall follow; and it has
been said that this is indispensable.

Knowledge Base Structure

The knowledge base was written in Production Rule
Language, a proprietary symbolic language distributed by
Information Builders Inc. The language is a subset of
the expert system development package Level 5 which also
contains a text editor, compiler, explanation generator,
and control mechanism. The control strategy of Level 5
is a backward chaining pattern matching algorithm
implemented in Turbo Pascal (a high level programming
language). The back chaining control strategy determines
the structure of the knowledge base.

The initial structures of TOTO provide the basic

declaration of fact types, control statements, and a goal

outline:

TITLE TORT (KNOWLEDGE BASE NAME)

STRING Plaintiff's Name (STRING is a fact declaration

AND Defendant's Name indicating that the listed
facts are characters

AND Agency or firm name representing words)

AND What type of activity

AND Hazard

NUMERIC Today's date (NUMERIC is a fact declaration

AND Months since alleged indicating the facts are

injury numbers)

AND How many other
injuries have occurred
at the site

66

AND AGE

AND M

AND B

AND WW

CONFIDENCE Common knowledge (CONFIDENCE enables
AND Expert knowledge certainty factors
AND The Defendant warned of to be assigned to

hazard in question facts)

AND The Defendant instructed
P of hazards

AND

AND

AND

AND

AND

AND

AND

AND

AND

AND

AND

AND P knowledge

AND P ability to avoid

AND P's omission

AND D had knowledge

AND D has ability to avoid

AND D's omission

THRESHOLD - 1 (THRESHOLD sets the minimum value
where a confidence factor
associated with as fact will be
considered true)

wmmCDOVCKQ'U
<0)

'0
O
O
D
Q.
I:
0
d-

1 Setup (This is the goal outline which
1.1 Sum drives the system)

The STRING fact declaration initiates a query to the
system user to enter text so that the system may use it
in the diagnostic session. Quite simply, the text
entered for names will. be 'utilized by the system. to

create customized dialogue during the consultation

session.

 

67

NUMERIC fact declaration preforms two functions in
TOTO. (Months since alleged injury), (How many other
injuries have occurred at the site), and (Age) act
similar to STRING facts in that they are specified by the
user. (M), (B), and (WW) are facts that will be assigned
values based on user responses. These are the major
"fuzzy" operators that will be discussed later.

CONFIDENCE is a control statement that provides a
means as to assess the user's confidence in a state of
fact. This value is also introduced to facts by the
programmer and by mathematics preformed by the system in
response to the system state. In TOTO, CONFIDENCE is
often utilized to map fuzzy numeric clauses to symbols.
This effort basically is an attempt to translate ordinal
(symbolic) data into an interval scale to facilitate
TOTO'S understanding of open texture predicates in
relevance to case facts.

THRESHOLD is another control statement that sets the
minimum value at which the system will consider a fact
true. THRESHOLD works exclusively with facts that have
CONFIDENCE associated with them. The default setting of
THRESHOLD IS 50 WHERE confidence < 50 indicates false and
CONFIDENCE > indicates true. In TOTO, However, the
THRESHOLD is set at 1 because in the majority of cases
CONFIDENCE has been modified by the programmer to preform

fuzzy logic.

68

The goal outline is the basic driving force behind
the system. The system's ultimate direction is
determined by the attempt to meet the goals in the
outline. TOTO queries the user and preforms calculations
until the goals are met or all alternatives are
exhausted. There are hundreds of paths TOTO can take to
satisfy these goals based upon the facts surrounding case
as reported by the user.

The control strategy is deemed backward chaining
because the system begins with the goal and searches the
set of interdependent rules to meet it. In TOTO the goal
(Setup) provides the impetus to gather factual data
relevant to the case prior to entering into legal
decision making. The secondary goal (Sum) provides the
impetus to proceed with legal analysis and the collection
of facts and attributes surrounding the case in hand.
TOTO's goal is to assess landholder negligence as a
defendant as well as the potential contribution to the
injury by the plaintiff. The (Sum) goal is satisfied in
the following rule.

RULE to sum
IF Page one response
AND Page two response
OR NOT Page two response
THEN Sum
(Page one response) is a condition to (RULE to sum)

as well as a conclusion to several other rules. (Page

one response) as a conclusion. to other rules is the

69

terminus to the evaluation of the defendant's negligence.
(Page two response) is also a terminal conclusion that is
satisfied by assessing the comparative negligence of the
plaintiff. Since not all cases involve comparative

negligence the OR clause is implemented.

Rule Structure

The structure of individual rules is based on
boolean logic which is naturally intuitive. Each rule
must have a title which assists in debugging and the
production of explanation facilities. The IF statement
is always the initial condition of the rule followed by
more conditions linked to an AND command or OR command.
The logic of the rule is strictly sequential. The THEN
statement is followed by the state of fact associated
with satisfying or "firing" the rule. The ELSE statement
may be placed after the THEN command to provide
information on the failure of the rule to fire for other
rules to act upon.

Most of the rules in TOTO are to some degree
dependent upon conclusions of other rules. Some have a
shallow structure where all conditions are virgin,
whereas others may be influenced by other dependent
rules. Rules that have the deepest structure become
recognizable as concepts and are in TOTO labeled

METARULES. The set of METARULES embody the basic legal

70

rules and inferencing strategies that constitute the

problem-solving behavior in the macro sense. In the

micro sense we see the difficult problems as well as

routine rules of fact as the foundation of the METARULES.

CONTROL RULES are a class above METARULES that reside

closest to the goal outline: and direct basic problem

solving as well as reporting facilities. The resulting

structure is simplified and illustrated in Figure 9.
Returning the (Sum) CONTROL RULE and its conditions

of:

IF Page one response

AND Page two response

OR NOT Page two response

we see that these. conditions are also» conclusions to

other control rules. For example, the (Page one

response) condition is a conclusion to several other

rules including (output no duty):

RULE output no duty

IF No Liability ordneg

OR No Liability grosneg

AND NOT Duty

THEN Page One response

AND DISPLAY No burden duty

This rule simple states that if TOTO cannot find

negligence of any kind because there is a failure to

demonstrate legal duty, then make a report discussing

these results in relation to the case in hand.

The function of (Page two response) is similar yet

is only activated when negligence by the defendant is

71

Recreation User
Trespasser Act
Statute of Umitation
Tort Claims Act \i:

Statutory

Limiting _D.~_Eir';2"f§ee

 

" '9 ’ ' trespasser
Elemfeni‘s —Duty Breached
Negligence
— Causation
— Injury

—— Assumption of Risk

 

Comparative Negligence
(case law)

 

— Waivers

 

— Intentional Nuisance

Figure 9. Modules in TOTO's Knowledge Base.

72

discovered. The rules that conclude with (Page two
response) are the series of METARULES that assign the
appropriate legal defenses for the defendant. For
example:
META RULE to invoke rec use statute

IF Liability ordneg

AND invoke rec use statute

THEN No liability

AND Page two response
AND DISPLAY rec use

META RULE to invoke comparative negligence

IF Liability grosneg

AND CN exists

THEN Partial Liability

AND CONF (LL):=100-CONF(CN)

AND Page two response

AND DISPLAY compneg
IN the first rule, the condition (AND invoke rec use
statute) sends the back chaining algorithm to search for
this as a conclusion of other rules. In this case TOTO
would discover the DEFENSES module and subsequently the
set of rules that assess the potential application of the
recreation use statute. If all conditions are met, the
(META RULE to invoke rec use statute) will fire and a
custom explanation of the case assessment will be
generated.

In the (META RULE to invoke comparative negligence),

the search process is similar, except that here numeric

73

values are discovered and comparative negligence is
calculated.

The DISPLAY operator acts to call a canned display
screen that has open areas in it that allow for
customizing the response. The names of litigants and
factual data gathered during the session are dropped into
the DISPLAY screen to make explanation of TOTO's
conclusions customized to the current consultation. The
TEXT command has a similar function, but is utilized to
customize specific queries during the session. For
example:

TEXT Permission

Did [Plaintiff's name] have permission from the
[Defendant's Name] to enter the premises owned and
controlled by [Agency or firm name]?

Lack of permission includes areas that have been marked
"Closed at Dusk" or "No Swimming."

The use of TOTO would view a screen similar to the

following:

 

Did Robbin Smith have expressed or implied permission
from Fred Nugent to enter the premises owned and
controlled by the Michigan Department of Natural
Resources?

Lack of permission includes areas that have been marked

"Closed at Dusk" or "No Swimming."

74

Under the (Setup) goal in the goal outline, TOTO has
elicited the factual information needed for this
particular DISPLAY.

The EXPAND operator is again similar to DISPLAY and
TEXT in that it provides information to the user. When
the EXPAND icon is clicked by the system user
supplemental explanatory information is presented. The
information is context sensitive and will provide
explanations relative to the current rule under

consideration.

Search Strategy

The hierarchical set of rules in the knowledge base
are searched by the back chaining inference mechanism to
find a combination of conditions that will ultimately
meet the goal statements in the goal outline. This
search strategy means TOTO will always attempt to find
the shortest path to a goal statement. As answers are
provided by the user, a new set of shortest paths are
brought into consideration. The full breadth of TOTO's
knowledge base is considered only when it encounters the
most difficult or unusual case circumstances. In such
cases the knowledge of TOTO has been exhausted and the
system will fail to reach a conclusion. The user than

has the ability to change his/her responses to TOTO's

75

queries in the attempt to discover why the system was
confounded.

The breadth of knowledge may be represented as the
"world according to TOTO." Sequential responses by the
system user continually modify the search space under
consideration. Thus in most cases most of TOTO's
knowledge base is never brought into consideration. For
example, if TOTO discovers that the plaintiff is a legal
trespasser upon the defendant's premises, the entire
branch of knowledge surrounding the issue of ordinary
care is cut off. This determinism comes into play when
straightforward rules of .law are ‘under consideration.
When the legal principles under consideration are subtle
or of open texture, small branches of the knowledge base
are pruned in a process of gradual refinement. Thus easy
problems in TOTO have the greatest ability to cut entire
knowledge branches, whereas difficult problems have less
pruning power. Difficult problems require a more

meticulous approach to pruning the search space.

Model Structure
The basic knowledge modeled in TOTO is separated
into two units. The first unit is the knowledge needed
to assess the burden of proof which lies upon the
shoulders of the plaintiff. In its simplest form, this

involves proving that the following elements are present:

76

A DUTY to meet a standard of care for the plaintiff
A BREACH of that duty

A CAUSAL connection

An INJURY

The second unit in TOTO assesses the potential
defenses available to the defendant. The burden of proof
unit must have determined that negligence exists in order
for the second unit to assesses defenses. In order to
demonstrate the actual content of the knowledge base the

following annotated outline is provided.

Elements of Negligence
DUTY

Establish relationship between players
Defendant type:
Public or
Private
individual
individual & agency/firm
Administrator
participates in tortuous conduct
hiring
training
discretionary function
ministerial function
Employee
act within scope of duty
while on duty
Volunteer
within scope of duty
viewed as an employee

Plaintiff type

Invitee
permission-yes
on D,s proprietorship
express invitation by D
paid fee to enter to D
or fee for activity

77

Licensee
permission—yes
no business transaction with D
no cash exchange with D
Trespasser
no permission
physical invasion of D,s premises
and "park closed" signage
or "no swimming after dark"

Minor
Under 13 years of age
BREACH
STANDARD

Ordinary negligence
only invitee can claim lack of ord care
based upon "reasonable prudent man"
OR reasonable prudent professional
inspect, repair, remove rule
warning or instruction
STANDARDS
written and customs, must be
certain
uniform
well known
obvious
Hand test

Magnitude of Risk v Burden of
Alternative conduct

Magnitude of Risk -
foreseeable probability
of harm + gravity of harm

Feasibility of Alternative
Conduct -
relative cost of
alternative conduct +

relative utility of
safer conduct +

relative safety of
alternative conduct +

78

Willful and Wanton Misconduct

players defined by statute

TRADITIONAL DEFINITION

D's knowledge of situation requiring
ordinary care to avert injury

D's omission of care when threat is
apparent

D's intent or omission of care

MICHIGAN DEFINITION

intent to harm

or indifference resulting SUI willingness
for harm to occur

Trespassers

Discovered
D discovers P
D continues risk activity
not natural land condition
creating hazard
D knows P will not discover
hazard
D has control of elements
creating hazard
D provides no warning or
instruction

Frequent

frequent intrusion to specific
area

active operation that create risk
of bodily harm

no warning of hazard

D knows hazard will not be
discovered

Minor

D knows or should know minors
will trespass

D aware of danger

hazard not natural land condition
D had power to limit risk

79

CAUSATION
Cause in Fact

"butfor" conduct (direct string of cause-effect
events)

coterminous—-hazard & injury

physical contact

avoidance of hazard

Proximate Cause--Unforseeable Consequences

consequences were foreseeable
P was in foreseeable zone of danger
or rescuers caused injury

Prox Cause-—Intervening Cause

intervening cause foreseeable

consequences of intervening causes foreseeable
Non—extraordinary weather conditions weather
third party negligence

third party criminal conduct

INJURY

person
chattels

Defendant's Defenses

Expressed Assumption of Risk
Signed waiver (useful evidence of assumption of
risk)
not against public policy
unambiguous language
voluntary participation
adult cannot sign away rights of minor

Implied Assumption Risk

P knows, appreciates, understands risk
voluntary participation
conduct manifests consent

Statute of limitation

P knows, appreciates, understands risk
voluntary participation
conduct manifests consent

80

Statute of limitation

3 years after injury
after 3 years must prove latent or delayed
causation

Michigan Recreation Use Statute

All lands where no fee is paid is excluded from
actions against D where ordinary negligence is
the issue

Thus P must seek willful and wanton misconduct by

D

State Park fee is for parking thus under
protection

IN MI all public agencies are protected (local,
state, fed)

Goal is to minimize reduction of recreational
opportunity

Comparative Negligence

Conduct by P enhancing CAUSATION of injury

Use willful and wanton misconduct rule against P

conduct shows indifference to harm self
P has knowledge that ordinary care would avert injury

P has ability to avoid injury

P conduct show omission of care

If found, D's negligence is reduced up to 50%

D 60%/P 40% as example

Damages (payment) is adjusted to fit proportion

Open Texture Predicates
The determination of reasonable conduct in ordinary

.negligence. cases is based 'upon. what. a reasonable and
prudent person would do in a given situation. The
"reasonable and prudent person" guideline is ultimately
assessed by a jury or judge from case facts and arguments

based on previous cases. This determination is not an

empirical one and thus remains problematical in the

81

determination of reasonableness. TOTO, however, is
limited to pretrial decision making and the behavior of
juries is not relevant.

The pretrial determination of breach of duty will
include an evaluation of reasonable conduct based on a
loose collection of legal guidelines. A. classic
formulation of determining when conduct is unreasonable
is the "Hand" test (United States v. Carrol Towing Co.
(2nd Cir 1947), L. Hand, J.).

a risk is unreasonable when the foreseeable

probability and gravity of harm outweigh the burden

to D [defendant] of alternative conduct which would

have prevented the harm (Kionka, 1988).

This test was selected to be formalized into the
knowledge structure because it can be broken into
components that fit the structure of production rules.
In the simplest form, the test stated as:

Magnitude of risk v Burden of alternative conduct

where

MAGNITUDE OF RISK = probability + gravity of harm

and

BURDEN OF ALTERNATIVE CONDUCT relevant costs

The relevant costs associated with the burden of

alternative conduct include:

1. the importance or social value of the activity
or goal of which D's conduct is part;

2. the utility of the conduct as a means to that
end;

82

. the feasibility of alternative, safer conduct;
the relative cost of safer conduct

the relative utility of safer conduct

the relative safety of alternative conduct
(Koinka, 1988)

O‘U'lbw

Thus, the burden of alternative conduct is the sum of the
six factors. In order to operationalize these variables
and make them usable in the knowledge base, they were
transformed into rule sets that incorporated confidence
ranking. The degree of membership for both MAGNITUDE OF
RISK and BURDEN OF ALTERNATIVE CONDUCT is determined by
values associated with the individual factors. During a
consultation session, the user is queried as to how case
conditions relate to these factors and to rank compliance
to the factor concept. A METARULE guides the process and
calculates BURDEN OF ALTERNATIVE CONDUCT and BURDEN OF
ALTERNATIVE CONDUCT membership values. By balancing the
opposing values, "reasonable conduct" is accepted or
rejected.

The same method was utilized to determine willful
and wanton misconduct. If the plaintiff is an invitee or
licensee, a general case assessment is invoked. The
determination of willful and wanton misconduct is based
on Michigan case law as follows:

1. knowledge of a situation requiring the exercise

of ordinary care and diligence to avert injury
to another

2. ability to avoid the resulting hann by ordinary

career and diligence in the use of means at hand
and

83

3. the omission to use such care and diligence to
avert the threatened danger when to the ordinary
mind. it must. be apparent that the result is
likely to prove disastrous to another (Thomas v.
Consumers Power Co.).

 

 

and

Willful and wanton misconduct is distinguished from
ordinary negligence by intent to harm or by an
indifference of the defendant of a defendant in the
presence of the probability of harm which is
tantamount to a willingness for that harm to occur
(Williams v. City of Cadillac).

 

The concepts for representing willful and wanton
misconduct were formalized into rules and METARULES
incorporating confidence ranks. When the confidence
numbers are gathered for each fact they are summed and
averaged to create an aggregate value. As the aggregate
value increases the potential for willful and wanton
conduct increases. If the value surpasses a threshold
then willful and wanton misconduct is confirmed. If the
aggregate value does not reach the threshold, it is
rejected.

The assessment of willful and wanton misconduct
where the plaintiff is a trespasser follows a similar
procedure, but incorporates a more extensive evaluation
of case facts. Special rules sets assess condition where
trespass is frequent, discovered by the defendant, or
involves a minor. These special cases require an
evaluation of the defendant's knowledge of trespass

conditions in order to assess his/her conduct.

84

Analysis and Results

TOTO has reached the stage of a research prototype
system. Figure 10 illustrates the range of development
stages for expert systems.

There are more than 120 rules and 300 fact
statements in TOTO. The rule set was condensed from more
than 300 by the creation of METARULES and attribute/value
(a/v) pairings. As discussed previously, METARULES were
implemented to provide a hierarchy of rule classes to
simplify the knowledge engineering process. The process
of organizing knowledge into rule classes created the
opportunity to isolate rules that have several special
functions. Thus, some rules are primarily collections of
facts, whereas others handle fuzzy math or provide
control to the program. This division of labor reduced
the need for redundant actions completed by several
rules.

The rule set was further condensed by extensive use
of attribute value pairings. The following rule set is
an implementation without a/v pairings.

RULE to determine public control
IF Public

AND Defendant is in the public trust

85

 

Development Stage Description
Demonstration The system solves a portion of
Prototype the problem undertaken,

suggesting that the approach is
viable and system development is
achievable. One to three months

to develop.
Research Prototype The system demonstrates credible
performance on the entire

problem but may be fragile due
to incomplete testing and
revision. One to two years
development time.

Field Prototype The system displays good
performance with adequate
reliability and has been revised
based on extensive testing in
the user environment. Two to
three years development time.

Production Model The system exhibits high
quality, is reliably fast; with
efficient performance in the
user environment . Two to four
years development time.

Commercial System The system is a production model
being used on a regular
commercial basis. Four to six

years development time.

 

Figure 10. Evolution of Expert Systems (From Waterman,

THEN

RULE

IF

THEN

RULE

IF

THEN

RULE

IF

AND

AND

THEN

86

Defendant is on the board of directors

Classified defendant

to determine public control

Public

Defendant is in the public trust
Defendant is the enterprise administrator

Classified defendant

to determine public control
Public

Defendant is in the public trust
Defendant is an employee

Classified defendant

to determine pubic control
Public

Defendant is in the public trust
Defendant is a volunteer

Classified defendant

Now with a/v pairings on rule handles the task.

RULE
IF
AND
AND

OR

to determine public control

Public

Defendant IS in the public trust
Defendant IS the enterprise administrator

Defendant IS on the board of directors

87

OR Defendant IS an employee
OR Defendant IS a volunteer

THEN Classified defendant

The IS statement enable the attribute "Defendant" to
take on a range of values. The status of "Defendant"

will be determined by the system user or deduced by TOTO.

Test Cases

 

During development, TOTO was continually tested by
running hypothetical case facts through the system and
monitoring its behavior. When problems were encountered,
the knowledge base was modified to correct the aberrant
behavior. As the knowledge base grew to more than 100
rules, the corrections and additions tended to introduce
more errors than they fixed. This problem was addressed
by standardizing test cases which would isolate the
effects of modifying the knowledge base.

At the stage of research prototype TOTO's ability to
identify the important elements in a case was tested by
using hypothetical case facts during consultations by
domain experts. Initially, the use of actual cases in
the Michigan civil law arena were proposed as testing
devices. This approach was problematical for the design
of much of the knowledge base was based on these cases.

There would be little use testing the system on what it

88

already should know. Thus hypothetical cases were
constructed by the domain experts to supply facts for
TOTO to process. The following test cases and
discussions of TOTO's reasoning illustrate the level of

performance the system has attained.

Test case 1. In this test case the plaintiff,

 

Robbin Smith, entered the defendant's premises to engage
in a game of golf. Phil Dirt's golf course charges a $15
green fee and requires golfers to sign a release
accepting the risk of being struck by golf balls. Phil
Dirt does not own the land upon which the course lies,
but leases the premises to run his business operation.
While driving her golf cart over a bridge that spanned a
stream, the back wheel slid over the edge and sent the
cart into the opposite bank of the stream striking a
cement foundation that supported a since removed bridge.
Smith suffered severe head injuries as a result of the
incident. other golfers have lost control in a similar
manner, yet remained "hung up" on the bridge. The
defendant had, on two separate occasions, used his
tractor to tow carts of the structure having one wheel
slide off the edge of the bridge. In order to proceed
with a civil case against the defendant, Smith has the
burden of proof to demonstrate that negligence exists and

consults TOTO to assist to do so.

89

The user began the session by entering the day's
date, the defendant' name, the plaintiff's name, and the
name of the firm or agency that the defendant is
associated with (see Appendix A for complete user session
report). Then TOTO elicits the activity type in question
for evaluation later. At this point, TOTO is only aware
of the rudimentary facts associated with the case. These
facts, however, will enable TOTO to address the litigants
by name and by facts associated with the case. TOTO uses
information learned in the session to customize queries
so that only relevant questions are posed to the user.
This process eliminates the asking of redundant or
irrelevant questions during a consultation.

The plaintiff, as identified as Robbin Smith, is
then assessed as to her legal status while on the
premises of the defendant, Phil. Dirt” 'The rule for
PERMISSION is invoked and TOTO determines whether or not
Robbin Smith is a TRESPASSER or of some other status.
TOTO determines that she is not a trespasser, and thus,
should be assessed as a licensee or invitee.

The sequence of TOTO's logic is not monotonic, and
thus, it pursues the solving of the case by the shortest
path available. This search strategy is called "depth
first" because TOTO will always search the entire
knowledge base for the simplest solution. The upper half

of Figure 11 illustrates the knowledge base as an

90

A. A protion of the decision
tree proir to fixing of the
'Pcnnission" rule.

 

B. Subsequent pruning of the
decision tree upon firing of the
"Pumhamfnm:

 

Figure 11. Firing of the Permission Rule and Subsequent
Pruning.

91

inverted decision tree. As TOTO gains information during
a session, the branches are pruned, narrowing the
available solutions. In the case at hand, the firing of
the PERMISSION rule invokes a search strategy that
disregards rules of law associated with trespassers and
the tree is pruned as in the lower half of Figure 11.

Now TOTO attempts to evaluate the defendant by
searching the METARULE "to determine landholder type."
The user supplies the fact that the premises in question
is owned by a private interest. Next TOTO presents a
list of options to the user to determine whether or not
the defendant owns or leases the premises, and was in
actual control of the premises (including land
conditions) at the time of the alleged injury.

The next series of queries attempt to determine the
nature of the defendant's operation. In this case, TOTO
determines that Phil Dirt runs a proprietary business
operation to provide golfing opportunities to the public.
Also, it becomes clear that the public is openly invited
to play at the course and that Phil charges a fee for the
activity. It is disclosed that the plaintiff, Robbin
Smith, entered the premises to play golf, and willingly
paid a feed to do so. Thus TOTO has completed assessing
both the plaintiff and defendant and classified Phil Dirt

92

as the proprietor and manager of land condition for the
course. Robbin Smith has been identified as a business
invitee.

This information is gathered or deduced for the
purpose of determining the relationship between the
litigants at the time of the alleged injury. The
determination of the nature of this relationship enables
TOTO to the DUTY requirement in negligence law. At this
point, TOTO has established that the. defendant, Phil
Dirt, has a duty of some sort to protect Robbin Smith
from harm. The exact level of care required by the
defendant will be assessed when TOTO proceeds.

A recurring problem in legal expert systems is the
requirement of the user to exercise a degree of legal
judgment during system consultation. In TOTO this
problem is first encountered when the user must decide if
the case is one of willful and wanton misconduct or one
involving ordinary negligence. The user is prompted to
decide based upon previous experience or the information
provided on screen. In some cases the user may select
the inappropriate type of negligence to consider. For
example, if the plaintiff has been classified as a
trespasser, they are unable to propose a suit based on
ordinary negligence. If 'this situation arises, TOTO
refuses the line of reasoning, explains why, and returns

to allow the user to try again.

93

In the case at hand, the plaintiff' will pursue
ordinary negligence against the defendant. 2hr order to
assess the standard of care, the defendant should provide
to ensure safety for the plaintiff TOTO examines his/her
level of expertise in the recreation activity and the
existence of relevant industry standards. Phil Dirt was
assessed to have a high level of expertise in golf and
golf course management. Also, there are design and
maintenance standards for the golf course industry. TOTO
queries the user as to the certainty of the standards,
their uniformity, and that they are well known and
obvious. Here the standards do not have to be written to
meet these criteria. TOTO then concluded that relevant
safety standards exist and that with the expert knowledge
of the defendant, there is a measurable standard of care
due to the plaintiff. The user is queried to estimate
the defendant's compliance to the standard on a scale of
1 to 100. This estimation will be combined with other
estimators to determine the reasonableness of Phil Dirt's
conduct.

The next series of queries revolve around the
attempt to assert that a breach of duty exists. The user
is asked to rank. his/her confidence in 'the «defendant
performance basic measures to ensure safety for the
recreating public. In this case, the user responded as

follows:

94

Defendant inspects the premises regularly SCORE
to discover potential hazards to the public . . . . 67
Defendant removes discovered hazards . . . . . . . 45
Defendant provided warning of hazards . . . . . . False

Defendant provided instruction on how to

avoid hazards . . . . . . . . . . . . . . . . . . False

To conclude the assessment of a potential breach of
duty the issue of’ foreseeability' is brought into the
session. Here TOTO queries the user as to the frequency
of injuries under similar circumstances and the
defendant's ability to anticipate future similar
injuries. Again, the user places confidence values in
association with his answers. These confidence values
are combined with the others obtained in the "breach"
module and used by TOTO to calculate membership in the
"breach" or "no breach" sets. Fuzzy numbers and set
theory is utilized to transform discrete interval data
into group memberships. Figure 12 illustrates the
difference between traditional logic and the fuzzy logic
TOTO utilizes to determine the reasonableness of the
defendant's conduct. The top graph maps the degree to
which reasonableness belongs to the value "low." As the
degree of membership increases, the reasonableness
decreased on a continuum. The bottom graph illustrates

the traditional representation of low reasonableness as a

95

 

 

 

 

 

 

Fuzzy Representation of Low Reasonableness

i .0 -
.9
r:
i
E
o
2
'6
8
a
a
o

0
Unreasonable Reasonable
Traditional Representation of Low Reasonableness

i .0 -
.9
E
o
o
E
o
2
‘6
t
8’
o

0

Unreasonable Reasonable

Figure 12. Traditional Logic v. Fuzzy Logic.

96

dichotomous true/false state of fact. TOTO uses both
methods to determine the state of "reasonableness" and
the determination of a breach of duty. In this case, it
was determined that there was a breach of duty and the
issue of the injury was brought into session.

The module that assesses the injury to the plaintiff
is straightforward in that the user is posed only yes/no
questions. The plaintiff, Robbin Smith, suffered
physical harm was substantiated by a physician and thus
the injury was considered valid.

The issue of the causal link between the breach of
duty and the injury focuses on both Space and time
dimensions. Initially, however, TOTO assists the user in
determining the type of causation to be considered.
Within the array of cause types, it was determined that a
"cause in fact" would be assessed. It was determined
that the injury would not have occurred "but for" the
existence of the hazard. It was also determined that the
hazard in question was in existence at the moment of the
injury and that the injury resulted from direct contact
with the hazard. TOTO subsequently concluded that there
was indeed cause in fact in this case.

At this point the line of reasoning report reveals
that TOTO has collected enough information to conclude
that ordinary negligence exists in this case. Thus, the

plaintiff's burden of proof to demonstrate duty, breach

97

of duty, cause, and injury' was successful“ .Now' the
system acts in behalf of the defendant to discover
potential defenses to the negligence.

The first TOTO pursues is to invoke the statute of
limitations. Since the injury occurred 13 months prior
to the consultation this attempt fails. The user then
responds that a waiver was signed prior to using the golf
course. The language of the waiver, however, is
discovered not to mention the type of risk that led to
the injury. Thus, the'assumption of risk defense is
defeated as well.

TOTO concluded analysis of the case by asserting
that negligence exists with no serious threat of defenses

proposed by the defendant.

Test Case 2. The plaintiff Fred Nugent was camping
(Ml a Michigan State Forest near an abandoned gravel pit
that contained a pond that was used for swimming.
Campers, as well as the local public, used the pond as a
swimming hole, even though the entry road was posted with
a "no swimming" sign. Nugent was swimming and drinking
beer with friends and driving from a platform that was
constructed on an overhanging tree. Nugent made several
dives from the platform and commented to the others that
they should be careful to avoid some cement fragments

that were farther out than the normal point where one

98

enters the water when diving. After a few more beers
Nugent, an accomplished swimmer and diver, was challenged
to attempt to dive over the cement blocks into the water
beyond. The first dive was successful, yet Nugent
scraped his knee on the cement. The second dive was not
successful, and Nugent struck his head on the cement and
was paralyzed.

The DNR posted the "no swimming" sign, and at one
point, removed the rdiving platform because a swimmer
broke her ankle upon jumping from the platform. Since
the defendant paid a fee for camping or swimming, the DNR
assumed that protection from the state's recreation use
statute would relieve them from civil actions.

These facts and others supplied by the user were
input into TOTO to assess the potential for negligence
and potential defenses. The complete line of reasoning
report is presented in Appendix B.

The first module in TOTO gathered basic information
to customize subsequent queries and results screens. The
defendant is identified as a manager of the lands for the
public trust that does not charge a fee for dispersed
camping. Thus, in this case, the defendant does not
operate a business operation which would influence the
degree of due care owed to the plaintiff. The land was
unrestricted state property and by default, the plaintiff

had permission to enter and ‘use the jpremises. 'TOTO

99

determined that Fred Nugent was licensee on the premises
which precludes the using for ordinary negligence. If
ordinary negligence was pursued at this point TOTO would
allow the attempt, yet would advise that this path of
reasoning is futile. Guidance would be provided until
the user pursued willful and wanton misconduct on the
part of the defendant.

The next series of queries focuses on addressing the
potential for, willful and wanton misconduct by the
defendant. As with test case 1, this assessment utilizes
fuzzy sets to assign membership or lack thereof. The

user's responses are summarized below.

SCORE
Did D have knowledge of the hazard in question: 100
Did D have the ability to avoid P's contact with
the hazard? 70
Did D's omission to remove the hazard result in
the injury? 60
Did D show intent to harm P 0
Did D's conduct show indifference to weather
the hazard would harm P? 90

These responses in combination with the level of
care owed to the plaintiff led TOTO to assign the
defendant to the willful and wanton conduct set. As with
tests case 1, it was .next determined that there: was
injury in fact. The next step was to assess the

causation for the injury. With prompting by TOTO the

100

user selected the cause in fact path for analysis. In
this case linking the hazard to the injury was not
difficult. The water depth was sufficient for diving and
the injury would not have occurred but for the existence
of the submerged cement blocks. From the line of
reasoning report, it can be seen that at this point, TOTO
has enough information to assert negligent conduct on the
part of the defendant. Immediately the search for
potential defenses is conducted.

Since it has only been 15 months since the injury,
the statute of limitations defense was eliminated. The
next defense considered was that of comparative
negligence. Here the plaintiff is assessed in the same
manner as the willful and wanton misconduct of the
defendant. The responses to TOTO's queries are

summarized below.

SCORE
Did P have knowledge that ordinary care would
have averted the injury? 69
Did P's conduct show indifference to whether
harm would result? 75
Did D have the ability to avoid the injury with
ordinary care? 83

Was there an omission of such care when the

threat was apparent? 90

101

From these responses , TOTO concluded that
comparative existed and weighed that against the
negligence of the defendant. In the final report screen,
TOTO summarized its findings and allocated 79% of the
negligence to the plaintiff. Comparative negligence in
Michigan, however, relegates fault to the plaintiff to
where it is "not greater than" the defendants. Thus, the

defendant will be responsible for 50% of damages.

Test case 3. The third test case was designed to
enter the branch of the knowledge base that handles the
situation where the plaintiff is a trespasser. The full
history of the case is displayed in Appendix C.

As with the other cases, the beginning screens
gather factual information to enable TOTO to generate
custom text for the session. In this case the plaintiff
trespassed upon the defendant's gravel pit operation to
fish in a pond therein. The user asked TOTO to pursue
ordinary negligence, yet TOTO refused, and recommended
that alleging willful and wanton misconduct is the only
avenue open to trespassers. This route was selected. It
was subsequently determined that the plaintiff' was a
frequent trespasser that the defendant was aware of. In
order to determine the defendant's duty to the plaintiff,
TOTO queried the user as to the nature of the injury.

These queries disclosed that there were active operations

102

that could injure the plaintiff, that these conditions
are man made (rather than natural) and that they would
not readily be discovered by the plaintiff. This led
TOTO to assert that there was willful and wanton
misconduct by the defendant.

Next it was determined that there was injury in fact

and that there was a causal link between the hazard and

the injury. The causation type selected included an
intervening element. This element was considered
foreseeable so the causal link was substantiated. At

this point, TOTO had gathered enough information from the
interview to assert that there was willful and wanton
misconduct by the defendant.

Now relevant defenses were considered. Since TOTO
knows the status of the litigants and the case
conditions, there is not consideration of defenses that
will not apply. For example, this case is one of
alleging willful and wanton misconduct by the defendant,
and the language of the Michigan recreation use statute
excludes protection from this type of negligence. Thus
TOTO ignores the defense. Ultimately, the comparative
negligence defense was invoked which reduced the claim on
the defendant by 50%.

This test case uncovered an error in TOTO's
knowledge base. There was a failure to correctly assess

the facts surrounding the trespass. The user was forced

103

to select from a range of conditions surrounding the
trespass including as follows:

Select the appropriate conditions of the trespass:

A. Trespass was frequent by the plaintiff

B. Trespass was discovered by the defendant

C. Trespass was by a minor

D. None of the above
TOTO forced the user to select the statement that most
closely resembled the case conditions even though more
than one may apply. In this test case, the trespass was
frequent and by a minor yet the user selected option (A).
This posed no problem in assessing the negligence, but
did so for the defenses. Remember that TOTO asserted
that there was comparative negligence on the part of the
plaintiff. This essentially means that the plaintiff
demonstrated willful and wanton disregard for his own
safety. By selecting option (A) TOTO never considered
the age of the plaintiff which is this case was six
years. If TOTO's protocol was correct the comparative
negligence defense would not be entertained for legally,
a minor is incapable of willful and wanton disregard for
his safety. This problem in the rule base can be
corrected, but it highlights one of the difficulty of
testing program at this stage of development.

104

Discussion

The three test cases presented here are a small
portion of the full array used to validate the program.
They represent both adequate and poor performance in
identifying key issues of a case. The tests are
qualitative assessments of the program's performance and
are judged by subjective measures. Testing of the
knowledge base is problematical because there are an
extensive number of combinations of rules that TOTO may
utilize in analyzing a case. The number of potential
combinations can be automatically calculated by creating
a knowledge tree (a function in the Level 5 expert system
shell). A knowledge tree was created, yet it could not
be viewed or printed in its entirety. The document was
too large to fit in the computer's memory. Less than
one-forth of the knowledge tree constituted more than 500
pages of text. A sample of the knowledge tree output is
included in Appendix D.

The combinatorics of the knowledge base makes
exhaustive testing very time consuming. To develop the
system beyond research prototype, the knowledge base
could be tested module by module. The METARULES then
could be tested on how they manipulate the modules.

Hays-Roth, Klahr, and Mostow (1986) have outlined

sources of error that may be uncovered during the testing

105

process as listed in Table 1. The problems with TOTO's
performance fall into some of the categories listed.

The first problem of excess generality is manifest
in TOTO's ability to determine who is the defendant in a
case. For example, the rules specify that a public
sector defendant is either an individual, the agency at
large, or both. Here the case of multiple defendants is
ignored.

The fourth and sixth problems were demonstrated when
a domain expert and the programmer's expectations were
not met during a test case. TOTO quite simply misapplied
the Michigan recreation use defense for two reasons: the
rule was conceptually incorrect and the syntax of the
rule was invalid.

The seventh problem listed in Table 1 of inadequate
integration is clearly displayed in test case three
above. In the search for applicable defense, TOTO
selected the path of comparative negligence because the
plaintiff was declared a frequent trespasser. In reality
the plaintiff was both a minor and a frequent trespasser.
Thus TOTO would not reject the comparative negligence
defense as it should in this case.

The aforementioned problems are. expectable, given
the developmental stage of TOTO as a research prototype.
Most of these deficiencies could be remedied by extensive

testing and revision. The matter of open texture

106

 

 

Table 1. Bugs Arising from Knowledge Programming
Type of Source of Problem Manifestation
Problem

 

1. Excess
generality

2. Excess
specificity
3. Concept
poverty

4. Invalid
knowledge

5 . Ambiguous
knowledge

6. Invalid
reasoning

7. Inadequate

Special cases
overlooked

Generality
undetected

Useful relationship
not detected and
exploited

Misstatement of
facts or
approximations

Implicit
dependencies not
adequately
articulated

Programmer
incorrectly trans-
forms knowledge

Dependencies among

Good rule occasionally
produces bad effects

Rules fail to cover
enough cases

Limited power and
capability of system

Expert's expectations
violated

Conflicts arise in
some situations about
what is best to do

Knowledge programmer's
expectations violated

Rejected action

 

integration multiple pieces of alternatives satisfy
advice actually more criteria than
incompletely selected action
integrated does.

8. Limited Consequences of Judgmental logic

horizon recent past or seems static, not
probable future sensitive to
events not changing or
exploited foreseeable

situations

9. Ego— Little attention No apparent adaptation

centrity paid to probable of one's behavior to
meaning of other's exploit knowledge of
actions other's plans

Source: Hays-Roth, Klahr, and Mostow, 1985.

107

predicates, however, has a deeper conceptual problem.

The scheme TOTO uses to bring meaning to open
texture predicates is not without problems.
Specifically, the concepts of "reasonableness" and
"willful and wanton misconduct" proved to be
problematical. The process of gradual refinement narrows
the meaning of these concepts into operational terms, yet
to a degree is based on the judgment of the user. The
expertise guiding the knowledge base is valid, and legal
expertise is not required of the user. It is the ranking
of confidence to case facts by the user that requires
discretion. For example in the willful and wanton
misconduct module, the user may be queried as follows
(the user can input any number from 1 to 99):

Did the defendant Jane Smith have the ability to
avert injury to the plaintiff Duck Jones, by exercising
reasonable care?

Rank your response below.
No ability Little ability Some ability Great ability

1 25 75 99

 

And next:

 

108

Was there an omission of care by Jane Smith to avert
an injury to Dick Jones when the threat of injury was
apparent?

Rank your confidence in the truth of this statement.
No Not for sure Quite sure Yes

1 25 75 99

 

In each case user judgment is required to gather case
facts. The method was implemented to provide analysis
beyond yes/no responses.

The difficulty in bringing meaning to open texture
predicates is not unique to expert system development.
The problem is identical to decision making by legal
experts. For example, when attorneys and judges assess
"reasonable conduct," there is no discrete formula to
guide the process. Judgment is required to refine how
legal concepts relate to conditions surrounding a case.
The process implemented in TOTO refines the meaning of
vague concepts incrementally based on the gathering of
information. Since the goal of the system is to identify
key case issues, the focus is on whether TOTO identifies
key elements of the concepts.

In the civil arena the determination of negligence
is based on the "more likely than not" rule. This means

that the determination of negligence is based on the

109

preponderance of evidence rather than "beyond a doubt" as
in criminal law. TOTO attempts to weigh case conditions
to create a balance of evidence which will favor either
the plaintiff or the defendant. The evidence is gathered
based on TOTO's reasoning, yet the ‘user' is sometimes
required to supply judgment on case elements. The use of
the refinement process in conjunction. with confidence
factors serves to approximate the process that a legal
expert may determine what "reasonable conduct" is. This
process is obviously a gross implication in comparison to
expert legal decision making.

When the system assesses for comparative negligence,
a value is given indicating the degree of negligence for
which the defendant is responsible. Thus, if the
defendant is found negligent and the plaintiff
contributed to the injury, the degree of negligence is
reduced by a percentage. TOTO reports a discrete number
from 0 to 99 indicating the defendant's percentage of
negligence. This number should be viewed as a guideline
only and is precisely calculated because of the way the
software computes mathematics. This degree of negligence
could be softened by implementing fuzzy set techniques to
result in statements such as, "The defendant is slightly
negligent." In civil law, however, there is an actual
percentage that is determined by the proceedings.

Therefore, the averages of the confidence rankings for

110

comparative negligence queries were used to produce the
numeric result. These are only estimations based on the

user's responses and must be viewed as generalizations.

Summary and Conclusions
Summary

Providers of recreation need to know what negligent
conduct is. ZUT order to understand the legal decision-
making process that determines negligence, one must enter
into the civil arena as a defendant in a suit. This ex
post facto understanding is of marginal value because an
injury has already occurred. There are legal, economic,
and ethical motivations for recognizing .negligent
conditions and remedying them prior to injury.

A knowledge based system was developed to model the
decision-making' process litigants use to determine if
negligence exists prior to entering the civil arena. The
program TOTO focuses on assessing potential negligent
conditions of the land for defendants that are private
individuals, public agencies, or proprietary operations.
The overall goal of the system is to aid the user in
identifying key elements in a case (real or hypothetical)
that determine the potential for negligence. TOTO
assesses negligence as well as legal defense to

negligence and reports the results to the user.

111

In order to meet these goals, the following
objectives were stated:

1. To provide a systematic means to evaluate legal
decision making in recreation liability cases. To
understand the problem-solving task.

2. To construct and evaluate a knowledge based
system that is theoretically and practically valid.

3. To create extensive explanation facilities in
the system to enhance understanding of the problem-
solving task by the end user.

These objectives were met in varying degrees.

First objective. The first objective was, in

 

general, satisfied. The software is a systematic model
of potential decision-making processes in negligence
cases. The decision-making processes is disclosed when
the user interacts with the system. Since the reasoning
of TOTO is not static, the evaluation of the decision
making occurs after a session when one views the line of
reasoning report.

Understanding the problem-solving task is the basis
for constructing the system. Here the task is
diagnostic, and requires two levels of reasoning. The
first level is based on rules of law, civil, and
statutory interpretations, and legal rule of thumb. The

second level, bringing meaning to open texture

112

predicates, is more difficult in that there is little
theoretical basis for doing so. This work does not
presume to provide a theoretical basis, but provides a
basis for understanding the difficulties of the problem-

solving task.

Second objective. The second objective was met
given the developmental level of TOTO as a research
prototype. The system successfully identifies key issues
of a case and makes a rough judgment as to the existence
of negligence and defenses. Its performance, however, is
fragile in that it makes mistakes. This is due to three
major reasons:

1. Limited testing and revision of code: Time
constraints limited the amount of evaluation of the
system performance. The specific limitations of
performance are manifest as items 1, 4, 6, and 7 in Table
1 (as discussed earlier). Some areas of the knowledge
base are subsequently more robust than others.

2. The difficulty of bringing meaning to legal
concepts that are contextual and nebulous: The approach
taken to address open texture predicates has no

theoretical basis and requires judgment by the user.

Third objective. The last objective was met in
part. The software shell that TOTO operates within has

facilities to explain the behavior of the system. User

113

session reports and line: of reasoning reports can. be
produced automatically. The user may also stop during a
session. to view the basis for' the current reasoning.
These explanation facilities, however, do not provide the
user with advice on how to assess a question TOTO poses.
It was intended that at any point in a session, the user
could ask for help by simply pressing the "EXPLAIN"
button. Programming this feature is not difficult, yet
is very time consuming. For each state of fact, text
would be entered that could be called by the user. This
would mean entering up to three hundred "explain" files.
This was not achieved due to the time needed to correct
the problems encountered in. programming the knowledge

base.

Conclusions

 

The modeling of decision making in recreation law
via knowledge base programming was an exploratory
pursuit. This exploration yields conclusions that
manifest possibilities rather than facts.

The expert system approach is a viable too for
modeling the legal assessment of negligence and other
types of recreation law. The ability of a system to
identify key legal issues and make recommendations is new

in the field of recreation. This research demonstrates

114

that the technique is feasible with no great expense or
extensive background in computer science.

The method is also a viable means of augmenting risk
management strategies. A fully developed system could
aid providers of recreation services in determining real
or hypothetical negligence. Thus economic, legal, and
ethical benefits would ensue.

The knowledge base of the system is readily
modifiable. This feature yields the potential for
exploring how hypothetical or real changes in the law may
impact the assessment of negligence. This ability would
enhance the ability to assess the impacts of changes is
in civil and statutory law and the subsequent impact upon
recreation providers.

The last conclusion is that there is weak
theoretical basis for bringing meaning to open texture
predicates. For example, the determination of what is
reasonable is rooted in the context of the situation.
Van der Smissen (1990), has suggested that. the
situational elements of' activity, environmental
conditions, and the participants impact the assessment of
what is reasonable. The final assessment is by a jury of
laymen. Attempts to emulate this type of determination
with the expert system approach are simplistic, yet may
hold promise as a research tool to disclose the nature of

this decision making.

115

Limitations

 

The major limitation of this study is the incomplete
testing of the knowledge base. Although the goal of the
research was not to complete a fully implemented system,
doing so would have shed light on a wider array of
problems and possibilities of the approach. This
incomplete testing resulted in fragility of the system
that may be expected at the research prototype stage.

Another limitation of the study is the
implementation of only one strategy to bring meaning to
open texture predicates. This is a result of
undertaking a problem that was too broad and the
limitations of the expert system shell selected. By
focusing on the problem of open texture predicates a
deeper understanding of the problem would have resulted.
The expert system shell selected was suitable for the
easy programming tasks, but not for the difficult ones.
The accessibility and price of the shell were primary
considerations for its selection. Perhaps greater
attention to assessing knowledge representations and

inference strategies would yield improved performance.

Recommendations for Further Study
The limitation of this study imply avenues for
further research. The structure implemented in the

system should be fully developed and tested to enable a

116

complete assessment of the validity and reliability of
the approach. This could be achieved by further formal
testing and field testing by practitioners.
Implementation in the field would provide an opportunity
to assess the usefulness of the approach as perceived by
recreation professionals.

Research to develop and analyze new approaches to
dealing with open texture predicates is in order. As a
fundamental limitation of this system, open texture
predicates deserve focused attention. This could be
driven by thorough analysis of legal theory and
artificial intelligence techniques.

Another avenue for research is the exploration of
emerging artificial intelligence methods to study legal
decision making. Machine learning algorithms and neural
networks may hold promise in understanding the dynamics
of bringing meaning to open texture predicates. These
methods could be utilized to extensively examine facts of
thousands of negligence cases to disclose latent

variables, or patterns that impact legal decision making.

APPENDIX A

TEST CASE 1

118

Line of Reasoning Report for: 'torl.KNB'
01:45:20 PM.

The following goal was pursued :
Setup

The following string fact was obtained :
Today's date = True
string = '8/16/90'

The following string fact was obtained :
Plaintiff's Name = True
string = 'Robbin Droppings“

The following string fact was obtained :
Defendant's Name = True
string = 'Phil Dirt'

The following string fact was obtained :
Agency or firm name = True
string = 'Phil Dirt's Golf Club'

The following Attribute-Value fact was obtained
Activity
Golf = True

RULE: to select activity fired.

As a result the following conclwlon was reached :
Know activity] = True

RULE: for setup fired.

As a resuit the following concimion was reached :
Setup = True

The following goal was pursued :
Sum

The following Attribute-Value fact was obtained
Permission
Yes = True

The following Attribute-Value fact was obtained
The premises
is(are) owned by a private interest = True

RULE: To determine landholder type fired.

As a result the following conclmlon was reached :
Private = True

The following Attribute-Value fact was obtained
Defendant
leases the premises = True

The following “mph my was obtained :

Defendant is in actual control of the premises =

The following amp). m. was obtained :

08/]6/1990

True

Land conditions are managed by the defendant or an employee =

True
RULE: To determine private control fired.

119

120

As a result the following conclmlon was reached :
Control Private = True

As a result the following conch-Ion was reached :
Assessed Defendant = True

The following .ympy. my was obtained :
Defendant is a business = True

The following my... my was obtained :
Defendant charges fee to enter premises = True

RULE: to assess proprietorship fired.

As a result the following conclusion was reached :
Proprietorship = True

The following .ympy. my was obtained :
Plaintiff expressly invited = True

The following ,ympy. my was obtained :
Plaintiff paid a fee to enter prem = True

RULE: for invitee fired.

As a result the following conciuion was reached :
Invitee = True

As a result the following common was reached :
licensee = False

As a resutt the following conclusion was reached :
trespasser = False

The following Attribute-Value fact was obtained :
issue
is of reasonableness = True

RULE: METARULE to determine ordinary type invitee fired.

As a result the following condition was reached :
assess for ordneg = True

RULE: METARULE to assess plaintiff fired.

As a result the following condition was reached :
Plaintiff isa player = True

As a result the following concluion was reached :
Assessed Plaintiff = True

RULE: METARULE Burden of Proof fired.

As a result the following condition was reached :
Assessed Duty = True

As a result the following mission was reached :
Duty = True

The following Attribute-Value fact was obtained :
The defendant
Type 2 = True

RULE: to determine expert knowledge fired.

As a result the following conclmion was reached :
Expert knowledge CF = 100

The following .ympy. my was obtained :
adopted written std = True

121

The following .ympy. my was obtained :

The industry standards are certain. = True
The following .ympy. my was obtained :

The industry standards are uniform. = True
The following .ympy. my was obtained :

The industry standards are well known and obvious. = True
RULE: for industry standard fired.

As a resuit the following common was reached :

Industry standard = True
RULE: Burden of Proof fired.

As a resuit the following emission was reached :

Standard exists = True
The following .ympy. my was obtained :

Def met standard for expert knowledge CF = 43
RULE: to meet standard expert knowledge fired.
As a result the following condition was reached :

Standard = True
RULE: METARULE to determine neg type to Proceed fired.
As a result the following conclmion was reached :

Pmceed = True
The following .ympy. my was obtained :

D inspects CF = 67
The following .ympy. my was obtained :

D repairs CF = 45
The following .ympy. my was obtained :

D removes CF = 46
The following .ympy. my was obtained :

warning = False
The following .ympy. my was obtained :

Instruction = False
RULE: get ordneg facts fired.

As a result the following conchnlon was reached :
have data = True

RULE: to have ordneg fired.

As a result the following conclmion was reached :
ordneg = True

The following my... my was obtained :

Def should anticipate CF = 100
The following .ympy. my was obtained :

injuries have occurred CF = 66
RULE: METARULE to avoid breach deformity fired.
As a result the following conclusion was reached :

there is ordinary breach = True
The following ,ympy. my was obtained :

the plaintiff suffered actual physical harm = True
The following Attribute-Value fact was obtained :

122

Injury occurred
upon the defendants premises = True

RULE: defs prem fired.

As a result the following common was reached :
Injury in space = True

The following .ympy. my was obtained :
the gravity of this harm is substantiated by a physician = True

RULE: to determine injury fired.

As a result the following comm was reached :
Injury valid = True

The following Attribute-Value fact was obtained :
Cause
Type A = True

The following “my. my was obtained :
injury would not have occurred but for conduct = True

The following .ympy. my was obtained :
conduct coterminous with injury = True

The following .ympy. my was obtained :
conduct contact with injury = True

RULE: for Cause in fact fired.

As a result the following common was reached :
Causation f = True

As a result the following conclmlon was reached :
Cause in fact = True

RULE: METARULE to find cause topology fired.

As a result the following conclmlon was reached :
Causation = True

RULE: for liability ordneg fired.

As a result the following condition was reached :
Liability ordneg = True

RULE: output ordneg fired.

As a result the following emission was reached :
Page one response = True

The following numeric fact was obtained :
Months since alleged injury = True
Value = 13.00

The following numeric fact was obtained :
Age = True
Value = 32.00

RULE: to legitimize assumption fired.

As a result the following common was reached :
AssumoRisk Potential = True

The following simple my was obtained 1
Plaintiff signed a waiver or signed release to accept risks = True

The following .ympy. my was obtained :
Language of the waiver is unambiguous = False

 

123

RULE: to sum fired.
As a result the following conclusion was reached :
Sum = True

 

APPENDIX B

TEST CASE 2

124

Line of Reasoning Report for: 'torI.KNB' 01/26/1991
02:15:55 PM.

The following goal was pursued :
Setup

The following string fact was obtained :
Today's date = True
string = '1 1—8-90'

The following string fact was obtained :
Plaintiff's Name = True
string = 'Fred Nugent'

The following string fact was obtained :
Defendant's Name = True
string = 'Edward Maloney'

The following string fact was obtained :
Agency or firm name = True

string = 'Mi DNR'
The following Attribute-Value fact was obtained :
A . .

Swimming = True
RULE: to select activity fired.
As a result the following comm was reached :
Know activity] = True
RULE: for setup fired.
As a result the following “mum was reached :
Setup = True
The following goal was pursued :
Sum
The following Attribute-Value fact was obtained :
Permission
Yes = True
The following Attribute-Value fact was obtained :
The premises
is(are) owned by local, county. state or federal government = True
RULE: To determine landholder type fired.
As a result the following ”my...“ was reached :
Public = True
The following simple fact was obtained :
Defendant manages the premises in the public trust = True
The following Attribute-Value fact was obtained :
Defendant is
the enterprise administrator = True
RULE: METARULE To determine public control fired.
As a result the following concym was reached :
Control Public = True
The following simple fact was obtained :
Defendant is a business = False
The following simple tact was obtained :
Plaintiff no business transaction = Tme

125

1 2 6

The following “mph fact was obtained :
No monetary transaction occurred for recreation = True

RULE: for licensee fired.

As a result the following comm was reached :
licensee = True

As a result the following comm was reached :
Invitee = False

As a result the following comm was reached :
trespasser = False

The following Attribute-Value fact was obtained :
issue
is of willfull and wanton misconduct = True

RULE: METARULE to determine neg type to Proceed fired.

As a result the following comm was reached :
Proceed = True

RULE: to determine willwant for no invitee fired.

As a result the following comm was reached :
assess for willwant = True

RULE: METARULE to assess plaintiff fired.

As a result the following comm was reached :
Plaintiff isa player = True

As a result the following comm was reached :
Assessed Plaintiff = True

The following simple fact was obtained :
Defendant participates, ratified. condones tortious act = True

The following simple fact was obtained :
Negligence arose from hiring, = False

RULE: to asses public administrator fired.

As a result the following ”..ch was reached :
Administrator not responsible = True

RULE: METARULE to make agency defendant fired.

As a result the following cm...” was reached :
Identified defendant = True

As a result the following comm was reached :
Def is
Agency at large = True

As a result the following comm was reached :
Assessed Defendant = True

RULE: METARULE Burden of Proof fired.

As a result the following comm was reached :
Assessed Duty = True

As a resutt the following comm was reached :
Duty = True

The following simple fact was obtained :
D had knowledge CF = 100

The following simple fact was obtained :
D had ability to avoid CF = 70

The following slmpb incl was obtained :

1 2 7

D's omission CF = 60

The following simple fact was obtained :
D Conduct shows intent CF = 0

The following simple fact was obtained :
D Conduct shows indifference CF = 90

RULE: to assess intent or indifference fired.

As a result the following comm" was reached :
lorD = True

RULE: for willful wanton misconduct fired.

As a result the following cmjmn was reached :
assessed willwant facts = True

RULE: to math] willwant facts fired.

As a result the following comm was reached :
do mathl = True

RULE: to math2 willwant facts fired.

As a result the following condom, was reached :
do math2 = True

RULE: to pick wilwant math formulaior2 fired.

As a result the following conch...“ was reached :
Have value = True

RU LE: to accept Wilwant fired.

As a result the following comm was reached:
willful and wanton misconduct by the defendant: True

The following simple incl was obtained:
the plaintiff suffered actual physical harm = True

The following Attribute-Value fact was obtained :
injury occurred
upon the defendants premises = True

RULE: defs prem fired.

As a resuit the following comm was reached :
Injury in space = True

The following simple fact was obtained :
the gravity of this harm is substantiated by a physician = True

RULE: to determine injury fired.

As a result the following cmlm was reached :
injury valid = True

The following Attribute-Value fact was obtained :
Cause
Type A = True

The following sliano fact was obtained :
injury would not have occurred but for conduct = True

The following simple fact was obtained :
conduct coterminous with injury = True

The following simple fact was obtained :
conduct contact with injury = True

RULE: for Cause in fact fired.

As a result the following “new“, was reached :
Causation f = True

As a result the following comm was reached :

1 2 8
Cause in fact = True

RULE: METARULE to find cause topology fired.

As a result the following comm was reached :
Causation = True

RULE: for liability grosneg fired.

As a result the following comm was reached :
Liability grosneg = True

RULE: output grosneg fired.

As a resutt the following conclusion was reached :
Page one response = True

The following numeric fact was obtained :
Months since alleged injury = True
Value = 15.00

The following “mole fuel was obtained :
The plaint contributes = True

The following simple fact was obtained :
P Conduct CF = 69

The following simple loci was obtained :
P knowledge CF = 75

The following simple fact was obtained :
P ability to avoid CF = 83

The following simple fact was obtained :
P omission CF = 90

RULE: for comparative negligence weight fired.

As a result the following common was reached :
Comparative negligence assessed = True

RULE: for existence of comparative negligence fired.

As a result the following comm was reached :
CN exists = True

RULE: to invoke comparative negligence fired.

As a result the following cmjm was reached :
Partial Liability = True

As a result the following cm.% was reached :
Page two response = True

RULE: to sum fired.

As a result the following comm was reached :
Sum = True

APPENDIX C

TEST CASE 3

129

Line of Reasoning Report for: 'tort.KNB'
01:22:03 PM.

The following goal was pursued :
Setup

The following String fact was obtained :
Today's date = True
String = '8-23-90'

The following String fact was obtained :
Plaintiff‘s Name = True
String = 'Thomas Rhude'

The following String fact was obtained :
Defendant's Name = True
String = 'Bud Jordahl'

The following String fact was obtained :
Agency or firm name = True
String = 'Bud's Gravel Co.‘

The following Attribute-Value fact was obtained
Acliviiy
Fishing = True

RULE: to select activity fired.

As a result the following conclusion was reached :
Know activityl = True

RULE: for setup fired.

As a result the following conclusion was reached :
Setup = True

The following goal was pursued :
Sum

The following Attribute-Value fact was obtained
Permission
No = True

The following Attribute-Value fact was obtained
The premises
is(are) owned by a private interest = True

RULE: To determine landholder type fired.

As a resuii the following conclusion was reached :
Private = True

The following Attribute-Value fact was obtained
Defendant
is owner of the premises = True

The following smug fact was obtained :

Defendant is in actual control of the premises =

The following simpje fact was obtained :

08/ 23/ I 990

True

Land conditions are managed by the defendant or an employee =

True
RULE: To determine private control fired.

130

131

As a resutt the following conclusion was reached :
Control Private = True

As a result the following conclusion was reached :
Assessed Defendant = True

RULE: for Trespasser fired.

As a result the following conclusion was reached :
trespasser = True

As a result the following conclusion was reached :
Invitee = False

As a result the following conclusion was reached :
licensee = False

This: following Attribute—Value fact was obtained :

ue

is of reasonableness = True

RULE: to determine willwant for no invitee fired.

As a result the following conclusion was reached :
No action = True

A CYCLE command was performed.

The following goal was pursued :
Sum

The following Attribute-Value fact was obtained :
Issue
is of willfull and wanton misconduct = True

RULE: to determine willwant for no invitee fired.

As a result the following conclusion was reached :
assess for willwant = True

RULE: METARULE to determine neg type to Proceed fired.

As a result the following conclusion was reached :
Proceed = True

RULE: METARULE to assess plaintiff fired.

As a result the following conclusion was reached :
Plaintiff isa player = True

As a result the following conclusion was reached :
Assessed Plaintiff = True

RULE: METARULE Burden of Proof fired.

As a result the following conclusion was reached :
Assessed Duty = True

As a result the following conclusion was reached :
Duty = True

The following Attribute-Value fact was obtained :
Trespass
was frequent = True

RULE: to determine trespasser type fired.

As a resuit the following conclusion was reached :
Have trespass type = True

The following gmpje fact was obtained :
D is aware of frequent intrusion in a specific area = True

132

The following mph fact was obtained :

active Operations creates risk for bodily harm to T = True
The following ample fact was obtained :

D not warn T of artificial conditions that create the danger = True
The following mpg fact was obtained :

D knows that conditions will not be discovered by T = True
RULE: for Trespasser Breach for frequent trespasser fired.
As a result the following conclusion was reached :

Assessed Plaintiff = True
As a result the following conclusion was reached :

trespasserl = True
As a result the following conclusion was reached :

willful and wanton misconduct by the defendant = True
The following 34mph fact was obtained :

the plaintiff suffered actual physical harm = True
The following Attribute-Value fact was obtained :

Injury occurred

upon the defendants premises = True
RULE: defs prem fired.

As a result the following conclusion was reached :

Injury in space = True
The following slmp'e fact was obtained :

the gravity of this harm is substantiated by a physician = True
RULE: to determine injury fired.

As a result the following conclusion was reached :

Injury valid = True
The following Attribute-Value fact was obtained :

Cause

Type C = True
The following gmpje fact was obtained :

lntervening cause was foreseeable: True
RULE: for Proximate Cause intervening cause fired.

As a result the following conclusion was reached .

Causation i- - True
As a result the following conclusion was reached :

PC foreseeable = True
RULE: METARULE to find cause topology fired.

As a result the following conclusion was reached :

Causation = True
RULE: for liability grosneg fired.

As a result the following conclusion was reached :

Liability grosneg = True
RULE: output grosneg fired.

As a result the following conclusion was reached :

Page one response = True
The following numeric fact was obtained :

133

Months since alleged injury = True
Value = 14.00
The following mpg fact was obtained :
The plaint contributes = True
The following simple fact was obtained :
P Conduct CF = 87
The following simple fact was obtained :
P knowledge CF = 73
The following “mph fact was obtained :
P ability to avoid CF = 52
The following simple fact was obtained :
P omission CF = 96
RULE: for comparative negligence weight fired.
As a resuit the following conclusion was reached :
Comparative negligence assessed = True
RULE: for existence of comparative negligence fired.
As a result the following conclusion was reached :
CN exists = True
RULE: to invoke comparative negligence fired.
As a result the following conclusion was reached :
Partial Liability = True
As a result the following conclusion was reached :
Page two response = True
RULE: to sum fired.
As a result the following conclusion was reached :
Sum = True

APPENDIX D

EDITED KNOWLEDGE BASE IN TOTO

134

TITLE TOTO

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

SETUP RULES
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE for setup

IF PAINT TOTO.PICT

AND ASK Today's date

AND ASK Plaintiff’s Name <>"

AND ASK Defendant's Name <>"
AND ASK Agency or firm name <>"
AND Know activity]

THEN Setup

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
CONTROL RULE CONTROL RULE CONTROL RULE

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE to sum

IF Page one response

AND Page two response

OR NOT Page two response

THEN Sum

AND FILE LORR line 0 reasoning
IIIlIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIlIlIII

RULE output ordneg
IF Liability ordneg
THEN Page one response

RU LE output grosneg
IF Liability grosneg
THEN Page one response

RULE output no duty

IF No Liability ordneg

OR No Liability grosneg

AND NOT Duty

THEN Page one response
AND DISPLAY No burden duty

135

136

RULE output no std

IF No Liability ordneg

AND NOT Standard

THEN Page one response
AND DISPLAY No burden std

RULE output no ordbreach

IF No Liability ordneg

AND NOT there is ordinary breach
THEN Page one response

AND DISPLAY No burden ordbreach

RULE output no grossbreach

IF No Liability grosneg

AND NOT willful and wanton misconduct by the defendant
THEN Page one response

AND DISPLAY No burden grossbreach

RULE output no injury

IF No Liability ordneg

OR No Liability grosneg

AND NOT Injury valid

THEN Page one response
AND DISPLAY No burden injury

RULE output no cause

IF No Liability ordneg

OR No Liability grosneg

AND NOT Causation

THEN Page one response
AND DISPLAY No burden cause

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
BIG RULES
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE for no liability ordneg

IF assess for ordneg

AND NOT Duty

OR NOT Standard

OR NOT there is ordinary breach
OR NOT Injury valid

OR NOT Causation

THEN No Liability ordneg

137

RULE for liability ordneg

IF assess for ordneg

AND Duty

AND Standard

AND there is ordinary breach
AND Injury valid

AND Causation

THEN Liability ordneg

RULE for no liability grosneg

IF Proceed

AND assess for willwant

AND NOT Duty

OR NOT willful and wanton misconduct by the defendant
OR NOT injury valid

OR NOT Causation

THEN No Liability grosneg

RULE for liability grosneg

IF Proceed

AND assess for willwant

AND Duty

AND willful and wanton misconduct by the defendant
AND Injury valid

AND Causation

THEN Liability grosneg

RULE to reject trespasser ordneg
lF trespasser

AND issue\ is of reasonableness
THEN Reject

AND DISPLAY tres no ordneg

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\1\\\\\\\\\\\\\\\\\\\\\\\\\\\
Implement Defenses
I\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\I\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE to invoke statute of limitation
IF Liability ordneg

OR liability grosneg

AND Statute of limitation imposed
THEN No liability

AND Page two response

AND DISPLAY stat Iim

138

RULE to invoke rec use statute
IF Liability ordneg

AND invoke rec use statute
THEN No liability

AND Page two response
AND DISPLAY rec use

RULE to invoke valid waiver
IF Liability ordneg

AND valid waiver

THEN No liability

AND Page two response
AND DISPLAY waiver

RULE to invoke comparative negligence
IF Liability grosneg

AND CN exists

THEN Partial Liability

AND LL:=100-CONF(CN)

AND Page two response

AND DISPLAY compneg

RULE to acuate comparative negligence I II!!!
IF CONF (LL)> 50
THEN CONF LL=50

RULE to invoke Assumption of risk
IF Liability grosneg

AND Assumption of risk

THEN Partial Liability

AND DISPLAY assrisk

AND Page two response

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
DUTY
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE METARULE Burden of Proof-Duty
IF Control Private

OR Control Public

AND Assessed Plaintiff

AND Assessed Defendant

THEN Assessed Duty

AND Duty

139

\\\\\\\\\\\\\\\\\I\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
Determine Defendant(s)
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE METARULE to make agency defendant
IF Board member not responsible

OR Administrator not responsible

OR Employee not responsible

OR Volunteer not responsible

THEN Identified defendant

AND Def is \ Agency at large

AND Assessed Defendant

RULE METARULE to make individual responsible
IF Public

AND Indiv defendant\ Board member

OR Indiv defendant\ Administrator

OR Indiv defendant\ Employee

OR Indiv defendant\ Volunteer

THEN Identified defendant

AND Def is \ Agency and Individual

AND Assessed Defendant

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
Classify Players
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE METARULE to assess plaintiff
IF Invitee

OR licensee

OR trespasser

THEN Plaintiff isa player

AND Assessed Plaintiff

RULE METARULE To determine public control

IF Public

AND Defendant manages the premises in the public trust
AND Defendant is \ on the board of directors

OR Defendant is \ the enterprise administrator

OR Defendant is\ an employee

OR Defendant is \ a volunteer

THEN Control Public

RULE To determine landholder type
IF The premises IS owned by a private interest
THEN Private

140

RULE To determine landholder type

IF The premises IS owned by local. county. state or federal
government

THEN Public

RULE To determine private control

IF Private

AND Defendant \ is owner of the premises

OR Defendant \ leases the premises

OR Defendant \ is an independent contractor on the premises
AND Defendant Is in actual control of the premises

AND Land conditions are managed by the defendant or an
employee

THEN Control Private

AND Assessed Defendant

RULE to assess public board member

IF Control Public

AND Defendant is\ on the board of directors
AND The negligent act is within scope

THEN Board member not responsible

AND Authority for conduct was discretionary

RULE to make board member a player
IF NOT Board member not responsible
THEN Individual defendant

AND Indiv defendant\ Board member

RULE to asses public administrator

IF Control Public

AND Defendant is \ the enterprise administrator

AND NOT Defendant participates. ratified. condones tortious act
OR NOT Negligence arose from hiring.

OR NOT Negligence arose from training

OR NOT Negligence arose from retaining an employee

THEN Administrator not responsible

RULE to make administrator a player
IF NOT Administrator not responsible
THEN Individual defendant

AND Indiv defendant\ Administrator

141

RULE to assess public individual

lF Control Public

AND Defendant is \ an employee

AND NOT Employee Is responsible for tortious act
AND The negligent act is within scope

AND Employee's act was while on duty

THEN Employee not responsible

RULE to make employee a player
IF NOT Employee not responsible
THEN Individual defendant

AND indiv defendant\ Employee

RULE to assess contractor

IF Control Public

AND Defendant is \ an independent contractor on the premises
AND Tortious condition is created by independent contractor
AND Agency does not have significant control over tASK

OR Agency does not retain supervision Ior control over the
employees!

THEN Contractor responsible

AND Indiv defendant\ Contractor

RULE to asses volunteer

IF Control Public

AND Defendant is \ a volunteer

AND NOT Volunteers actions created tortious condition
OR Volunteer's actions created tortious condition

AND The negligent act is within scope

THEN Volunteer not responsible

RULE to make volunteer a player
IF NOT Volunteer not responsible
THEN Individual defendant

AND indiv defendant \ Volunteer

RULE to assess proprietorship

lF Public

OR Private

AND Defendant Is a business

AND Defendant charges fee to enter premises

OR Defendant charges fee for activity on premises
THEN Proprietorship

142

RULE for invitee

IF Permission \ Yes

AND Control Public

OR Control Private

AND Proprietorship

AND Plaintiff expressly invited
AND Plaintiff paid a fee to enter prem
OR Plaintiff paid a fee for activity
THEN Invitee

AND NOT licensee

AND NOT trespasser

RULE for licensee

IF Control Public

OR Control Private

AND Permission \ Yes

AND Plaintiff no business transaction

AND No monetary transaction occurred for recreation
THEN licensee

AND NOT Invitee

AND NOT trespasser

RULE for minor
IF Age<13
THEN minor

RULE for Trespasser
IF Permission \ No
THEN trespasser
AND NOT Invitee
AND NOT licensee

\\\\\\\\\\\\\\\\11\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
Standard
I\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE to meet standard common knowledge
IF Common Knowledge
THEN Standard

RULE to meet standard expert knowledge

IF Standard exists

AND Expert knowledge

AND Def met standard for expert knowledge
THEN Standard

143

RULE Burden of Proof-breach

IF Invitee

AND Common knowledge

AND CONF (Common knowledge) >50
AND Industry standard

OR Industry custom

THEN Standard exists

RULE Burden of Proof-breach2

IF Invitee

AND Expert knowledge

AND CONF (Expert knowledge) >50
AND Industry standard

OR Industry custom

THEN Standard exists

RULE to determine common knowledge
IF The defendant \ Type 1
THEN Common Knowledge

RULE to determine expert knowledge
IF The defendant \ Type 2
THEN Expert knowledge

RULE for industry standard
. IF adopted written std
AND The industry standards are certain.
AND The industry standards are uniform.
AND The industry standards are well known and obvious.
THEN Industry standard

RULE for industry custom

IF Customs

AND The industry customs are certain.

AND The industry customs are uniform.

AND The industry customs are well known and obvious.
THEN Industry custom

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
ordinary or willful and wanton
I\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE METARULE to determine neg type to Proceed
iF NOT No action

OR assess for ordneg

OR assess for willwant

THEN Proceed

144

RULE METARULE to determine ordinary type invitee
IF Invitee

AND issue \is of reasonableness

THEN assess for ordneg

RULE to determine willwant type invitee

lF Invitee

AND issue \is of willfull and wanton misconduct
THEN assess for willwant

RULE to determine willwant for no invitee

IF licensee

OR trespasser

AND issue\is of willfull and wanton misconduct
THEN assess for willwant

RULE to determine willwant for no invitee
IF licensee

OR trespasser

AND issue \is of reasonableness

THEN No action

AND DISPLAY not actionable

AND FORGET Issue

AND CYCLE

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
ORDBREACH
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE METARULE to avoid breach deformity
IF Proceed

AND assess for ordneg

AND ordneg

AND NOT Reasonable

THEN there is ordinary breach

RULE METARULE to deform breach
IF Proceed

AND assess for ordneg

AND not ordneg

AND Reasonable

THEN NOT there is ordinary breach

1.45

RULE METARULE to deform breach
IF Proceed

AND assess for ordneg

AND not ordneg

AND NOT Reasonable

THEN NOT there is ordinary breach

RULE get ordneg facts

IF Invitee

AND D inspects

AND D repairs

OR D removes

OR D clear warning

THEN have data

AND 06 :=(CONF(D inspects)+CONF(D repairs)+CONF(D removes):
+CONF(D clear waming))/3

RULE to have ordneg
IF have data
AND CONF(OG) <= 50
THEN ordneg

RULE to reject ordneg
IF have data

AND 06 > 50

THEN not ordneg

RULE Failure to warn or instruct

IF warning !or instruction!

OR Instruction

THEN D clear warning

AND CONF (D clear waming):=100

RULE for REASONABILITY

IF Invitee

lAND Probability!

AND Magnitude of risk

AND Burden of attemative conduct
AND CONF(M) >CONF(B)

THEN NOT Reasonable

RULE for UNREASONABILITY

IF Magnitude of risk

AND Burden of alternative conduct
AND CONF (M) < CONF(B)

THEN Reasonable

146

RULE for foreseeable probability

IF Def should anticipate

AND Injuries have occurred

AND common knowledge

THEN Probability

AND P := (CONF (Def should anticipate)+ CONF (Injuries have
occurred):

+ CONF(common knowledge» /3

RULE for magnitude of risk

IF Probability

AND G

THEN Magnitude of risk

AND M := (CONF(P)+CONF(G)) l2

RULE for Burden of alternative conduct

IF U

AND A Ifeasibility of attemative!

AND C Irelative cost of safer conduct!

AND S Irelative utility of safer conduct!

AND 08 Irelative safety of alternative conduct!

THEN Burden of alternative conduct

AND B := ((CONF (A)+CONF (C)+CONF (S)+CONF (08)) M) - CONF
(U)

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
CAUSATION
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE METARULE to find cause topology
IF Causation f

OR Causation u

OR Causation i

THEN Causation

RULE for Cause in fact

IF Cause\ Type A

AND Injury would not have occurred but for conduct
AND conduct coterminous with Injury

AND conduct contact with injury

OR avoidance of conduct yields injury

THEN Causation f

AND Cause in fact

147

RULE for Proximate Cause unforeseeable consequences
IF Cause \ Type B

AND consequences of D conduct were foreseeable
AND P was within the foreseeable zone of danger

OR P Isa rescuer

THEN Causation u

AND PC foreseeable

RULE for Proximate Cause intervening cause

IF Cause \ Type C

AND lntervening cause was foreseeable

OR Non extraordinary weather conditions or changes
OR Third party negligence

OR Third party criminal conduct

OR P isa rescuer

THEN Causation i

AND PC foreseeable

RULE to determine 3rd party criminal conduct
IF Criminal action by third party involved

AND D conduct exposed P to that risk

THEN Third party criminal conduct

RULE to determine rescuer
IF P was attempting rescue as a result of conduct
THEN P isa rescuer

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
INJURY
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE to determine injury

IF the plaintiff suffered actual physical harm

AND Injury in space

AND the gravity of this harm is substantiated by a physician
THEN Injury valid

RULE defs prem\ adj prem

IF NOT Injury occurred \on premises adjacent to defendants premises
OR Injury occurred \upon the defendants premises

THEN injury in space

RULE to validity adajacent

IF Injury occurred \on premises adjacent to defendants premises
AND Defendant control of adj premisis

THEN Injury valid

148

RULE to validify def control of adj prem

IF Defendant leases adjacent premisis

OR Def retains control overland conditions on adjacent prem
AND The hazard in question lies on adjacent premisis

THEN Defendant control of adj premisis

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
WILLFUL/ WANTON BREACH
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE for willful wanton misconduct
IF Proceed

AND assess for willwant

AND licensee

OR Invitee

AND D had knowledge

AND D had ability to avoid

AND D's omission

AND lorD

THEN assessed willwant facts

RULE to assess intent or indifference
IF D Conduct shows intent

OR D Conduct shows indifference
THEN lorD

RULE to mathl willwant facts

IF assessed willwant facts

THEN do mathl

AND Wilwant := (CONF (D had knowledge) + CONF (D had ability to
avoid):

+CONF (D's omission) + CONF (D Conduct shows intent))/4

RULE to math2 willwant facts

IF assessed willwant facts

THEN do math2

AND Wilwant := (CONF (D had knowledge) + CONF (D had ability to
avoid):

+CONF(D's omission) + CONF (D Conduct shows indifference))/4

RULE to pick wilwant math formulalor2
IF do mathl

OR do math2

THEN Have value

149

RULE to reject Wilwant

IF Have value

AND CONF (Wilwant) < 50

THEN NOT willful and wanton misconduct by the defendant

RULE to accept Wilwant

IF Have value

AND CONF (Wilwant) >50

THEN willful and wanton misconduct by the defendant

\\\IIIIIIIIIIIIIIIIIIIIIIII\IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII\IIII
WIIlwant Trespasser
IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII\\\\I\IIIIIIII

RULE to determine trespasser type

IF Proceed

AND assess for willwant

AND trespasser

AND Trespass\ was frequent

OR Trespass\ discovered by defendant
OR Trespass\ involves a minor

OR Trespass\ none of the above

THEN Have trespass type

IIIIIIIIIIIIIIIII\II\\IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
Wilwant for normal trespasser
\IIIIIIIIIIIIIIIIIIIIIIII\IIIIIIII\\\\\I\\I\\\I\I\I\\\\III\I\III\I

RULE for willful wanton misconduct
IF Proceed

AND assess for willwant

AND Have trespass type

AND Trespass\ none of the above
AND D had knowledge

AND D had ability to avoid !
AND D's omission

AND lorD

THEN assessed willwant facts

RULE to assess Intent or indifference
IF D Conduct shows Intent

OR D Conduct shows indifference
THEN lorD

150

RULE to mathl willwant facts

IF assessed willwant facts

THEN do mathl

AND Wilwant := (CONF (D had knowledge) + CONF (D had ability to
avoid):

+CONF (D‘s omission) + CONF (D Conduct shows intent))/4

RULE to math2 willwant facts

IF assessed willwant facts

THEN do math2

AND Wilwant := (CONF (D had knowledge) + CONF (D had ability to
avoid):

+CONF(D's omission) + CONF (D Conduct shows indifference))/4

RULE to pick wilwant math formulalor2
IF do mathl

OR do math2

THEN Have value

RULE to reject WIlwant

IF Have value

AND CONF (Wilwant) < 50

THEN NOT willful and wanton misconduct by the defendant

RULE to accept erwant

IF Have value

AND CONF (Wilwant) >50

THEN willful and wanton misconduct by the defendant

\IIIIIIIIIII\II\IIIIII\IIIIIIIIIIIIIIIIIIIIIII\\\III\\\III\IIIIIII
willwant for frequent, discovered or minor
IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

RULE for Trespasser Breach for frequent trespasser

IF Have trespass type

AND Trespass\ was frequent

AND D is aware of frequent intrusion in a specific area

AND active operations creates risk for bodily harm to T

AND D not warn T of artificial conditions that create the danger
AND D knows that conditions will not be discovered by T

THEN Assessed Plaintiff

AND trespasser]

AND willful and wanton misconduct by the defendant

151

RU LE for Trespasser Breach 2

IF Have trespass type

AND Trespass\ discovered by defendant

AND D discovers T on the premises

AND D continues conduct that creates risk for injury to T

AND NOT natural conditions on land

AND D knows that hazardous cond will not be discovered by 1
AND D did not control elements within control to limit risk

OR D did not warn T of artif cond on the land that create risk
THEN Assessed Plaintiff

AND trespassefz

AND willful and wanton misconduct by the defendant

RULE for Trespasser Breach 3

IF Have trespass type

AND Trespass\ involves a minor

AND minor

AND D knows or has reason to know of T is likely
AND D is aware of bodharm condition to minor T
AND NOT natural conditions on land

AND condition creates risk of bodily harm or death
AND D not control within power to limit risk

THEN Assessed Plaintiff

AND trespasser3

AND willful and wanton misconduct by the defendant

RULE for Landhoider natural conditions

IF land is in a natural condition

AND Land has not been altered by D or others
THEN natural conditions on land

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
DEFENSES
\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

RULE Statute of limitations

IF Months since alleged Injury >36
AND Age >18

THEN Assessed Defenses

AND Statute of limitation imposed

152

RULE rec use statute

IF Control Private

OR Control Public

AND NOT Invitee

AND Plaintiff was on def prem for 0 rec purpose

AND NOT Social guest

AND NOT Fishing from nav h20

AND No monetary transaction occurred for recreation
AND a state park

THEN invoke rec use

RULE to determine state park

IF Def prem 0 MI state park

AND The injury occurred within park boundaries
THEN a state park

RULE to deten'nine fishing exclusion

IF Activity \ Fishing

AND NOT a private lake

AND NOT there are streams entering or leaving the lake
THEN Fishing from nav h20

RULE for existence of comparative negligence

IF The plaint contributes

AND Comparative negligence assessed

THEN CN exists

AND CN := (CONF(P Conduct)+CONF(P knowledge)+CONF(P ability
to avoid):

+CONF(P omission))/4

RULE for comparative negligence weight
IF P Conduct shows indifference
AND P knowledge

AND P ability to avoid

AND P omission

THEN Comparative negligence assessed

RULE for rec activity
IF Know activityl
THEN Plaintiff was on def prem for a rec purpose

153

RULE DEFENSES Waiver

IF AssumoRisk Potential

AND Plaintiff signed a waiver or signed release to accept risks
OR Plaintiff agreed verbally to accept risks

AND Language of the waiver is unambiguous

AND release includes hazard type in question

THEN Assessed Defenses

AND valid waiver

AND Express assumption of risk

AND D liable for grossneg

RULE to legitimize assumption-o—risk
IF NOT minor

OR NOT Parental consent

THEN AssumoRisk Potential

RULE Expressed assumption of risk

IF AssumoRisk Potential

AND Plaintiff read waming signs prior to activity

AND Plaintiff understood and appreciated the warning
THEN Expressed assumption of risk

RULE Implied assumption of risk

IF AssumoRisk Potential

AND P knows. appreciates and understands existence of the risk
AND P voluntarily entered into the risk situation

AND P conduct manifests consent

THEN Implied assumption of risk

RULE METARULE DEFENSES assumption of risk
IF Expressed assumption of risk

OR Implied assumption of risk

THEN Assessed Defenses

AND Assumption of risk

APPENDIX E

PORTION OF TOTO'S KNOWLEDGE TREE

154

Knowledge Base : TORT
Compled:07/23/1990 10:43:39 AM.

1. Setup
Fraane:forsetIp

Uses fact : Know activityl
From rule : to select activity

M question = Activity Hmthg
A“ W = Activity Playground
m ”"0" = Activity XC Skiing
“‘3 question = Activity Alphe Sklng
Asia cuestton : ACTIVITY Skating
“6 question = Activity Golf
1.1. Sun
From rile : to sun
From rule : Page one response
Fran rule: output adneg
Uses fact : Liability ordneg
From rile : for liability adneg
Uses fact : assess for ordneg
From Me: METARULEto determine ordinary type Invitee
Uses fact : Invitee
Fran n19 : for Invitee
Asia question: Permission Yes
Uses fact : Control PIbiic
From Me: METARULETo determine public control
Uses fact : PIbllc
From Me: To determine landholder type
qu= The premises owned by local. county. state or federal government
M wallow Defendant manages the premises h the public trust
Mmbm Defendant Is on the board of directors
M0 MM= Defendant Is the enterprise admhlstrator
Mm= Defendant Is an employee
M3 W= Defendant is a volunteer
Uses fact : Control Private
Fran rile: To determine private control
Uses fact : Private
Fran i119: To determine landholder type
Ms question: The premises owned by a private interest
Nils mallow Defendant ls owner of the prernlses
mm= Defendant leases the premises
Asia mostlon: Defendant Is an Independent contractor on the premises
M question = Defendant Is In actual control of the premises
MW= Land conditions are managed by the defendant or an employee
Uses fact : Proprietorshp
From Me: to assess proprietorship
Uses fact : Plbllc
Fran rile: To determine landholder type
MW: The premises owned by local. county. state or federal government
Uses fact : Private

Fran rile: To determine landholder type
MW= The premises owned by a private Interest
MW= Defendant Is a business

155

156

Alia Walton: Defendant charges fee to enter premises
Alia dilation: Defendant charges fee for activity on premises
Ms W = Plaintiff expressly Invited
mutation: Plaintiff paid a fee to enter prem
Asioqucstlom Plaintiff paid a fee for activity
Fran rUe : for licensee
was fact : Control Plbllc
Fran rile: METARULETo determine prllc control
Uses fact : PIbllc
Fran rile: To determine landholder type
Mention: The premises owned by local. comfy. state or federal government
MW= Defendant manages the premises In the public trust
MW: Defendant Is on the board of directors
No question = Defendant Is the enterprise administrator
Mm= Defendant Is an employee
Ms W: Defendant Is a volunteer
Uses fact : Control Private
Fran Me: To determine private control
Uses fact : Private
From Me: To determine landholder type
M question: The premises owned by a private interest
As"! m= Defendant ls owner of the premises
Mmllon: Defendant leases the premises
M WU“ Defendant Is an Independent contractor on the premises
M Walton: Defendant Is In actual control of the premises
“swallow Land conditions are managed by the defendant or an employee
M quasiiom Permission Yes
MW: Plaintiff no business transaction
Ask-squash: No monetary transaction occurred for recreation
Fran rile : for Trespasser
No question: Permission No
Mutation: Issue Is of reasonableness
Uses fact : Duty
Fran rile: METARULEBurden of Proof
Uses fact : Control Private
Fran rile: To deterrnlne private control
Uses fact : Private
Fran Me: To determine landholder type
No Mb": The premises owned by a private Interest
Moussflom Defendant Is owner of the premises
Moussflom Defendant leases the premises
Nils wallow Defendant Is an Independent contractor on the premises
mm: Defendant is In actual control of the premises
Mowiiom Land conditions are managed by the defendant or an employee
Uses fact : Control Plbllc
Fran i119: METARULETo determine pubic control
Uses fact : Piblc

From rule : To deterrnlne landholder type
Mow-lion: The premises owned by local. comfy. state or federal government
Mouosiiom Defendant manages the premises In the public trust
M questioni Defendant Is on the board of directors
Moussflon: Defendant Is the enterprise administrator
AsloouesIbn: Defendant Is an employee
MW= Defendant Is a volunteer
Uses tact : Asseaed Plaintiff
Fran rile: METARULEto assess plaintiff
Uses fact : Invitee
Fran Me: for invitee

MlBursitis":Permission Yes 157

Uses fact : Control Plbllc
Fran rUe: METARULETo determine public control
Uses fact : Pibllc
Fraane: To determine landholder type
“is W = The premises owned by local. county. state or federal government
Moussfiom Defendant manages the premises In the public trust
Mowiiom Defendant Is on the board of directors
M9W= Defendant Is the enterprise administrator
MW: Defendant Is an employee
Moussflon: Defendant Is a volunteer
Uses fact : Control Private
From Me: To determine private control
Uses tact : Private
From Me: To determine landholder type
Ms question: The premises owned by a private Interest
MW= Defendant Is owner of the premises
A“ quaibm Defendant leases the premises
Mowibm Defendant Is an Independent contractor on the premises
“0 9W= Defendant Is In actual control of the premises
MOWM= Land conditions are managed by the defendant or an employee
Uses fact : Proprietorship
Fran rile: to assess proprietorship
Uses fact : PIbiic
Fran Me: To determine landholder type
Mansion = The premises owned by local. county. state or federal government
Uses fact : Private
Fran rile: To determine landholder type
No Milo" = The premises owned by a private Interest
mm= Defendant Is a business
Mousslbm Defendant charges fee to enter premises
“Omission: Defendant charges fee for activity on premises
M“ “"0" = Plaintiff expressly Invited
Asks question: Plaintiff paid a fee to enter prem
M question = Plaintiff paid a fee for activity
Fran i119: for licensee
Uses fact : Control Ptbilc
From Me: METARULETo determine prIIc control
Uses fact : PLbIlC

Fran rile: To determine landholder type
MW: The premises owned by local. comfy. state or federal government
MGW= Defendant manages the premises In the public trust
“swallow Defendant Is on the board of directors
“swallow Defendant Is the enterprise administrator
MWM= Defendant Is an employee
Ms oussilom Defendant Is a volunteer
Uses fact : Control Private
Fran Me: To determine private control
Uses fact : Private
Fran rile: To determine landholder type
M6 W= The premises owned by a private Interest
Aslsquosiion: Defendant Is owner of the premises
M quosibm Defendant leases the premises
Miami Defendant Is an Independent contractor on the premises
Moussflom Defendant Is In actual control of the premises
“equation: Land conditions are managed by the defendant or an employee
m mallow Permission Yes
Asks mails": Plaintiff no business transaction

158

m W= No monetary transaction occurred for recreation
Fran Me: for Trespasser
m mallow Perrnlsslon No
Uses fact : Icensee
Fran Me: for Invitee
Nils Gallon: Permission Yes
Uses fact : Control Public
Fran Me: METARULETo determine public control
Uses fact : PIbIIc
Fran Me: To determine landholder type
“satiation: The premises owned by local. comfy. state or federal government
Mm= Defendant manages the premises In the public trust
Moussiiom Defendant Is on the board of directors
Moduuilon: Defendant Is the enterprise administrator
qu= Defendant Is an employee
“Nation: Defendant Is a volunteer
Uses fact : Control Private
Fran Me: To determine private control
Uses fact : Private
Fran Me: To determine landholder type
“*3 ”mi The premises owned by a private Interest
Mutation: Defendant Is owner of the premises
Mumbm Defendant leases the premises
MOWM= Defendant Is an Independent contractor on the prernlses
Moussiiom Defendant Is In actual control of the premises
Moussflom Land conditions are managed by the defendant or an employee
Uses fact : Proprietorship
Fran Me: to assess proprietorship
Uses fact : PLbIIc
Fran Me: To deferrnlne landholder type
m Milo": The premises owned by local. county. state or federal government
Uses fact : Private

From rule : To determine landholder type
Mm= The premises owned by a private Interest
Mquosiiom Defendant Is a business
Mowbm Defendant charges fee to enter prernlses
MW= Defendant charges fee for activity on premises
Asia question = Plaintiff expressly Invited
Asia motion: Plaintiff paid a fee to enter prem
M “lion = Plaintiff paid a fee for activity
Fran Me: for licensee
Uses fact : Control Public
Fran Me: METARULETo determine public control
Uses fact : PLbllc
Fran Me: To determine landholder type
Nils Giulio" = The premises owned by local. county, state or federal government
Asioquesiion: Defendant manages the premises In the public trust
Mowiiom Defendant Is on the board of directors
M 909m": Defendant Is the enterprise administrator
Moussflon: Defendant Is an employee
M0 quosiiom Defendant Is a volmteer
Uses fact : Control Private
Fran Me: To determine private control
Uses fact : Private
Fran Me: To determine landholder type
Ms question: The premises owned by a private Interest
“equation: Defendant Is owner of the premises
mmi Defendant leases the premises
Nils mallow Defendant Is an Independent contractor on the premises

159

Nils oussibm Defendant Is In actual control of the premises
Astoquestton: Land conditions are managed by the defendant or an employee
M3 mallow Permission Yes
Asia question = Plaintiff no business transaction
Nils “MI No monetary transaction occurred for recreation
Fran Me: for Trespasser
Ash ouosfiom Perrnlsslon No
Uses fact : trawasser
Fran Me : for Invitee
M W: Permission Yes
Uses fact : Control Plbilc
Fran Me: METARULETo determine public control
Uses fact : PLbItC
Fran Me: To determine landholder type
m Walton = The premises owned by local. comfy. state or federal government
MQWM= Defendant manages the premises In the public trust
Mowiiom Defendant Is on the board of directors
MQWW= Defendant Is the enterprise administrator
Moussilom Defendant Is an employee
“omission: Defendant Is a volmteer
lbes fact : Control Private
Fran Me: To deterrnlne private control
Uses fact : Private

From Me : To determine landholder type
Asia question: The premises owned by a private Interest
Mooussilom Defendant Is owner of the premises
Mmm= Defendant leases the premises
Moussflom Defendant Is an Independent contractor on the premises
MGW= Defendant Is In actual control of the prernlses
M9W= Land conditions are managed by the defendant or an employee
Uses fact : Proprietorship
Fran Me: to assess proprietorship
Uses fact : PLbIIc
Fran Me: To determine landholder type
”swallow The premises owned by local. county. state or federal government
Uses fact : Private
Fran Me: To determine landholder type
m quesiion = The prernlses owned by a private Interest
M9W= Defendant Is a business
Mutation: Defendant charges fee to enter premises
MW: Defendant charges fee for activity on premises
Ms W = Plaintiff expressly Invited
Asia mostlon: Plaintiff paid a fee to enter prem
Mo Citation: Plaintiff paid a fee fa activity
Fran Me: for licensee
Uses fact : Control Public
Fran Me: METARULETo determine public control
Uses fact : PLblIc
Fran Me: To determine landholder type
Nils Citation: The premises owned by local. county. state or federal government
Mwm= Defendant manages the premises In the public trust
Asks oussiionI Defendant Is on the board of dlrectas
Nils question: Defendant Is the enterprise administrator
Aslaquestlon: Defendant Is an employee
fluctuation: Defendant Is a volmteer
Uses fact : Control Private
Han Me: To determine private control
Uses fact : Private

APPENDIX F

SAMPLE SESSION REPORTS

160

Session Report

Session Date: 9-11-90

Plaintiff: Bill Smith

Plaintiff age: 43

Defendant: Tom Mather

Agency or firm: Saginaw County Park System

Defendant is a: Volunteer

Responsibility lies with: Volunteer

Activity type: Swimming

Month since Injury: 11

Type of negligence pursued: willful and wanton misconduct

Causation: ? (A) cause in fact
(B) Proximate cause. unforeseeable
consequences
(C) Proximate cause. intervening cause

TOTO has determined that there is willful and wanton misconduct by
Tom Mather. In this case liability of the defendant is reduced
because of Bill Smith's conduct contributed to the injury.

Tom Mothers liability: 50%

161

162

SESSION REPORT

Session Date: 8-6-90
Plaintiff: Kerrin O'Brien
Plaintiff age: 23
Defendant: Thord Sundstrom
Agency or firm: Thord's Mountain
Defendant Is a: Owner of the premises
Responsibility lies with: Thord Sundstrom
Activity type: Skiing
Month since injury: 7
Type of negligence pursued: Ordinary negligence
Causation: ? (A) cause in fact
(B) Proximate cause. unforeseeable
consequences
(C) Proximate cause, intervening cause

In order for legal negligence to exist. the plaintiff Kerrin O'Brien must
fiemonsfrate a preponderance of evidence that fulfills basic criteria.
ese
criteria include:
- a duty by the defendant to act or refrain from acting
- a breach of that duty by the defendant's failure to meet
a standard or level of conduct.
- a causal connection between the negligent conduct
and the plaintiff’s injury.
- Actual injury that is measurable

In this case Kerrin O'Brien demonstrated that the first two elements

of negligence exist. In order for there to be negligence there must be
a causal link between negligent conduct and the injury. In this case
the causal connection has not been demonstrated and thus there is
no negligence.

B I BLI OGRAPHY

163

BIBLIOGRAPHY

Allen, L. E. (1980). Language, law and logic: Plain
drafting for the electronic age. In B. Niblett (Ed.),
Computer Science and Law, 75-100. Cambridge:
Cambridge University Press.

Anderson v. Brown Bros., Inc., 65 Mich App 409 (1975).
Ashley, K. D., & Rissland, E. L. (1988). A case-based

approach to modeling legal expertise. IEEE Expert
(Fall), 70-77.

 

Black, H. C. (1983). Black's law dictionary. St. Paul:
West Publishing.

 

Bronstein, D. (1985). Legal perspectives of recreation
liability and insurance. In S. M. DeRosa (Ed.),
Recreation and tourism liability and insurance forum
proceedings I, 17-18. Michigan Travel, Tourism and
RecreaEIcn Resource Center, Michigan State University.

Buchanan, B. G., & Shortliffe, E. H. (1984). Rule-based
expert systems: The experimentsrof the Stanford
heuristic programming project. Reading, MA: Addison
Wesley.

 

Caudill, M. (1987). Neural networks primer, part 1. A;
Expert (December), 46-52.

Chandrdasekaran, B. (1985). Expert systems: Matching
techniques to tasks. In W. Reitman (Ed.), Artificial
intelligence applications for business. Ablex.

 

Clancey, W. J., & Shortliffe, E. H. (1984). Readings in
medical artificial intelligence: The first decade.
Reading, MA: Addison Wesley.

 

Coughlan, J. C., & Running, W. R. (1990). An expert
system to aggregate biophysical attributes of a
forested landscape within a geographic information
system. AI Applications in Natural Resource
Management, 2(4), 34-43.

 

164

165

Cullen, I. (1985). Expert systems in planning analysis.
Town Planning Review, §1(3), 239—251. .

 

Davis, J. R., & Clark, J. L. (1989). A selective
bibliography of expert systems in natural resource
management. AI Applications in Natural Resource
Management, 3(3), 1-18.

 

Duda, R. J., Gasching, J., & Hart, P (1979). Model
design in the prospector consultant system for mineral
exploration. Edinburgh: Edinburgh University Press.

Forsberg, P. F. (1988). Recreational use statutes and
civil law interpretations. Department of Park and
Recreation Resources, Michigan State University,
unpublished report.

Forsberg, P. K. (In presS). An expert system for
marketing natural resources. Proceedings, 3rd
International Conference on Computers in Agricultural
Extension Programs, March 1990, Lake Buena Vista, FL.

 

Forsberg, P. K. (1990). Expert systems in outdoor
recreation. Proceedings of the Outdoor Recreation
Trends Symposium III: Dialogue for the 90's.
Indianapolis, IN.

Gardner, A. (1983). The design of a legal analysis
program. In Proceedings, the National Conference on
Artificial Intelligence, Vol. 1 (pp. 114-118).
Stanford.

Gardner, A. L. (1987). An artificial intelligence

approach to legal reasoning. Cambridge, MA: MIT
Press.

Hafner, C. D. (1981). Representation of knowledge in a
legal information retrieval system. In R. Oddy, S.
Robertson, C. Van Rijsbergen, & P. Williams (Eds.),
Information retrieval research. London: Butterworths
& Co.

Harris, C. C., & Swanson, C. D. (1990). Developments in
modeling choice using the policy referendum format.
In Bla Bla (Eds.), Proceedings of the Outdoor
Recreation Trends Symposium III: Dialogue for the
90's. Indianapolis, IN.

 

 

166

Hays-Roth, F. Klahr, P., & Mostow, P. (1986). Knowledge
acquisition, knowledge programming, and knowledge
refinement. In P. Klahr and D. Waterman (Eds.),
Expert system: Techniques, tools, and applications.
Reading, MA: Addison Wesley.

Hays-Roth, F., Waterman, D. A., & Lenat, D. (1983).
Building expert systems. Reading, MA: Addison
Wesley.

Holecek, D. F., & Westfall, R. D. (1977). Public
recreation onprivate lands--the land owners

 

 

perspective. Report No. 335). East Lansing:
Michigan State University Agricultural Experiment
Station.

Hughes, T. E. Personal communication, May 1987.

Josin, G. (1988). Integrating neural networks with
robots. AI Expert (August), 50-58.

Kaiser, R. A. (1986). Liability and law in recreation,
' parks, and sports. Englewood Cliffs: Prentice Hall.

Kionka, E. J. (1988). Torts. St. Paul: West
Publishing Company.

Klahr, P., & Ellis, J., et a1. (1985). Twirl: Tactical
warfare in the ross language. In P. Klahr, D. &
Waterman (Eds.), Expert systems: Techniques, tools,
and applications. Reading, MA: Addison Wes1ey.

Koldner, J. L. (1987). Extending problem solver
capabilities through case-based inference. In P.
Langley (Ed.), Proceedings of the fourth international

workshop on machine learning. Los Altos, CA: Morgan
Kaufman.

Kozlowski, J. C. (1987). A "cut and paste" of model
recreational use law to include public. Parks and
Recreation, pp. 22-28.

 

Kozlowski, J. C. (1987). California and Florida
recreational use statutes don't apply to public
entities. Parks and Recreation, pp. 16-23.

Kozlowski, J. C. (1986). WI, NJ, and LA, limit public
recreational use immunity to "true outdoors." Parks
and Recreation, pp. 18-22.

167

McAvoy, L., Dustin, D., Rankin, J., & Frakt, A. (1989).
Wilderness and legal liability: Guidelines for
resource managers and program leaders. Journal of
Park and Recreation Administration, 41-49.

 

McCarty, L. T., & Sridharan, N. A. (1982) A
computational theory of legal argument. Report LRP-
TR-13. New Brunswick, N.J.: Laboratory for Computer
Science Research.

McNeal v. Department of Natural Resources, 140 Mich App
625.

McPhee v. Dade County, Fla App, 362 So.2d 74.

Meldman, J. A. (1975). Apreliminarystudy in gomputer
aided legal analysis. Ph.D. Dissertation, MIT,
Cambridge, MA.

Michaelson, R. (1984). An expert system for federal tax
planning. Report University of Nebraska.

Mich. Stat. Ann. ss 13.1485.
Mich. Stat. Ann ss 13.1482 (1) to (11).

Miller v. United States Department of Interior, 649 F
Sup. 444 (W. D. Mich 1986).

Moline, W. J. (1985). Welcome. In S. W. DeRosa (Ed.),
Recreation and tourism liability and insurance forum
proceedings, 1, 1-2. East Lansing, MI: Michigan
State University, Michigan Travel, Tourism, and
Recreation Resource Center.

Nelson v. City of Gridley, 113 Cal. App 3d 87, 169 Cal.
Rptr. 787 (1980).

Ortolano, L., & Perman, D. (1987). A planner's
introduction to expert systems. American Planning
Association Journal (Winter), 98-103.

Popp, W. G., & Schlink, B. (1975). JUDITH, a computer
program to advise lawyers in reasoning a case.
Jurimetrics Journal, 1§(4), 303-314.

Prosser, W. (1984). Cases and materials on torts. New
York: Foundation Press.

168

Rauscher, H. M., & Hacker, R. (1989). Overview of AI
applications in natural resource management.
Artificial intelligence and growth models for forest
management decisions conference. Vienna, Austria.

Richer, M. (1989). AI tools and techniques. Norwood,
N.J.: Ablex Publishing Company.

 

Rissland, E. L., & Shalak, D. B. (1989). Interpreting
statutory predicates. In E. L. Rissland (Chair), The
Second International Conferenge on Artificial
Intelligence and the Law, 46—53. New York: The
Association for Computing Machinery.

 

Rissland, E. L., Vakarce, E. M., & Ashley, K. (1984).
Explaining and arguing with examples. In AAA1-84,
Proceedings, National Conference on Artificial
Intelligence, 288—294, Austin, Texas.

Shulman, L. B. (1985). Legal perspectives of recreation
liability and insurance. In S. M. DeRosa (Ed.),
Recreation and tougism ngbility and insurance from
proceedings, 1., 17-18. East Lansing MI: Midhigan
State University, Michigan Travel, Tourism, and
Recreation Center.

 

Siderelis, C. D. (1985). Park maintenance management
systems: Past and present. Journal of Park and
Recreation Administration, 3(3), 29-36.

 

Sprowl, J. A. (1979). Automating the legal reasoning
process: A computer that uses regulations and
statutes to draft legal documents. American Bar
Foundation Research Journal, 1-18.

 

Susskind, R. E. (1989). The latent damage system: A
jurisprudential analysis. In E. L. Rissland (Chair),
The Second International Conference on Artificial
Intelligence and the Law. New York: The Association
fdr Computing Machinery.

Thomas v. Consumers Power Co., 58 Mich App 486 (1975).

Trelease, R. B. (1988). Connectionism, cybernetics, and
the cerebellum. AI expert. August, 30—36.

 

United States v. Carrol Towing Co. (2nd Cir 1947), L.
Hand, J.

169

van der Smissen, B. (1990). Legal liability and risk
managementfor public and private_entities. Vols. 1,
2, and 3. Cincinnati: Anderson Publishing Company.

 

 

Waterman, D. A. (1986). A guide to expert systems.
Reading, MA: Addison Wesley.

 

Waterman, D. A., & Peterson, M. A. (1986). Models of
legal decision making. In P. Klahr & D. Waterman
(Eds.), Expert systems: Techniques, tools, and
applications. Reading, MA: Addison Wesley.

 

Weilinga, B. J., Bredeweg, B., & Breuker, J. A. (1989).
Knowledge acquisition for expert systems. In J.
Seikmann (Ed.), Advanced_topics in artificial
intelligence. Berlin: Springer-Verlag.

 

White, H. (1989). Neural network learning and
statistics. Al expert, December, 48-53.

 

Williams v. City of Cadillac, 148 Mich App 86 (1985).
Winiecki v. Wolf, 147 Mich App 742 (1985).

Wymer v. Holmes, 144 Mich. App. 192, 375N.W.2d 384
(1985).

Yahrling, v. Belle Lake Assoc., 145 App 620, 378N.W.2d
772 (1985), rev'd 429 Mich. 66, 412N.W.2d213 (1987).

"liliiiil'lliiilll