FACTORS THAT AFFECT GROUP GENERATIVE COLLABORATIONS IN ENTERPRISE SOCIAL MEDIA By Elisavet Averkiadi Innovation is closely linked to a compan y s ability to survive and thrive. As a result, companies are becoming increasingly interested in group generative collaborations the conception of novel ideas and solutions through group exchanges. This explorator y study investigates the prevalence of generative collaborations in ESM - based groups and identif ies antecedents of such collaborations in groups. Additionally, the nature (i.e. language) of these collaborations is explored. A mixed methods approach is empl oyed for this study , comprised of a content analysis of text - based data from an ESM platform , building a machine learning classifier model , and a predictive regression model . The results of this exploratory study show that approximately 44% of exchanges on an ESM platform contain elements of generative collaborations . Furthermore , the predictive negative binomial regression model illustrate s open), the bonding and bridging behaviors of group members , and the size of a group are significant antecedents for group generative collaborations . These outcomes are valuable insights that help advanc e our theoretical understanding of ESM - based group generative collaborations and could also help improve these c ollaboration behaviors in the context of ESM platforms, and possibly beyond . ................................ ................................ ................................ ......................... ................................ ................................ ................................ ...................... ................................ ................................ ................................ .......................... ................................ ................................ ................................ ................ ................................ ................................ ................................ ..................... ................................ ................ ................................ ................................ ................................ ................................ .............. ................................ ................................ ................................ ................................ ... ................................ ................................ ................................ ........... ................................ ................................ .............. ................................ ................................ ................................ ......... ................................ ................................ ................................ ..... ................................ ................................ ................................ ........... ................................ ................................ ................................ ................................ ...... ................................ ................................ ................................ ........... ................................ ................................ ................................ ......... ................................ ................................ ...................... ................................ ................................ ...................... ................................ ................................ .................. ................................ ................................ ................................ ........ ................................ ................................ ................................ .... ................................ ................................ ................................ ..... ................................ ................................ ................................ . ................................ ................................ ................................ ............................... ................................ ................................ ................................ .... ................................ ................................ ................................ ................... ................................ ................................ ..................... ................................ ................................ ................................ ............................. ................................ ................................ ................................ .............................. ................................ ................................ ................... ................................ ................................ .......... ................................ .... ................................ ................................ ................................ ............................. ................................ ................................ ............... ................................ ................................ .............. ................................ ................................ .................... ................................ ................................ . ................................ .............. ................................ ...................... ................................ ...................... ................................ ................................ ................. ................................ ................................ .................. ................................ ................................ ............ ........................ ............................ ................. ............................ ............. .................... ................................ ................................ ....... ................................ ....... ................................ ................. ................................ ................ ................................ ................................ ............................. ................................ ................................ ................ ......................... ................................ .................... ................................ .................... ................................ .............. ................................ ........ ................................ ........ ................................ .... ................................ ................................ .......................... ................................ ................................ .......................... ................................ ................................ ................................ ........... ................................ ................................ .. In the last few years, Enterprise Social Media (ESM) have become an omnipresent workplace IT tool. ESM are web - based platforms that aid organizations to conduct internal corporate business processes and objectives, by allowing users to communicate, network, organize, leverage information capital, and collaborat e . ESM are encouraging knowledge sharing within the workplace by making it simpler to find and work with people with mutual interests and complementary expertise (Leonardi, 2014). These platforms are also equipped with affordances that have the potential to foster group generative collaborations the co - creation of novel ideas and solutions through group exchanges (Avital & Te'eni, 2009). ESM offer a unique opportunity for researchers interested in studying online team behaviors. Visibility, a unique affordance of ESM, allows all contributions and activities to the platform to become visible to anyone, enhancing not only the possibilities for employees to access knowledge and ideas from anywhere in the organization, but also the availa bility of such data for investigation . This opportunity to observe how people colla borate is not only an opportunity to improve the theoretical understanding of the nature of group exchanges occurring in ESM, but also a chance to improve these types of exchange s in the context of ESM platforms, and potentially beyond. As the root - cause of innovation (Avital & Te'eni, 2009), generative collaborations are of particular interest for companies, given the direct link between their ability to innovate and their chance t o survive and thrive (Cowen, 2011; Abernathy & Clark, 1985). The objective of this study is to examine the forms of generative collaborations that occur in Enterprise Social Media and identify the factors that predict their success. To do this, I pose two research questions: RQ1: What types of generative collaborations occur in ESM? RQ2: What are the antecedents to the generative collaborations that occur in ESM? To conduct this study, I will be utilizing data that contains thread message s from a multinati onal organization that uses an ESM platform for internal corporate processes. This data set includes a total of over 2 0 ,000 messages from various groups on the platform . Given its exploratory nature, a subset of 2,230 messages were used for the purpose of this study. conceptual change, where the author identifies three distinct types of creativity (i.e., generativity): reframing, expansion, a nd combination. Additionally, to identify antecedents of generative collaborations, I will implement the forms of generative collaborations as a dependent variables, and I will be incorporating characteristics of groups found in ESM ( as independent variabl es ) ; including their size (number of members), their privacy setting (i.e., whether they are open (visible to network) or closed (invisible to network) groups ) , and various network measures ( e.g., particularly the bridging and bonding structures of the group network ). 1 1 The case organization has recently migrated to a new platform called Igloo. In this study, the focus remains on the data from the Jive platform; the platform migration did not affect this study. Figure 1 : Average Group Size by Group Visibility Figure 2 : Distribution of Group Generativity Type by Group Visibility Figure 3 : Distribution of Group Generativity by Group Bridging Figure 4 : Distribution of Group Generativity by Group Bonding Table 1 : Content Analysis Training Timeline Figure 5 : Qualtrics Example Survey 2 2 at a threshold of 0.5 and above. In this study however, due to the multiclass and multilabel nature of the data, the complexity of inter - reliability between coders is higher. Thus, the alpha scores produced at nearly ~0.5 were considered satisfactory enoug h for the purposes of this exploratory study. Table 2 : Inter - Coder Reliability Highest Pairs Figure 6 : Machine Learning Classifier Steps Table 3 : Regression Models Variable List Table 4 : Operationalization of Group Level Variables Model 1 : Negative Binomial Regression Model with Bonding and Group Visibility Interaction Terms Model 2 : Negative Binomial Regression Model with Bridging and Group Visibility Interaction Terms Figure 7 : Distribution of Categories of Generativity from the Content Analysis more than the need to share data with external parties. I use it primarily as an easy backup for files on my pc that I would hate to lose, and as way to share files between multiple devices that use my work pc and my home pc. For example services like google drive Dropbox and SkyDrive let me map a folder on my pc hard drive to a cloud storage location that is kept constantly synchronized so I can access these files from any of my devices, and I can also be sure that ev en if my pc hard drive completely fails - which has happened to me in the past In the following example of expansion, the employee who wrote the message is discussing with colleagues the process of new product creation. Here, the employee is expanding the use of a quote they mention, to fit the topic of discussion in the thread : here we can find Taking this quote a step further, and looking at the [case organization] family as whole company have made me look good in front of customers. We have so many knowledgeable individuals and the fact that anyone and everyone is willing to pull The example of c ombination below is from an employee contributing to a thread about moves on to combine what was mentioned by the colleague into a new idea for implementing gamification theory : the same thing. Drive really describes why certain elements of game theory work as well as they do . Maybe gamification game theory helps one think through how to apply the concepts. If one were to take an eleme nt of performance such as Mapp objective or maybe dimension of the ITCDPP, and structure it as more of a game, how might that look? keeping in mind a game does not always mean video game, but can come in different varieties, such as group oriented vs. solo and tangible real vs. virtual group. Tangible baseball solo, tangible bowling, group virtual world of war craft solo, virtual pac man. Games also tend to have meaningful status, and an intrinsic reward component. How might this be built into the mechanics Figure 8 : Reframing Word Cloud with Content Analysis Data Figure 10 : Expansion Word Cloud with Content Analysis Data Figure 9 : Combination Word Cloud with Content Analysis Data Table 5 : Reframing Logistic Regression Classifier Model Resu lts Table 6 : Expansion Random Forest Classifier Model Results Figure 11 : Reframing Machine Learning Model Linguistic Indicators Figure 12 : Expansion Machine Learning Mo del Linguistic Indicators Table 7 : Combination Naive Bayes Classifier Model Results Figure 13 : Combination Machine Learning Model Linguistic Indicators Table 8 : Reframing Model 1 & 2 Estimates Figure 14 : Reframing Interaction Plot Figure 15 : Expansion Interaction Plot Table 9 : Expansion Model 1 & 2 Estimates Table 10 : Combination Model 1 & 2 Estimates 3 3 Percentage breakdown from the 44% of data that contain elements of generative activity. Figure 16 : Example Thread APPENDIX B Figure 17 : Example Thread for Final Coding Manual APPENDIX C Table 12 : Reframing Naïve Bayes Machine Learning Classifier Performance Table 11 : Reframing Random Forest Machine Learning Classifier Performance Table 13 : Expansion Logistic Regression Machine Learning Classifier Performance Table 14 : Expansion Naive Bayes Machine Learning Classifier Performance Table 15 : Combination Logistic Regression Machine Learning Classifier Performance Table 16 : Combination Random Forest Machine Learning Classifier Performance