Abstract This paper investigates the potential of existing methods for analyzing collaboration in wiki environments. After a short description of the motivation for this research paper a presentation of analysis methods in CSCL will be given, including a special focus on Social Network Analysis. The next section points out the main characteristics of wikis and their differences compared to other CSCL tools. In the following step, the methods for collaboration analysis are transferred to wiki contexts. The paper concludes with open issues and an outlook on future research on this topic area. Keywords Wikis, CSCL, Collaboration Analysis, Social Network Analysis ACM Classification Keywords H.4.m. Information Systems Applications: Miscellaneous. H.1.2. User/Machine Systems. Motivation Since the emergence of the Web 2.0, wikis have been applied in various areas. Because of their openness to Copyright is held by the author/owner(s). CHI 2009, April 49, 2009, Boston, Massachusetts, USA. ACM 978-1-60558-247-4/09/04. Saskia-Janina Kepp Information Science Hildesheim University Marienburger Platz 22 31141 Hildesheim, Germany keppsa@uni-hildesheim.de
Heidemarie Schorr English Language and Literature Hildesheim University Marienburger Platz 22 31141 Hildesheim, Germany schorr@uni-hildesheim.de
CHI 2009 ~ Spotlight on Works in Progress ~ Session 2 April 4-9, 2009 ~ Boston, MA, USA 4201
every user and every topic without restriction of access they have been applied in higher education to encourage self-determined, self-organized and self- regulated learning, e.g. [3,4,7]. But the existence of wikis alone, like any other software, is not enough to guarantee successful learning scenarios. Support and moderation are important in every learning platform, especially those allowing constructivist learning, in order to guide and scaffold the learning process, e.g. [6]. With every user becoming a potential author and the amount of content rising exponentially, the lost in hyperspace syndrome and the cognitive or information overload problems that are well-known in the context of hypermedia have an even stronger impact [14]. Thus far, there have been attempts to motivate and encourage participation as well as collaboration in collaborative settings by giving incentives. One example for an incentive is rewarding contributions and giving ranks to the participants according to the amount of their contributions [5]. However, our assumption is that motivation is not the main problem. From our evaluations of wiki-based learning scenarios [8], we concluded that the students are very motivated, but feel overwhelmed by the number of articles and authors in the Wikis in use. Moreover, they feel they destroy their peers work if they edit and change the contents others have created. Therefore, in our wiki-based learning scenarios, editing and adding to other peoples work has been made an explicit task and the common goal of collaboratively creating a knowledge base is emphasized over the individual goals of externalizing ones knowledge. In addition to that, different technical support channels are made available to the students. While these basic conditions are creating a framework for using wikis for collaborative learning, they need extensive moderation accompanying and scaffolding the learning process. Since this is too much work for one lecturer alone, supportive software could be employed. According to [13], there are different possibilities for realizing automatic moderation to some degree: the software can either visualize the current state of collaboration (mirroring) or visualize the current state as well as an ideal state of collaboration so that the users can adapt their behavior in order to get closer to the ideal state (metacognitive tools). A third option is to let the system interpret the current state of collaboration and infer feedback for the users from it (guiding systems). Irrespective of the approach, the state of collaboration has to be measured first. This work-in-progress paper describes existing methods for the analysis of collaboration in CSCL systems focussing on Social Network Analysis. These methods will then be transferred to wiki environments. Open issues and an outlook on future research will conclude the paper. Collaboration Analysis in CSCL Before taking a look at the existing methods for analyzing collaboration, a definition for the term collaboration is given. According to [12], collaboration and cooperation can be distinguished by the way tasks are completed. Cooperative tasks, on the one hand, are divided into subtasks that are being accomplished separately by single group members. The result of cooperation is therefore the sum of the results of these subtasks. Collaboration, on the other hand, means that the group members directly work together towards a common goal. The result of collaboration is more than the sum of the parts, because every person adds their CHI 2009 ~ Spotlight on Works in Progress ~ Session 2 April 4-9, 2009 ~ Boston, MA, USA 4202
own knowledge and extracts new knowledge by working on the task. Cooperation, therefore, mainly involves communication, while collaboration involves communication as well as co-construction [12]. When analyzing collaboration it is important to differentiate between these two possibilities. If the participants communication is to be analyzed, conversation analysis can be applied. The co- construction of knowledge can be analyzed using activity-based methods, for example. With Social Network Analysis it is possible to consider both communication and co-construction. Using this method, each user is represented by a node in the network and the relationship between them is represented by edges between these nodes. A relationship is consequently formed by working on a wiki entry together, i.e. someone edits an entry somebody else has worked on. Another way of establishing a relation between two network nodes is to comment on an entry created and/or edited by one or more peers. This network can be analyzed on the individual, the group or the community level. By pointing out the main differences between Wikis and other CSCL tools, the next section will show the effectiveness of this method for analyzing wiki collaboration. Differences between Wikis and other tools Wikis are websites consisting of several interconnected pages that can be created, edited and commented on by every visitor to the site [9]. Each wiki page can be viewed in read mode, like regular websites, or in edit mode, where the content of the page can be changed or added to by using so-called wiki syntax [9]. Wiki syntax is simplified HTML code that allows users to format and layout the page without having to be familiar with HTML. Moreover each page usually has a discussion or commenting section where the communication about the contents takes place. Nevertheless, co-construction is the main focus of wikis. For learning scenarios they therefore offer the opportunity for collaborative learning and let students take an active role in their learning process. As a consequence, conversation-oriented methods are not sufficient for analyzing wiki contents. Another difference between wikis and traditional CSCL software is the absence of fixed groups and roles. Instead, wikis are characterized by being open to every user, by not restricting access rights and by allowing free collaboration within the community. That means that one user can either work alone on an article or in a group, together with others. The composition of the group, however, can change constantly, because a user can switch groups at any given time and can also be in several groups at the same time. As a result, techniques that analyze collaboration on the group level are not sufficient. Instead, techniques analyzing wiki collaboration have to focus on the individual and the community level rather than on groups. A wikis openness also extends to the topics covered. For instance, with many conversation-oriented methods there comes a restriction to specific sentence openers in order to map predesigned conversation paths to the analysis results to discover conflicts. Such restrictions of usage should be avoided in order to maintain the wikis openness. CHI 2009 ~ Spotlight on Works in Progress ~ Session 2 April 4-9, 2009 ~ Boston, MA, USA 4203
One last aspect is the fact that using wikis is an asynchronous process. Therefore the analysis methods have to take into account that the collaboration has to be tracked over an undetermined period of time. Analysis of Wiki Collaboration There have been attempts to analyze wiki collaboration by using different methods. [1] for example, have considered an entrys longevity to infer the quality of contributions to a wiki article. [15] have developed a way of measuring the degree of co-authorship of wiki articles. However, when wikis are deployed in learning settings additional measures have to be considered. For example it is important to measure the amount and types of activity of all learners to identify those who are active participants and those who are passive. While passive browsing, i.e. reading the contents, can still be considered learning, it mainly enables the readers to memorize and reproduce knowledge [2]. For higher order goals, according to [2], the active processing and construction of knowledge is important. Since our goal is to enable competence development, we need to identify passive learners in order to be able to guide them towards being more active. In order to determine the degree of collaboration, the number of contributions can be counted. Contributions can be, for example, the creation of new wiki entries, editing or commenting on existing ones. Social Network Analysis can be used to describe this collaboration. It has been applied in CSCL contexts before, but has so far mainly focused on the aspect of communication. [11], for example, have used cluster and clique analyses to monitor e-mail communication and forum discourse. Another example is SAMSA, a tool that analyzes the actors degree centrality, the network density and the network degree centralization in order to study participatory aspects of learning [10]. This tool has been applied to various settings in which synchronous as well as asynchronous cooperation and collaboration was supported. Thus, it was possible to analyze communication as well as co-construction. The results show that Social Network Analysis is especially useful in asynchronous settings and therefore could possibly be transferred to analyzing collaboration in wiki contexts. Regarding the analysis object, SAMSA mainly applies Social Network Analysis to the group level. As has been said before, the focus concerning the analysis of wiki collaboration should be on the individual and the community levels, which can also be analyzed with this method. To give a working example: we measured the actors degree centrality in one of our wikis which is a glossary that is filled collaboratively. During the last semester students worked on creating wiki entries and thereby produced a set of log data consisting of 1773 records. After cleaning the dataset, 167 records could be identified as collaborative activities. The activity types considered as collaborative are as follows: edit, comment, upload and tag. According to this data, 11 out of the 21 participating students were actually collaborating whereas the other 10 participants were mainly working by themselves. The resulting degree centralities for the nodes therefore lie between 0 and 0.3. CHI 2009 ~ Spotlight on Works in Progress ~ Session 2 April 4-9, 2009 ~ Boston, MA, USA 4204
Social Network Analysis should be supplemented by qualitative measures that look at the quality of the contents created. The approach developed by [1], i.e. considering the longevity, is not necessarily a useful indicator for the quality of an article in respect of learning settings. Instead, an entrys longevity could result from a lack of motivation or courage of peers to edit it, a lack of knowledge on how editing is done or a lack of factual knowledge on the topic of the entry. Therefore other means of quality measurement need to be found that are not biased by these factors. Important measurements that can be used to infer the quality of a wiki page are formal aspects like appropriate formatting, and especially the mentioning of references. Another indicator that especially applies in the wiki context is the interconnectedness of a wiki article, i.e. the amount of hyperlinks to related concepts within the text of the wiki page. A rather simple measuring device for determining the quality of a wiki entry in a learning setting can be the length of an entry. It often takes more than a few lines of text to describe a concept in a scholarly way. Therefore a very short entry is, in many cases, unlikely to be a complete one. Open issues and future work This paper investigates the applicability of Social Network Analysis to wiki-based learning scenarios based on a literature review. This theory-driven study and a working example show the potential the method could have in wiki contexts. The next step is an extensive study to verify the assumptions. In this study further measuring methods will be applied to the wiki log files and the results will be discussed with pedagogues in order to evaluate their validity. By describing how Social Network Analysis can be applied to model collaboration in wikis we have not yet dealt with the fact that wiki entries, and thereby their authors, can be connected by setting hyperlinks inside the wiki article. That relationship is not as strong as the direct collaboration by working on the same entry and therefore would need to be defined as another type of relationship. As a consequence we are comparing two different kinds of relationships: direct collaboration (working on the same wiki entry) and indirect collaboration (working on related wiki entries that are connected to each other through hyperlinks). Commenting should then be defined as the third type of relationship. Future work will address these issues. The results of this analysis can be used to support collaborative processes either by visualization of the results or by interpreting them in order to give automatic guidance to the learners and users of the wiki. This guidance can, for example, lessen the lost-in- hyperspace effect mentioned in the beginning and help learners organize their self-regulated work. CHI 2009 ~ Spotlight on Works in Progress ~ Session 2 April 4-9, 2009 ~ Boston, MA, USA 4205
Acknowledgements I thank the German county of Lower Saxony for funding the CELEB project that is realized within the ELAN context and in which the use of wiki and other tools fr self-organized learning is studied. Citations [1] Adler, B.T., de Alfaro, L., Pye, I., and Raman, V. Measuring Author Contributions to the Wikipedia. In Proceedings WikiSym 2008. http://www.wikisym.org/ws2008/proceedings/research papers/18500027.pdf [2] Bloom, B. S. Taxonomy of Educational Objektives. Handbook I: Cognitive Domain. New York, 1956. [3] Duffy, P., Bruns, A. The Use of Blogs, Wikis and RSS in Education: A Conversation of Possibilities. In Proceedings of the Online Learning and Teaching Conference, Brisbane, 2006, 31-38. [4] Ebner, M., Kickmeier-Rust, M., Holzinger, A. Utilizing Wiki-Systems in Higher Education Classes: a Chance for Universal Access? In Universal Access in the Information Society, volume 7, issue 4, 2008, 199-207. [5] Farzan, R., DiMicco, J., Millen, D. R., Dugan, C., Geyer, W., and Brownholtz, E. Results from deploying a participation incentive mechanism within the enterprise. In Proceeding of the Twenty-Sixth Annual SIGCHI Conference on Human Factors in Computing Systems (CHI '08). ACM Press (2008), 563-572. [6] Hmelo-Silver, C. E., Duncan, R. G., Chinn, C. A. Scaffolding and achievement in problem-based and inquiry learning: A response to Kirschner, Sweller, and Clark (2006). Educational Psychologist, volume 42, 2007, 99-107. [7] Jadin, T. Social Software fr kollaboratives Lernen. In Batinic, B., Koller, A., Sikora, H. (Eds.). E-Learning in Obersterreich. Digitale Medien und lebenslanges Lernen. Trauner, Linz (2008), 2335. [8] Jaksch, B., Kepp, S.-J., Womser-Hacker, C. Integration of a Wiki for collaborative knowledge development in an E-Learning context for university teaching. In A. Holzinger (Ed.). USAB 2008, LNCS 5298, Springer, Berlin, 2008, 7796. [9] Lamb, B. Wide Open Spaces: Wikis, Ready or Not. In Educause Review, volume 39, issue 5, 2004, 36-48. [10] Martnez, A., Dimitriadis, Y., Gmez-Snchez, E., Rubia-Avi, B., Jorrn-Abelln, I., Marcos, J. A. Studying participation networks in collaboration using mixed methods. In International Journal of Computer- Supported Collaborative Learning, volume 1, issue 3, 2006, 383-408. [11] Reffay, C., and Chanier, T. How Social Network Analysis can help to Measure Cohesion in Collaborative Distance-Learning. In Proceedings of Computer Supported Collaborative Learning Conference 2003 (CSCL'2003), Kluwer Academic Publishers, Dordrecht, NL, 2003, 343-352. [12] Roschelle, J., Teasley, S. D. The construction of shared knowledge in collaborative problem solving. In OMalley, C. (Ed.) Computer Supported Collaborative Learning. Berlin, Springer, 1995, 69-97. [13] Soller, A., Martinez, A., Jermann, P., and Mhlenbrock, M. From Mirroring to Guiding: A Review of State of the Art Technology for Supporting Collaborative Learning. In International Journal of Artificial Intelligence in Education, volume 15, 2005, 261-290. [14] Stickel, C., Ebner, M., Holzinger, A. Useful Oblivion Versus Information Overload in e-Learning Examples in the Context of Wiki Systems. In Journal of Computing and Information Technology, volume 16, issue 4, 2008, 271-277. [15] Tang, L.V.-S., Biuk-Aghai, R.P., and Fong, S. A Method for Measuring Co-authorship Relationships in MediaWiki. In Proceedings WikiSym 2008. http://www.wikisym.org/ws2008/proceedings/research papers/18500125.pdf CHI 2009 ~ Spotlight on Works in Progress ~ Session 2 April 4-9, 2009 ~ Boston, MA, USA 4206