Collaborative Learning

A method of group formation for collaborative learning.

Inputs: Feedback on past users' collaborative interactions
Output: Group formation for upcomming rounds of collaboration

Addressed Problems

Promoting and maintaining effective collaborative interactions in a web-based system is not an easy task. As a first step in a collaborative activity, the group is formed. With each activity having a specific requirements on participants’ abilities to sustain a well-functioning interaction a suitable selection and assignment of users to roles is needed.

To this end we propose an idea of opportunistic collaboration driven by user feedback that progressively explores the social structure identifying users’ skills and characteristics relevant for effective collaboration, and allowing it to allocate suitable collaboration peers for activities executed in a series of collaboration rounds.


After each collaborative activity the group performance and peer feedback are evaluated facilitating the exploration of the social graph of users. Users being nodes of the social graph are connected by edges representing the activities (see Figure below). The process starts with a social graph without edges. By participating in many short collaborative activities the groups are recombined frequently, seemly very random at first, providing numerous graph edges weighted by the peer feedback on the quality of participation in selected dimensions of relevant users’ characteristics. After several rounds of collaboration, the social structure allowing to identify suitable collaboration peers emerges.

method overview

Collaborative interaction visualization (students A,B,C,D,E,F).

The dimensions of users’ characteristics to be controlled for are not established yet. While we could use the ontology of collaborative learning for an exhaustive enumeration of possible characteristics, we deem necessary to keep the complexity of peer feedback low to not to discourage the participants to use it comfortably. With the graph structure in place, missing characteristics of users can be inferred on the basis of the characteristics in the neighborhood while every other user feedback provides us with more information to capitalize on.


By a rather loose process of group formation in the early stages of social graph exploration, the method gets diverse users to try working together possibly discovering novel types of collaborative organization within a team. By placing the users in the social structure a kind of assessment of collaboration is provided by which they can identify their strengths and weaknesses and improve upon them.


  1. Tvarožek, J. (2008): Effective Collaborative Interactions in Personalized Web-Based Systems. In: Proc. in Informatics and Information Technology Student Research Conference IIT.SRC 2008, pp. 171-176.