where we are: we have catchbob! a collaborative and mobile game platform running on ipaq and pcs. we want to use it as a testbed to study collaborative processes, namely mutual modeling and division of labor mutual modeling = mutual belief/mutual knowledge/mutual understanding = awareness (evidence) + assumption/induction = representation of the partner cognitive state (knowledge, goals, purposes, intents, understanding).
group modeling = representation of the group itself: task-level features (e.g. group representation of the problem state) and interaction-level features (e.g. who is more active...)
usual method to study mutual modeling 1. ask questions (on-task interviews), explicitly during the collaboration process. Methodological BIAS: participants hence pay more attention to their partners than they would normally do. 2. find behavioral cues during the game: for instance if A goes in the same place B already visited or if C performed already done actions (when it is not relevant to perform the action several times). the point is to find signs that a participant did something because he misunderstood what the others did or wanted him to carry out. 3. ask question after the game: replay of specific episodes, self-confrontation (one participant or focus group). The material that shoulkd be shown could be as rich as possible (path, users' actions, users' interactions...) so that they can elaborate on it and explain their behavior. We can also ask them to DRAW stuff (paths, interactions location...). BIAS: (inherent of stimulated recall) what participants will explains is different from what really happened. 4. during the game: it is possible to assign specifi roles to each participants or either to give them different piece of knowledge.
-> solution 2 and 3 are the most interesting we are working on a catchbob! replay tool and log parser to meet this end. the replay tool provides an interface to navigate through the game, to edit and mix the history of collaborative episodes/interactions.
replay possibilities (ask participants why they found the interaction episode interesting, why they did that...): - selective/all: the experimenter OR the subjects choose specific episodes to comment. - individual/group replay - replay/mixed replay: the groups's episode/another group episode (ask them if they recognize each episodes as part of the group history or not and which elements of the replay determined their answers) - replay mode: mono/stereo. in the mono mode, A sees his/her own action and B+C sees A's action. In the stereo, they all see the actions carried out by the three persons.
episode selection: how to select relevant episodes, criteria for interesting episodes selection: rich/poor interaction: conflict solving, explanation, "i got it episodes", silence/lost episodes