We talked about the CatchBob data analysis. We're going to use both qualitative and quantitative techniques. Note: we have to be careful to balance the number of Bob's location: there should be the same number of Bob at 'la coupole' as in 'faculte d'informatique'
1. about the tabletPC screenshots: their map annotations (from the 3 players): we came up with the following coding scheme to analyse the map annotation: - content: position/direction/signal strength(proximity sensor)/strategy(stay calm)/off-task notes/acknowledgement/content-free acknowledgement - mode: textual (+numbers)/graphical - position on the map: site-specific/non site specific - intent/pragmatics: announcement/request
Let's have a quantification of each messages: - number of messages per player/group - number of messages per category per player/group check if there are still position request in the condition "with location awareness tool". It if it's the case, it might be an indicator that this query and the answer are not only related to the position but also to something higher in terms of intent modeling...
2. Path drawings: I have the path of each participant (thanks to client+server logfiles) + their drawing (A drew is path and the path of B and C)
Calculate the number of errors about A's path: path drawn by A/real path by A - places where they have not been - place they were but forgot to draw PLUS: real path by A/A's path drawn by B and real path by A/A's path drawn by C The tricky thing here is to define what is an error, we need some criteria like: - distance (higher than the maximum size of epfl corridors) - no visibility between the two representations - door/wall/glass - walking back just a while is not an error
Then: we can compute the number of errors A did about B: E(A,B) as well the number of errors A did about C: E(A,C) Thanks to this errors evaluation we have an relevant evaluation: the quality of the spatial representation that A had about B and C: an indication of Spatial Mutual Modeling MM(A-BC)= [E(A,B)/E(A,A)] + [E(A,C)/E(A,A)] (we use the error about A own's position as a weight here that indicate the accuracy of his representation of space).
3. Self-confrontation to the replay tool look for explicit stuff like: “I understood that you/he/she/they wanted to…", “I did not get that you/he/she/they", “you/he/she/they did not understand that…": - critical incidents ('I thought you were doing this') - 'best practices' (good modeling)
then we can have a catalogue/typology of mutual modeling acts -> transcript just those moment. That will be good to illustrate + to find new variables we did not think about...