Interesting discussion today with Patrick about the fundamental issues of CSCL. We started from the discussion about the differences of a guiding /tutoring system to a mirroring system. The limits of the former are the complexity of a computational model able to acknowledge group interaction. The latter has the limit of the design choice of what elements of the interaction are relevant to show and provide as feedback to the users.

Fundamentally, the problem is that building a model intelligent fight against the difficulty of translating a cognitive model into a third order logic. So, either you loose the discreteness and concreteness of the elaboration you get form the system, either you find yourself in a finite state machine (FSM), which do not acknowledge the complexity of human interaction, therefore is of no use.

I was questioning myself wether AI is dead. Patrick’s current point is to find good algorithms able to show relevant information to the users which are able to propel the learning process. More so, he is interested in finding algorithms able to detect the change of state of a group. Problem here is to find exactly what to measure to be able to detect those kind of changes, and secondly how to define such changes.