Traces Based Reasoning ( TBR ) Definition , illustration and echoes with story telling

The question is wether it is possible to exploit a computer environment to facilitate user’s tasks, whatever they are. The goal is not to develop a new kind of ”intelligent” systems using a ”world representation” to reason about it as the user himself according to some adapted cognitive model. The question is rather to design computer environment able to work ”in understanding” with the user. This system’s ”empathy” for the user cannot be reduced at building a user profile, typifying his/her according to pre-established factors. Actually, the user works in context of an environment which cannot be reduced to the computer environment and it is vain to ask him/her to describe precisely the significant context of each use. Hence, the main idea is to try to progressively learn useful contexts from user interactions which are mediated by the computer environment. This paper does not address the use of traces as data to mine and analyse in order to build general rules on the ”world domain”, which is an other research field. The question is to exploit traces as they are ”recognizable” by users to make easier user’s tasks when mediated by a computer environment. The first part of the paper focuses on this approach, trying to define precisely what we propose to call ”Traces Based Reasoning”; second part describes shortly the MUSETTE (Modeling USEs and Tasks to Trace Experience) approach common base of several TBR oriented systems. MUSETTE theorizes the notion of ”use trace” and its exploitation by an Explained Task Signature (ETS) mechanism; third part presents shortly a TBR oriented application which is an ITS system reusing learner traces to help the learning process . Discussion is an opportunity to place our work in the context of other approaches