fidélité cognitive

OBS: Voir modélisation cognitive.

CONT: The evolution of the SOPHIE project can be viewed as a search for the right ingredients of ""cognitive fidelity"" in modeling expert reasoning. SOPHIE-I abandons the automata of the METEOROLOGY tutor to take advantage of mathematically based computational formalisms, which place the emphasis on physically faithful simulations. In an effort to gain cognitive fidelity, SOPHIE-III moves away from the structurally isomorphic representations of quantitative simulations to concentrate on mental models, but in some sense it goes too far and becomes too specialized to a given form of reasoning in the context of a specific task.
CONT: Both internal and external representational mappings can have varying degrees of epistemic fidelity, that is, they can reflect the epistemic source a little or a lot.
CONT: An expert system is cognitively faithful if its own internal solutions correspond to the solutions a human would produce. MATHPERT solves math problems in the way we teach students to do, rather than using high-powered algorithms. Cognitive fidelity must be designed for from the beginning, as the demand for cognitive fidelity complicates the construction of an expert system considerably.