"Tunnel effects" in cognition : A transfer mechanism from known conceptual domains to new ones

Antoine Cornuéjols (1) , Andrée Tiberghien (2) , Gérard Collet (3)

1 LRI - Laboratoire de Recherche en Informatique
2 ICAR - Interactions, Corpus, Apprentissages, Représentations
3 COAST - COmmunication Apprentissage des Savoirs scientifiques et Techniques

This paper reports on a multidisciplinary head-on approach to the problem of learning new ways to interpret the world by relying on (and relating to) old ones. By studying how high school students address problems in conceptual domains that are new to them, we were led to analyze mechanisms that seemed to be at play in their segmenting the world, and constructing models of the situation, as well as the (re)conceptualization efforts that-sometimes-followed. In this paper, we focus on a reasoning mechanism that we hypothesize does explain part of the students behavior and may be at play within scientific discovery processes. Like analogy, this mechanism that we call tunnel effect by reference to transport effects that can take place in quantum physics, allows the transfer of knowledge from one conceptual domain to another one. Unlike analogy however, it does so without having to resort to two situations or cases, but only considers the one at hand, and it does not necessitate to specify beforehand a hierarchy of representation primitives in both domains (one being mostly unknown), nor to define how similarity between the two represented cases must be computed.

Domaines: Apprentissage [cs.LG]; Intelligence artificielle [cs.AI]; Education

Antoine Cornuéjols, Andrée Tiberghien, Gérard Collet. "Tunnel effects" in cognition : A transfer mechanism from known conceptual domains to new ones. Conference on Artificial Intelligence and Simulation of Behaviour (AISB-99), Apr 1999, Edinburgh, United Kingdom. ⟨hal-02482181⟩