Grounded Models as a Basis for Intuitive Reasoning: the Origins of Logical Categories

Grounded models (Siena 2001b) differ from axiomatic theories in establishing explicit connections between language and reality that are learned through language games (Wittgenstein 1953). This paper describes how grounded models are constructed by autonomous agents as a side effect of their activity playing different types of language games (Steels 1999), and explains how they can be used for intuitive reasoning. It proposes a particular language game which can be used for simulating the generation of logical categories (such as negation, conjunction, disjunction, implication or equivalence), and describes some experiments in which a couple of visually grounded agents construct a grounded model that can be used for spatial reasoning.