Logic and social cognition. The facts matter, and so do computational models (Q843769)

The paper offers an overview of the literature on higher-order belief and sketches some perspectives of convergence of the different approaches. A zero-order belief is a belief about a fact; a first-order belief is a belief about another agent's belief, etc.; the `limit case' of such higher-order beliefs is common belief. An agent's first-order and higher-order beliefs make up his theory of mind. The latter is instrumental for intelligent social behaviour. The author gives an extensive overview of the subject, with 171 references covering the relevant literature in cognitive psychology, empirical economics, philosophy of mind, epistemic logic, and multi-agent systems. She focusses on a fault line between epistemic logic and cognitive science: while standard epistemic logic has the omniscience property (it prescribes that agents reason perfectly with beliefs of any order), experimental evidence as collected in psychology and economics tells us that this is not the case. In particular it takes children quite a long time to move from first-order to second-order theories of mind. Moreover, even when a child \(i\) has learned to attribute to an agent \(j\) a belief about agent \(k\) then \(i\) is often unable to apply such reasoning in game-like scenarios where \(i\) has to `put to work' such beliefs in order to take decisions. She finally points out several research avenues at the intersection of logic and cognitive science, such as (1) the implementation higher-order beliefs in computational cognitive models such as the cognitive architecture ACT-R; (2) the improvement of existing logical accounts of resource-bounded reasoning (in particular the integration of dynamic epistemic logics) in view of improving its cognitive realism.