Hilfe Wegweiser Impressum Kontakt Einloggen





A Model of Theory-Of-Mind Based on Action Prediction


Schultz,  J
Department Human Perception, Cognition and Action, Max Planck Institute for Biological Cybernetics, Max Planck Society;

Externe Ressourcen
Es sind keine Externen Ressourcen verfügbar
Volltexte (frei zugänglich)
Es sind keine frei zugänglichen Volltexte verfügbar
Ergänzendes Material (frei zugänglich)
Es sind keine frei zugänglichen Ergänzenden Materialien verfügbar

Schultz, J. (2006). A Model of Theory-Of-Mind Based on Action Prediction. Poster presented at 9th Tübingen Perception Conference (TWK 2006), Tübingen, Germany.

Theory-of-Mind, or mentalising, is defined as a cognitive process used to understand other peoples’ actions based on mental states. Twomain theories of mentalising have been put forward in recent years: Simulation Theory and Theory-Theory. We propose a model of mentalising based on action prediction and semantic representation. The model would be triggered whenever a human observer detects a potential agent (particularly other humans, but also other animals or active entities). On the basis of their actions, it would associate a possible mental state to the observed agent and predict its future behaviour. To do this, first a search engine would look for a potential mental state matching an observed action in a look-up table containing actionmental state associations acquired through experience. Then, a predictor would calculate a possible next action for the observed agent on the basis of the mental state, and a comparator would compare this predicted action to the actual next action of the agent. If the discrepance between predicted and actual behaviour is greater than a threshold, the mental state is rejected and the process repeated until a conclusive match or abandon. The predictor is postulated to be similar to mechanisms thought to underlie motor learning or reinforcement learning, while the look-up table could resemble semantic representations of objects or faces. The model could also be used for active interaction with other agents: the search engine would find an action to be executed by the observer in order to induce a particular mental state in the observed agent. Success could be assessed by the model through observation of the other agents’ reaction. The neural correlates for this model are likely to be distributed and could include the posterior part of the superior temporal sulcus, the medial prefrontal cortex, the temporal poles, the premotor cortex and the cerebellum. To assess the plausibility of the model and test possible associations between particular neural structures and the components of the model, we review previous studies of the neural correlates of mentalising and some associated processes.