Abstract: The talk will explore progress over the past 15 years on testing the hypothesis that human behaviour can be predicted by models that find the optimal policy for a partially observable sequential decision process with bounds imposed by ecology, cognition, and utility. Progress will be measured against Newell’s (1992) criteria for a unified theory of cognition: problem solving, decision making, routine action, memory, learning, skill, perception, motor behaviour, language, motivation, emotion, imagining, dreaming, and day-dreaming. Prospects for the future of such models in interactive AI will be discussed.

Speakers:  Andrew Howes

Affiliation: School of Computer Science. University of Birmingham

Place of Seminar:  Zoom (now available on YouTube)