Prediction and prediction error
Prof. Rob Honey, School of Psychology, Cardiff University
Time, date and venue: April 17th, 12.30, Conference Room 1, CIMCYC
Language: English
Abstract: Models of learning emphasize the importance of prediction and prediction error. This emphasis is often justified by the difference in effectiveness of forward and backward conditioning trials: When a conditioned stimulus precedes and thereby predicts the unconditioned stimulus marked changes in behaviour are observed, but when the unconditioned stimulus precedes the delivery of the conditioned stimulus the changes are weak and transient. However, our recent research suggests that the emphasis on prediction and prediction error should be revisited: Backward conditioning trials in human and nonhuman animals can affect marked changes in behaviour (Guitard, Burgess, Foo, Navarro & Honey, 2025; Navarro, Dwyer & Honey, 2023). These results provide support for Asratian’s (1965) informal account of learning and recent formal instantiations of his ideas (Honey & Dwyer, 2022; Honey, Dwyer & Iliescu, 2020).