A bioinspired model to motivate learning of appetitive signals’ incentive value under a Pavlovian conditioning approach

Abstract

Nowadays, the general artificial intelligence field aims to emulate some human behavior features in computational models for different objectives. This article proposes a model grounded in neurosciences focusing on individual behavior under appetitive circumstances. Environmental stimuli inspire human behavior because they represent a (learned) incentive value (IV) linked to the satisfaction of motivational states (MSs). That is, the expected behavior is goal-directed. We describe the generation of the incentive learning process. To reach this objective, we propose a computational-bioinspired model exemplifying the IV encoding process of a perceived smell stimulus for its setup, recovery and updating using Pavlovian conditioning. The proposed model highlights the main processes involved in Pavlovian incentive learning (PIL) and complementaries, e.g., an early integration between a smell stimulus and an MS, the generation of a temporal reward history, the prediction of a reward, the calculation of the obtained reward, the evaluation of reward error, among others. The model’s functionality was validated in two phases: the first stage was the acquisition of the IV and the second stage was a go/nogo task based on an incentive value and its MS.