Modelling basic needs as agent motivations

摘要

Any autonomous agent behaviour generation mechanism should incorporate as a core module, a source of internal motivation that functions as a start point for agent behaviour to commence. Intelligent virtual agents are typically respondent to external stimuli, however, their behaviour becomes repetitive and trivial when these stimuli are missing. We argue that it is necessary for virtual agents to be equipped with intrinsic motivations that energise and direct their behaviour, in order to function in a coherent and believable way. Adopting the general principles of hierarchical motivation theories, in the current work, we attempt to model physiological needs as the lowest and basic level of motivations, in a layered motivational architecture. Based on readings from physiology, we present the mechanisms underlying the function of four basic needs and propose a model that allows the incorporation of plausible human-like needs in an intelligent virtual agent.