Bio-Inspired Virtual Populations: Adaptive Behavior with Affective Feedback

In this paper, we describe an agency model for generative populations of humanoid characters, based upon temporal variation of affective states. We have built on an existing agent framework from Sequeira et al. [17], and adapted it to be susceptible to temperamental and emotive states in the context of cooperative and non-cooperative interactions based on trading activity. More specifically, this model operates within two existing frameworks: a) intrinsically motivated reinforcement learning, structured upon affective appraisals in the relationship of the agents with their environment [19,17]; b) a multi-temporal representation of individual psychology, common in the field of affective computing, structuring individual psychology as a tripartite relationship: emotions-moods-personality [7,15]. Results show a populations of agents that express their individuality and autonomy with a high level of heterogeneous and spontaneous behaviors, while simultaneously adapting and overcoming their perceptual limitations.This project has received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 655226