Emergence of Emotional Appraisal Signals in Reinforcement Learning Agents

Pedro Sequeira, Francisco S. Melo and Ana Paiva
Autonomous Agents and Multi-Agent Systems, 29(4), 2015, pp. 537-568. DOI 10.1007/s10458-014-9262-4 Emergence of Emotional Appraisal Signals in Reinforcement Learning AgentsEmergence of Emotional Appraisal Signals in Reinforcement Learning Agents

Abstract The positive impact of emotions in decision-making has long been established in both natural and artificial agents. Emotions complement the perceptual information acquired through the agent’s sensors, coloring our sensations and thus guiding our decision- making. However, when designing autonomous agents, are emotions the best complement to the perceptions? Mechanisms investigated in affective neuroscience provide support for this hypothesis in biological agents. In this paper, we look for similar support in artificial systems. We adopt the intrinsically motivated reinforcement learning framework (IMRL) to investigate different sources of information that can guide decision-making in learning agents, and an evolutionary approach based on genetic programming to identify a small set of such sources that have the largest impact on the performance of the agent. We then show that these sources of information: (i) are applicable in a wider range of environments than those where the agents evolved; (ii) exhibit interesting correspondences to appraisal-like signals previously proposed in the literature, pointing towards our departing hypothesis that emotions might indeed provide essential information to complement perceptual capabilities and thus guide decision-making.



Note: Accepted manuscript version. The final publication is available at link.springer.com.


Eat-all-pellets scenario


PacMan scenario


Power-pellet scenario


Rewarding-pellets scenario

Videos of the experiments.
Posted on