Understanding Affective Cognition: Frontiers in modeling reasoning about others' emotions

Social life constantly requires us to decipher information about others into inferences about their emotional states: for example, we have to reason about what makes our romantic partners happy (a surprise gift?) or angry (not doing one’s chores?), and what they would do in those emotional states, in order to plan our upcoming interactions. Such affective cognition, or our ability to reason about others’ emotions, scaffolds everything from cooperation to the maintenance of social relationships. Affective cognition lies in the intersection of two foundational social cognitive topics, Theory of Mind (ToM; the ability to reason about others’ mental states) and Empathy (the ability to feel and understand others’ emotions). Although the past decade has seen much progress in understanding ToM and empathy using neuroscience (KosterHale & Saxe, 2013; Zaki & Ochsner, 2012), developmental (Meltzoff, 2011), and computational (Baker, Saxe, & Tenenbaum, 2009; Goodman & Stuhlmuller, 2013) approaches, somewhat less attention has been paid specifically to affective cognition and some of its foundational cognitive questions. How do we represent (cognitively and neurally) others’ emotional states? How do we reason with those representations? How do we make predictions and inferences about others’ future actions or desires based on their emotions? Finally, how does affective cognition shift across development? The aim of the symposium is to answer these, and other relevant questions, at the forefront of this field. In this symposium, we will discuss recent work in affective cognition done by some of the above-cited researchers and their labs who approach the topic from different disciplines and backgrounds. There will be ample time set aside at the end of the symposium for the speakers and the audience to discuss future research directions. Amy Skerry will kick off the symposium by discussing recent work on neural representations of others’ emotional states. Desmond Ong will then present a computational model of emotion attribution to other agents. Yang Wu will continue the discussion with a model of inference over an agent’s desires and beliefs from observed emotions, in both adults and children. Our final speaker, Andrew Meltzoff will describe and discuss recent developmental work on how children learn and make inferences about others’ emotional traits, and regulate their own behavior in response. Finally, open discussion, especially with regards to implications and future directions, will be led and moderated by Noah Goodman. Affective cognition is a key topic that is growing increasingly relevant to the cognitive science community in the past few years, and it is the perfect time to bring together leaders in the field of affective cognition to discuss the exciting state of the art of this field, moving forward. In addition to gaining deeper understanding of the core computations that underlie such reasoning, there are many promising applications of this work. In particular, this symposium and its applications resonate well with the theme of this year’s conference, Cognitive Science Meets Artificial Intelligence: Human and Artificial Agents in Interactive Contexts. First, having a better understand about basic human affective cognition will enable us to make progress in building interactive artificial agents that can reason about human emotions (Picard, 2000). Second, having models of working and dysfunctional affective cognition would enable us to build computational algorithms and agents to assist with diagnosing and treating clinical disorders (Montague, Dolan, Friston, & Dayan, 2012). Such applications of affective cognition are becoming more ubiquitous each year, and this dialogue between leaders in the cognitive science community will, we hope, spur future research in both the basic and applied disciplines.