Evolution of Affect and Communication

Introduction

Affect, or more precisely, affective control, is wide-spread in nature. From simple homeostatic control, to need-based control, to simple mood-based control, to basic and complex emotional control, and various other forms, affective control mechanisms of varying complexity underlie all behavior in animals. In humans, affective states are deeply intertwined with cognition and are an essential part of human mentality. And affective states often play a critical role in social behavior, from simple displays of prowess, to sexual attraction, to aggressive encounters, to social attachment, and many others. The challenge for cognitive science and its various defining disciplines including philosophy, psychology, artificial intelligence, and neuroscience is to explain what affective control is, what kinds of affective control occur in nature, how affective control can be implemented, and how it is implemented in biological organisms – we call this the “affect challenge”.

The “affect challenge” comes on the heels of much conceptual disagreement in psychology alone (but also in philosophy and artificial intelligence) about what affect concepts are. For example, the difference between moods and emotions has been explained in various non-exclusive ways: Ekman (1994) only sees them as differing in terms of time-scale with moods being longer-lasting than emotions, while for Davidson (1994) emotions bias actions while moods bias cognition; yet another explanation is offered by Frijda (1994) who distinguishes moods and emotions based on their intentionality, i.e., emotions have an object towards which they are directed, while moods are non-intentional states. To appreciate the extent of the disagreement, one does not even have to compare classes of affective states such as emotions or moods; it suffices to look at any of the classes itself, e.g., the class of emotions. As succinctly put by Delancey (2002, p. 3), “there probably is no scientifically appropriate class of things referred to by our term emotion. Such disparate phenomena—fear, guilt, shame, melancholy, and so on—are grouped under this term that it is dubious that they share anything but a family resemblance.” And, in fact, several authors have noted that there is not even agreement about what “basic emotions” are supposed to be (e.g., Ortony & Turner, 1990; Griffiths, 1997).

The “affect” challenge is, however, not limited to understanding affect concepts and the functional role of affective control processes instantiating these affect concepts in agent architectures. It also includes giving accounts of why affect is so pervasive in nature, and thus why certain forms of affective control might be better than other forms of control. In fact, we believe that understanding the dynamics of affective control processes and their utility for controlling and managing an agent’s body (e.g., against the backdrop of survival and procreation) will help in answering many open questions about the nature of individual and social behavior. For example, understanding the nature of affective control will help elucidate the different ways in which affective and cognitive processes (such as reasoning, problem-solving, and decision-making) can interact. Moreover, a detailed account of affective control processes in individuals will also help to explain the dynamics and regulatory roles of emotion processes in social interactions (e.g., in aggressive exchanges). Most importantly, the utility and limits of affective control will allow us to determine, at least in part, possible evolutionary pressures leading to the evolution of higher-level cognition.

While we are currently a long way from being able to answer the above questions, we believe that it is possible to make headway on a smaller set of questions whose answers will contribute to making progress on the larger picture. For example, we can investigate whether the display of affective states will lead to better performance in cooperative tasks or better conflict resolution strategies in competitive tasks in multi-agent environments. We can also attempt to determine which affective control states will or are likely to evolve in cooperative and competitive multi-agent environments. And we can investigate the trade-offs between simple affective control and more complex deliberative control requiring representational mechanisms in the control architecture.