Abstract

Current theories of emotions and emotional machines often assume too homogeneous a conception of what emotions are in terms of whether they are experienced as one's own emotions (“internal” emotions) or whether they are perceived as the emotions of other agents (“external” emotions). In contrast, this paper argues, first, that an answer to the question of whether machines can possess emotions requires such a distinction—the distinction between internal emotions-had and external emotions-perceived. Second, it argues that the emotions we perceive in other agents can be explicated as indicators of likely imminent paths or patterns of behavior of those agents. As will be shown, perceiving emotions in others does not necessarily involve the ontological attribution of corresponding emotions-had to the subject which is perceived as exhibiting a particular emotion-perceived. Thus, it will be shown that we can, for example, perceive an agent as angry without (ontologically) attributing anger (as an emotion-had) to that agent. If we apply this reasoning to emotional robots, it follows that there is no deception involved in perceiving a robot as exhibiting a particular emotion, as long as its behavior realizes and continues to realize the behavioral form of that external emotion-perceived. Thus, a robot’s behavior can fully instantiate emotions-perceived. In order to elaborate these claims, the presented view will be contrasted with Gibson’s conception of “affordances.” The final section discusses whether these considerations might have broader implications for our view of human perception in general.