Carson Reynolds, Rosalind W. Picard

Abstract

By integrating sensors and algorithms into systems that are adapted to the task of interpreting emotional states, it is possible to enhance our limited ability to perceive and communicate signals related to emotion. Such an augment ation would have many potential beneficial uses in settings such as education, hazardous environments, or social contexts. There are also a number of important ethical considerations that arise with the comput er's increasing ability to recognize emotions. This paper will survey existing approaches to computer ethics relevant to affective computing. We will categorize these existing approaches by relating them to different metaethical positions. The goal of this paper is to situate our approach among other approaches in the computer ethics literature and to describe its methodology in a manner that practitioners can readily apply. The result then of this paper is a process for critiquing and improving affective computing systems.