Affective computing: challenges

https://doi.org/10.1016/S1071-5819(03)00052-1Get rights and content

Abstract

A number of researchers around the world have built machines that recognize, express, model, communicate, and respond to emotional information, instances of “affective computing.” This article raises and responds to several criticisms of affective computing, articulating state-of-the art research challenges, especially with respect to affect in human-computer interaction.

Introduction

Many people think of Spock, the half-Vulcan, half-human character of Star Trek, as the patron saint of computer science. He was highly intelligent, highly rational, highly unemotional, and (at least at the university where I work) attractive to women. A popular image is that Spock did not have emotions: after all he almost never expressed emotion, excepting his characteristic pronouncement of the word “fascinating” upon pondering something new. In fact, as the actor Leonard Nimoy describes in his book (Nimoy, 1995) the character Spock did have emotion; he was just very good at suppressing its expression. But most people think of Spock as not having emotion. When someone never expresses emotion, it is tempting to think that emotion is not there.

In affective computing, we can separately examine functions that are not so easily separated in humans. For example, the Macintosh has been displaying a smile for years upon successful boot-up. But few people would confuse its smile—albeit an emotional expression—with a genuine emotional feeling. Machines can fake the appearance of an emotion quite well, without having any feelings similar to those we would have: They can separate expression from feeling. With a machine it is easy to see how emotion expression does not imply “having” the underlying feeling.

Machines that might actually “have” feelings is the key area of affective computing that I expressed serious doubt about the 1997 book Affective Computing (Picard, 1997). I think the discussions there, and in a later book chapter (Picard, 2003) on this topic are still timely and I will not plan to add to them here. For those who find it absurd that anyone would even propose giving machines such abilities, it might help to know that researchers in the last decade have obtained dozens of scientific findings illuminating important roles of emotion in intelligent human functioning, even when it looks like a person is showing no emotion. These findings have reshaped scientific understanding of emotion, inspiring many to consider that emotional mechanisms might be more valuable than previously believed. Consequently, a number of researchers have charged ahead with building machines that have several affective abilities, especially: recognizing, expressing, modelling, communicating, and responding to emotion. And, within these areas, a number of new criticisms and challenges have arisen. The rest of this paper addresses such matters.

Section snippets

Sensing and recognizing emotion

Criticism 1: The range of means and modalities of emotion expression is so broad, with many of these modalities being inaccessible (e.g., blood chemistry, brain activity, neurotransmitters), and many others being too non-differentiated. This makes it unlikely that collecting the necessary data will be possible or feasible in the near future.

Criticism 2: People's expression of emotion is so idiosyncratic and variable, that there is little hope of accurately recognizing an individual's emotional

Affect modelling

Criticism 3: Little real progress has been made with cognitive modelling. How can we hope to make progress in modelling the much more subtle, multimodal, and idiosyncratic processes that characterize emotional responses.

While some data exist regarding the effects of emotion on decision making and behavior, the majority of these data were collected in highly artificial lab environments, and both their robustness and their generalizability is questionable. IN addition, within the emotion research

Emotion expression

Criticism 4: The sine qua non of emotion expression is the physical body. To the extent that computers do not have physical bodies, they cannot reliably and believably express emotion. Existing attempts at expressing emotions in robots, which are at times referred to as embodied, are unrealistic and therefore unconvincing, and unable to generate the type of affective responsiveness characterizing human–human interaction.

The opening statement of this criticism rings true; but in a superficial

Ethics

Criticism 5: Emotions, perhaps more so than thoughts, are ultimately personal and private. Providing information about the most intimate motivational factors and reactions. Any attempts to detect, recognize, not to mention manipulate, a user's emotions thus constitutes the ultimate breach of ethics and will never be acceptable to computer users. Attempts to endow computers with these abilities will lead to widespread rejection of such computer systems and will help promote an attitude of

Utility of considering affect in HCI

Criticism 6: Airplanes do not flap their wings. Just because humans have evolved to use emotional abilities, and just because human–human interaction is replete with affective reactions, does not mean that computers should aspire to emulate what may in fact be an evolutionary dead-end or error. Emotions and passions tend to be more problematic than helpful in human–human interaction. There is no need to contaminate purely logical computers with emotional reactiveness.

First, let me address the

Acknowledgments

The ‘criticisms’ statements and questions were written by Eva Hudlicka and aim to represent a ‘devil's advocate’ perspective on affective computing research.

References (15)

  • R.W. Picard et al.

    Computers that recognise and respond to user emotiontheoretical and practical implications

    Interacting with Computers

    (2002)
  • Bickmore, T., 2003. Relational Agents: Effecting Change through Human–Computer Relationships. Ph.D. thesis, MIT Media...
  • Breazeal, C., 2003....
  • A. Damasio

    Descartes’ ErrorEmotion, Reason, and the Human Brain

    (1994)
  • A.M. Isen

    Positive Affect and Decision Making

  • J. Kagan

    The Nature of the Child

    (1984)
  • J.E. LeDoux

    The Emotional Brain

    (1996)
There are more references available in the full text version of this article.

Cited by (0)

View full text