Preparing Computers for Affective Communication: A Psychophysiological Concept and Preliminary Results

Article excerpt

INTRODUCTION

Computer technologies have advanced rapidly over the past decades, and their progress continues to accelerate. Computers have become smaller, faster, and, more importantly, much cheaper, so that they are now a relatively inexpensive commodity. Along with the technological progress of computer hardware, interfaces have also evolved into a more user-friendly format. This contributes substantially to the wide-spread use of computers. Furthermore, the focus in designing computer interfaces has shifted from the viewpoint of developers to the need of prospective users. Computers have become wearable, and a number of unusual and interesting devices such as handheld twiddler keyboards for input and private eyes for output are available.

Despite the rapid progress in establishing user-centered computer hardware and software, human-computer interaction (HCI) still lacks an essential part of human-human interaction. If people exchange information face to face, affect recognition and expression are implicitly included and play an important role in the success of such interactions (e.g., Isen, 1990; Lazarus, 1990). It is not uncommon for users to interact with their computer as if it were human. However, computer-based communication is still affect blind. For example, both novices and experts are prone to the development of negative emotions if the computer system does not act in accordance with their expectations. Nevertheless the system will continue to display its information, regardless of the user's frustration. Even experienced users may opt for a distancing/avoidance coping strategy against a computer that does not respond to their aversive feelings (Saarni, 1997). Thus disregarding human emotions in the design and development of computer interfaces may constitute a critical factor in HCI.

It has been empirically demonstrated that human emotions play an essential role in decision making, perception, learning, and a variety of other cognitive functions. A number of studies in social and cognitive psychology have investigated the effects of emotions on simple and complex tasks (for a review, see Isen, 1990). However, very limited research has been conducted in contexts in which computers are necessarily involved to perform such tasks, and the results remain contradictory (Kaufmann & Vosburg, 1997).

Several studies have demonstrated the beneficial effect of negative emotions on the performance of computational and other monotonous tasks, whereas others did not (for a review, see Forgas, 1995). Stone and Kadous (1997) reported that increasing task-related negative affect led to a more frequent use of scanning strategies, which increased the accuracy of choices in easy tasks but led to performance decrement in difficult tasks. Good moods often lead to faster, simpler, more flexible, and more superficial processing strategies, whereas negative moods cause slower, more systematic, and more vigilant processing strategies (Forgas, 1999). On the other hand, positive emotions may have detrimental effects on HCI as well, because happy people tend to simplify the problem space and employ shortcuts in their decision making (Isen & Means, 1983). In general, people tend to be more intuitive in their judgment when they are happy and more analytical when they are sad (Schwarz, 1990). Therefore, a careful investigation of optimal emotional states for a particular HCI has to be performed before designing computers that will respond appropriately to human emotions.

Exciting research is currently underway at the Massachusetts Institute of Technology Media Lab, where Picard (1997) has challenged the issue of affective computing. Affective computers will be designed not only to recognize the user's emotion but also to express some changes of their own state that may resemble what humans expect from a human counterpart in an emotion-sensitive interaction (Picard, 1997, pp. 60 ff.). Because a continuous monitoring of the user's current affective state by the computer system is essential to this approach, subjective reports that interrupt HCI cannot be used for this purpose. …