Gene Ball

Panel on affect and emotion in the user interface

INTRODUCTION Intelligence. So much of our technology revolves around intelligence: technology in support of intellectual activities; the goal of engineering artificial intelligence; the need for intelligence in the user interface. And yet, so much of everyday life is really about affect and emotion: differences in performance under conditions that are supportive, threatening, or punishing; the challenges of conflict resolution and cooperation among heterogeneous groups of people; the implicit messages of body language and conversational style; the spirit-sustaining texture of our affective relationships with family and friends.

Relating personality and behavior: posture and gestures

We have constructed a Bayesian model relating personality and emotion to externally observable behaviors, designed to be useful in generating natural and convincing communicative behaviors in a conversational agent. The same model can be used to diagnose the internal emotional state and personality type of the human user. This paper briefly recounts the motivation and structure of the overall model, and then considers the relationship between personality type and posture and gestures in more detail. As is well established in the psychology literature, people recognize characteristic body motions as reliable indicators of the personality type of others. We review the most likely sources of evidence for those judgments, and consider the feasibility of generating behavior in an animated computer agent that presents a consistent personality, as well as the more difficult task of recognizing personality type based on body movement.

Mixing scripted interaction with task-oriented language processing in a conversational interface

Natural conversational interaction with computers will require systems that can successfully process unconstrained spoken input. Within the domain of its competency, such a system must be able to process an utterance in a “deep” fashion, extracting the detailed information necessary to carry out a useful task. When users stray outside the supported domain, the system must still be able to respond to a “broad” range of plausible inputs to maintain basic conversational competency. This paper reports on an effort to combine simple pattern matching techniques which can provide broad coverage with deep processing based on robust natural language template matching.