Page 256 - Designing Sociable Robots

breazeal-79017 book March 18, 2002 14:24

Grand Challenges of Building Sociable Robots 237

explores pre-linguistic communication where important paralinguistic cues such as gaze
direction and facial expressions are used to perform key social skills, such as directing
attention and regulating turn-taking during face-to-face interaction between the human and
the robot. In related robotics work, an upper-torso humanoid robot called Robita can track
the speaking turns of the participants during triadic conversations—i.e., between the robot
and two other people (Matsusaka & Kobayashi, 1999). The robot has an expressionless face
but is able to direct its attention to the appropriate person though head posture and gaze
direction, and it can participate in simple verbal exchanges.

Personal recognition This challenge problem concerns the recognition and representa-
tion of people as individuals who have distinct personalities and past experiences. To quote
Dautenhahn (1998, p. 609), “humans are individuals and want to be treated as such.” To
establish and maintain relationships with people, a sociable robot must be able to identify
and represent the people it already knows as well as add new people to its growing set of
known acquaintances. Furthermore, a sociable robot must also be able to reflect upon past
experiences with these individuals and take into account new experiences with them.
Toward this goal, a variety of technologies have been developed to recognize people in
a variety of modalities such as visual face recognition, speaker identification, fingerprint
analysis, retinal scans, and so forth. Chapter 1 mentions a number of different approaches
for representing people and social events in order to understand and reason about social situ-
ations. For instance, story-based approaches have been explored by a number of researchers
(Schank & Abelson, 1977; Bruner, 1991; Dautenhahn & Coles, 2001).

Theory of mind This challenge problem addresses the issue of giving a robot the ability
to understand people in social terms. Specifically, the ability for a robot to infer, repre-
sent, and reflect upon the intents, beliefs, and wishes of those it interacts with. Recall that
chapter 1 discussed the theory of mind competence of humans, referring to our ability to
attribute beliefs, goals, percepts, feelings, and desires to ourselves and to others. I outlined
a number of different approaches being explored to give machines an analogous compe-
tence, such as modeling these mental states with explicit symbolic representations (Kinny
et al., 1996), adapting psychological models for theory of mind from child development
to robots (Scassellati, 2000a), employing a story-based approach based on scripts (Schank &
Abelson, 1977), or through a process of biographic reconstruction as proposed in
(Dautenhahn, 1999b).

Empathy This challenge problem speaks to endowing a robot with the ability to infer,
understand, and reflect upon the emotive states of others. Humans use empathy to know
what others are feeling and to comprehend their positive and negative experiences. Brothers
(1989, 1997) views empathy as a means of understanding and relating to others by wilfully