Page 257 - Designing Sociable Robots

breazeal-79017 book March 18, 2002 14:24

238 Chapter 13

changing one’s own emotional and psychological state to mirror that of another. It is a
fundamental human mechanism for establishing emotional communication with others.
Siegel (1999) describes this state of communication as “feeling felt.” More discussions of
empathy in animals, humans, and robots can be found in Dautenhahn (1997).
Although work with Kismet does not directly address the question of empathy for a robot,
it does explore an embodied approach to understanding the affective intent of others. Recall
from chapter 7 that a human can induce an affective state in Kismet that roughly mirrors
his or her own—either through praising, prohibiting, alerting, or soothing the robot. Kismet
comes to “understand” the human’s affective intent by adopting an appropriate affective
state.
For technologies that must interact socially with humans, it is acknowledged that the
ability to perceive, represent, and reason about the emotive states of others is important. For
instance, the field of Affective Computing tries to measure and model the affective states
of humans by using a variety of sensing technologies (Picard, 1997). Some of these sensors
measure physiological signals such as skin conductance and heart rate. Other approaches
analyze readily observable signals such as facial expressions (Hara, 1998) or variations in
vocal quality and speech prosody (Nakatsu et al., 1999). Several symbolic AI systems, such
as the Affective Reasoner by Elliot, adapt psychological models of human emotions in order
to reason about people’s emotional states in different circumstances (Elliot, 1992). Others
explore computational models of emotions to improve the decision-making or learning
processes in robots or software agents (Yoon et al., 2000; Velasquez, 1998; Canamero,
1997; Bates et al., 1992). Our work with Kismet explores how emotion-like processes can
facilitate and foster social interaction between human and robot.
Autobiographic memory This challenge problem concerns giving a robot the ability to
represent and reflect upon its self and its past experiences. Chapter 1 discussed autobi-
ographical memories in humans and their role in self-understanding. Dautenhahn (1998)
introduces the notion of an autobiographic agent as “an embodied agent that dynamically
reconstructs its individual ‘history’ (autobiography) during its lifetime.”
Autobiographical memory develops during the lifetime of a human being and is socially
constructed through interaction with others. The social interaction hypothesis states that
children gradually learn the forms of how to talk about memory with others and thereby learn
how to formulate their own memories as narratives (Nelson, 1993). Telling a reasonable
autobiographical story to others involves constructing a plausible tale by weaving together
not only the sequence of episodic events, but also one’s goals, intentions, and motivations
(Dautenhahn, 1999b). Cassell and Glos (1997) have shown how agent technologies could
be used to help children develop their own autobiographical memory through creating and
telling stories about themselves. A further discussion of narrative and autobiographical
memory as applied to robots is provided in (Dautenhahn, 1999b).