Page 79 - Socially Intelligent Agents Creating Relationships with Computers and Robots

62 Socially Intelligent Agents

quires defining: (1) which forms of social intelligence these agents should have
and how they may be translated in terms of personality traits; (2) how a trait
may be represented in the agent’s mental state and reasoning style; (3) how
various traits may be combined in the same individual; and finally (4) how one
or more traits may be manifested in the agent’s behaviour. In this chapter we
discuss our experience in building an Interface Agent that cooperates with the
user in performing software application tasks; we will focus our description on
the way that we formalised its cooperation attitude.

2. Cooperation personalities in XDM-Agent

Research on personality-based Human-Computer Interaction (HCI) has be
driven by results of studies about human intelligence—in particular, the Five-
Factor Model (FFM) and the Interpersonal Circumplex Model (IC). The FFM
[10] derives from the psychologists’ need of defining “the most important ways
in which individuals differ in their enduring emotional, interpersonal, experien-
tial, attitudinal and motivational styles” [10]. The five dimensions (Extraver-
sion, Emotional Stability, Agreeableness, Openness, and Conscientiousness)
are an interpretation of results of applying factor analysis to questionnaires
submitted to various groups of subjects; their meaning is a subjective inter-
pretation of the set of variables they “explain”, and is described with natural
language terms. “Sociability” or “Social closeness” is associated, in partic-
ular, with Extraversion. The second method employed to categorise human
personalities is Wiggins’ measure of IC [13], with axes “Dominance” and “Af-
filiation”. Whether the two factorisation criteria are related is not fully clear:
some authors identify Extraversion with Dominance, while others argue that
Extraversion is best seen as located midway between Dominance and Warmth
[10].
Researchers in HCI have employed the two mentioned factorisation criteria
to enrich interfaces with a personality. Some notable examples: Nass and col-
leagues studied graphical interfaces in terms of Dominance and agent-based
interfaces in terms of Extraversion [11]; Dryer adopted the IC model [8]; An-
dré and colleagues [1] attach Extraversion and Agreeableness to the members
of their “Presentation Teams”; Ball and Breese [3] included Dominance and
Friendliness in their modelling of personality-related observable behaviour. To
computer scientists, the advantage of referring to the two mentioned models is
to have a widely accepted frame of reference, with a definition of the way that
every personality factor manifests itself in the external behaviour. The main
disadvantage is that these personality traits refer to a characterisation of com-
munication styles rather than to mental social attitudes. They are therefore
very useful for endowing agents with a “pleasant” and “believable” appear-
ance, but not to express diversification in social relationships. Another diffi-