Page 71 - Socially Intelligent Agents Creating Relationships with Computers and Robots

54 Socially Intelligent Agents

tive architecture capable of modeling a variety of individual differences (e.g.,
affective states, personality traits, etc.) [5], and (2) developing an adaptive user
interface capable of recognizing and adapting to the user’s affective and belief
state (e.g., heightened level of anxiety, belief in imminent threat, etc.) [4].
In this chapter we focus on the area of affective adaptation and describe an
Affect and Belief Adaptive Interface System (ABAIS) designed to compensate
for performance biases caused by users’ affective states and active beliefs. The
performance bias prediction is based on empirical findings from emotion re-
search, and knowledge of specific task requirements. The ABAIS architecture
implements a four-phase adaptive methodology: (1) assessing user affect and
belief state; (2) identifying their potential impact on performance; (3) select-
ing a compensatory strategy; and (4) implementing this strategy in terms of
specific GUI adaptations. ABAIS provides a generic adaptive framework for
exploring a variety of user assessment methods (e.g., knowledge-based, self-
reports, diagnostic tasks, physiological sensing), and GUI adaptation strate-
gies (e.g., content- and format-based). We outline the motivating psycholog-
ical theory and empirical data, and present preliminary results from an initial
prototype implementation in the context of an Air Force combat task. We con-
clude with a summary and outline of future research and potential applications
for the synergistic application of the affect-adaptive and affect and personality
modeling methodologies within SIA architectures.

2. Selecting Affective States And Personality Traits

The first step for both the modeling and the adaptation research goals is to
identify key affective and personality traits influencing behavior. The affective
states studied most extensively include anxiety, positive and negative affect,
and anger. The effects of these states range from influences on distinct in-
formation processes (e.g., attention and working memory capacity, accuracy,
and speed; memory recall biases), through autonomic nervous system mani-
festations (e.g., heart rate, GSR), to visible behavior (e.g., facial expressions,
approach vs. avoidance tendencies, etc.) [9, 7, 1]. A wide variety of per-
sonality traits have been studied, ranging from general, abstract behavioral
tendencies such as the Five Factor Model or “Big 5” (Extraversion, Emo-
tional Stability, Agreeableness, Openness, Conscientiousness) and “Giant 3”
(Approach behaviors, Inhibition behaviors, Aggressiveness) personality traits,
through psychodynamic / clinical traits (e.g., narcissistic, passive-aggressive,
avoidant, etc.), to characteristics relevant for particular type of interaction (e.g.,
style of leadership, etc.) [3, 8]. Our initial primary focus in both the modeling
and the adaptation research areas was on anxiety, aggressiveness, and obses-
siveness.