Page 74 - Socially Intelligent Agents Creating Relationships with Computers and Robots

Adapting to Affect and Personality 57

Table 6.1. Examples of Task-Specific Rules for Strategy Selection.

Anxiety effects
IF (recent change in radar return status) THEN (emphasize change in status)
IF (attention focus = HUD) AND (incoming radar data) THEN (redirect focus to radar)
IF (attention focus = radar) AND (Incoming radio call) THEN (redirect focus to radio)
IF (likelihood of task neglect for <instrument> = high) & (has-critical-info? <instrument>) THEN (emphasize
<instrument> visibility)
IF (target = unknown) AND (target belief = hostile) THEN (emphasize unknown status) AND (collect more data)
Aggressiveness effects
IF (likelihood of premature attack = high) THEN (display all available info about enemy a/c) AND(enhance display of
enemy a/c info)
Obsessiveness effects
IF (likelihood of delayed attack = high) THEN (display all available enemy a/c info) AND (display likelihood of enemy
attack) AND (display vulnerability envelope) AND (display reminders for attack tasks)

Strategy Selection Module. This module receives as input the predicted
specific effects of the affective and belief states, and selects a compensatory
strategy to counteract resulting performance biases. Strategy selection is ac-
complished by rule-based reasoning, where the rules map specific performance
biases identified by the Impact Prediction Module (e.g., task neglect, threat-
estimation bias, failure-estimation bias, etc.) onto the associated compensatory
strategies (e.g., present reminders of neglected tasks, present broader evidence
to counteract threat-estimation bias, present contrary evidence to counteract
failure-driven confirmation bias, etc.). Table 6.1 shows examples of task-
specific rules for compensatory strategy selection.

GUI Adaptation Module. This module performs the final step of the
adaptive methodology, by implementing the selected compensatory strategy in
terms of specific GUI modifications. A rule-based approach is used to encode
the knowledge required to map the specific compensatory strategies onto the
necessary GUI adaptations. The specific GUI modifications take into consider-
ation information about the individual pilot preferences for information presen-
tation, encoded in customized user preference profiles; for example, highlight-
ing preferences might include blinking vs. color change vs. size change of the
relevant display or icon. In general, two broad categories of adaptation are pos-
sible: content-based, which provide additional information, and format-based,
which modify the format of existing information (see Figure 6.2).

4. Results
The ABAIS prototype was implemented and demonstrated in the context
of an Air Force combat mission, used a knowledge-based approach to assess