80 likes | 109 Views
Explore the tradeoffs in believability for virtual agents, including the balance between user expectations and system capabilities, expressivity in conveying emotions, believability versus fidelity, robustness in design, consistency for personality traits, persistence in consequences, and the importance of sociability for interactive systems.
E N D
Personality and Believability Panel at the2nd Workshop on Intelligent Virtual Agents(Virtual Agents 99) The Centre for Virtual Environments University of Salford, Salford, United Kingdom 13th September 1999 paolo@ai.univie.ac.at
Believability tradeoffs:BALANCE • What is elicited from user vs.system capabilities • Can go wrong both ways! • System raises too high/complex/… expectations in users • Systems places to high demands on users in terms of competence/ duration of interaction/...
Believability tradeoffs:EXPRESSIVITY • Action expression problem(Phoebe Sengers) • Convey a summary of the “story” behind a given situation, provide clues about ongoing task • E.g.:local expressive behaviour by itself is insufficient for discrimination of current emotional state!(cf. action tendencies, Nico Frijda)
Believability tradeoffs:BELIEVABILITY vs. FIDELITY • Impact of prior knowledge,“folk theories”, prejudices,… on subjective assessment of system performance • Emphasis on believability allows better exploitation of available resources • Emphasis on fidelity (usually) results in higher robustness and better consistency (see next slides)
Believability tradeoffs:ROBUSTNESS • “Brittle depth” vs. “Robust shallowness” • Depth (may be) required for “interestingness” • Lessons from knowledge-based systems design • Semantic vs. Architectonic Space (Nancy Kaplan) • Capitalize on ambiguities/under-determination • Leave space for multiple interpretation by users • Symmetric requirements for system: • Ability to make multiple/different interpretations/appraisals • Meta-reasoning
Believability tradeoffs:CONSISTENCY • Consistent behaviour over time entails predictability and facilitatesrecognition of “personality traits” • E.g.: consistency of problem solving and action selection across different situations (difficult!?) • May be achieved via an“inverse mapping” of dimensions of consistency to system lifeworld/ architecture • “Big 3” (5) traits, action tendencies…
Believability tradeoffs:PERSISTENCE • Persisting consequences of choices, actions, events • “No undo”
SOCIA(L A)BILITIY • Basic requirement for interactivity • integration of system+user into a “bigger whole”: interactive system • How? • E.g. via social psychology models, such as: • Power&Status (T. Kemper) • Emotional Competence (C. Saarni) • Seizure/relinquishment* of control at different levels • Acceptance of broad classes of inputs(recognition of affordances) *(difficult/unusual for system designers)