Discussions at the AAAI FS
Socially Intelligent Agents - The Human in the Loop
3-5 November 2000, Sea Crest Resort, North Falmouth, MA, USA.
Summarised by Kerstin Dautenhahn who is solely responsible for English spelling and grammatical errors, omissions and wrong interpretations.
Challenge Problems in Research on Socially Intelligent Agents
Effective/Affective Interfaces and Models of "whole" human beings (with personality, emotions, TOM, self-identity etc.)
Building an agent that can dynamically adapt to the user
Life-long socially situated learning/adaptation
Agents with interaction histories
Agents with a "life"
Truly embodied agents (is it sufficient to have a physical body?)
Evolvable social agents/relationships
User in the design loop
Adequate evaluation techniques
Interactive plots/drama with believable narratives
Challenge Scenarios for Socially Intelligent Agents:
Design spaces (dimensions e.g. autonomy/control )
"human in the loop" (personality, culture)
Examples of Socially Intelligent Agents
An agent that "cares" about me
An agent that can have conversations with me
An agent that has an "inner life" (possibly more believable for physical, robotic systems, "knowing where the brain is?")
Agents that are "happy" with their job
An agent that has common sense
An agent that recognises and remembers me individually
An agent that understand social scenarios (e.g. "having dinner")
An agent that can recognise subtle differences in emotions
An agent that can explain its behaviour/emotional states etc.
Agents that advertise/make explicit their social and cognitive competencies