Supporting Artificial Social Intelligence With Theory of Mind

Area of Science:

• Artificial Intelligence
• Cognitive Science
• Human-Computer Interaction

Background:

• Human collaboration relies on understanding others' mental states, a capability known as theory of mind.
• Current artificial agents lack sophisticated social intelligence, hindering seamless human-agent teaming.
• Artificial social intelligence (ASI) aims to equip agents with social understanding.

Purpose of the Study:

• To explore the foundational role of artificial theory of mind (AToM) in enabling agent collaboration with humans.
• To propose a framework for developing ASI by leveraging insights from human social intelligence and theory of mind.
• To enhance human-agent interaction through agents' ability to model, interpret, and predict human behaviors and mental states.

Main Methods:

• Drawing from social signals theorizing to identify core features of ASI.
• Developing a framework to guide the creation of agents with AToM.
• Analyzing how human social intelligence informs the development of artificial social intelligence.

Main Results:

• AToM provides a structure for agents to model human counterparts effectively.
• ASI requires capabilities to perceive, interpret, and generate social cues for team operation.
• Modeling AToM supports transparent communication, enhancing trust in artificial agents.

Conclusions:

• Artificial theory of mind is essential for socially intelligent agents to collaborate with humans.
• ASI development can be advanced by building upon human social intelligence principles.
• AToM fosters trust and predictability in human-agent interactions through better modeling and communication.