The SocialAI school: a framework leveraging developmental psychology toward artificial socio-cultural agents

Area of Science:

• Developmental Psychology
• Artificial Intelligence
• Cognitive Science

Background:

• Socio-cognitive abilities are crucial for human intelligence and cultural evolution.
• Current AI research on social agents often overlooks developmental psychology, focusing on culture emergence rather than entry.

Purpose of the Study:

• To bridge AI research and developmental psychology by studying socio-cognitive abilities for cultural entry.
• To propose a broader framework of socio-cognitive concepts for artificial social intelligence.
• To introduce The SocialAI School as a tool for empirical research.

Main Methods:

• Drawing inspiration from developmental psychologists Michael Tomasello and Jerome Bruner.
• Defining key socio-cognitive concepts: social cognition, communication, social learning, formats, and scaffolding.
• Utilizing The SocialAI School, a customizable environment for AI agents (RL and LLMs).

Main Results:

• Demonstrated the versatility of The SocialAI School through case studies with both RL and LLM agents.
• Showcased the tool's capability to facilitate research on socio-cognitive abilities in AI.

Conclusions:

• Advocates for developmental psychology-informed AI research on social intelligence.
• Provides a foundational framework and a practical tool for investigating how AI agents can enter and learn within cultural environments.