Revisiting Rogers' Paradox in the context of human-AI interaction

Area of Science:

• Cognitive Science
• Artificial Intelligence
• Evolutionary Psychology
• Computational Social Science

Background:

• Human learning occurs through individual exploration and social transmission of information.
• Rogers' Paradox demonstrated that readily available social learning does not always enhance population fitness.
• The rise of artificial intelligence (AI) introduces a new dimension to social learning dynamics.

Purpose of the Study:

• To investigate Rogers' Paradox in the context of human-AI interaction.
• To model the collective world model equilibrium in a network of humans and AI learning together.
• To assess the impact of different learning strategies on human-AI network dynamics.

Main Methods:

• Extended agent-based simulations to incorporate both human and AI agents learning in an uncertain environment.
• Analyzed the influence of various learning strategies on the societal 'collective world model'.
• Modeled potential negative feedback loops arising from humans learning from AI.

Main Results:

• The availability of social learning from AI did not confer a fitness advantage over individual learning in the simulated network.
• Different learning strategies significantly impacted the equilibrium of the collective world model.
• Identified potential negative feedback loops in human-AI social learning scenarios.

Conclusions:

• Human-AI social learning dynamics present novel challenges and considerations beyond traditional social learning models.
• Understanding and managing learning strategies is crucial for optimizing human-AI collaborative intelligence.
• The simulation framework provides a basis for exploring future research on human-AI cultural evolution and world models.