Human collective intelligence under dual exploration-exploitation dilemmas.

Area of Science:

• Behavioral economics
• Cognitive science
• Social psychology

Background:

• The exploration-exploitation dilemma presents a fundamental trade-off between seeking new opportunities and utilizing known ones, crucial for optimizing performance under uncertainty.
• While eusocial insects exhibit evolved collective mechanisms for this dilemma, humans face individual and group-level challenges, including potential 'information scrounging' via social learning.
• Social learning can reduce exploration costs but risks undermining benefits if individuals disproportionately benefit from others' discoveries.

Purpose of the Study:

• To experimentally investigate whether social learning opportunities enhance the performance of human groups facing the exploration-exploitation dilemma.
• To assess the impact of different types of social information (choice frequency vs. evaluations) on group decision-making and collective intelligence.
• To determine if human groups can effectively manage the dual exploration-exploitation dilemmas at both individual and group levels.

Main Methods:

• Human participants engaged in a group 'multi-armed bandit' task, a decision-making problem involving exploration and exploitation.
• Experimental conditions varied the availability of social learning information, specifically choice-frequency data and/or observed evaluations of chosen options.
• Group performance, exploration time, and resource harvesting were measured across different social learning conditions.

Main Results:

• Social learning opportunities significantly reduced total group exploration time and increased harvesting from better options, thereby improving overall collective performance.
• Despite the frequent emergence of 'information scroungers,' groups with social learning capabilities outperformed those without.
• Enriching social information by including observed evaluations (akin to rating systems) negatively impacted performance compared to using only choice-frequency data.

Conclusions:

• Human groups can effectively manage the dual exploration-exploitation dilemmas inherent in decision-making under uncertainty.
• Social learning, particularly through the sharing of simple choice-frequency information, facilitates collective intelligence and improves group performance.
• The type and richness of social information critically influence the effectiveness of social learning in group decision-making contexts.