AI Insight
This study investigates how humans decide between learning from others (social learning) versus learning through direct experience. Researchers developed a "Rational Mentalizing" model that uses Theory of Mind—the ability to understand others' mental states—to predict when people will choose social learning by estimating how informative another person's actions will be. Testing this model with a novel game, they found it accurately predicted human decisions about when to observe others versus explore independently, suggesting people strategically choose social learning based on its expected utility.
Why it matters
Understanding how humans balance social and non-social learning could improve educational strategies, training programs, and the design of AI systems that learn from human demonstrations. This research provides a computational framework for predicting when people will rely on others for information versus seeking direct experience.
Understand the Science
arXiv:2607.28601v1 Announce Type: cross
Abstract: Social learning is a powerful mechanism through which agents learn about the world from others. However, humans sometimes choose direct experience over social learning, which can carry time and cognitive resource costs. How do people balance social and non-social learning? We propose a Rational Mentalizing model of the decision to engage in social learning. This model estimates the utility of social learning by reasoning about another agent’s goal and the informativeness of their future actions. It then weighs the utility of social learning against the utility of non-social learning. Using a novel game where players choose between observing other agents or exploring the environment, we show that the Rational Mentalizing model can quantitatively capture human trade-offs between these strategies. These findings suggest that selective social learning is guided by ‘Theory of Mind’ in the service of utility maximization.
Source: Using Theory of Mind to Arbitrate between Social and Non-social Learning