Tom Griffiths

Games have long been a microcosm for studying planning and reasoning in both natural and artificial intelligence, often focusing on expert-level or even super-human play1, 2, 3, 4, 5–6. But real life also pushes human intelligence along a different frontier, requiring people to flexibly navigate decision-making problems that they have never thought about before. Here we use novice gameplay to study how people reason about new problem settings. Through a series of large-scale behavioural studies with over 1,000 participants and 121 two-player strategic board games (almost all novel to our participants), we show that people are systematic and adaptively rational in how they play a game for the first time or evaluate a game (for example, how fair or how fun it is likely to be) before they have played it even once. We explain these capacities via a computational cognitive model that we call the ‘Intuitive Gamer’: a model based on mechanisms of fast and flat (depth-limited) goal-directed probabilistic simulation. Our work offers insights into how people rapidly evaluate, act and make suggestions when encountering novel problems, and could inform the design of more flexible and human-like artificial intelligence systems that can determine not just how to solve new tasks but also whether a task is worth thinking about at all.

In the 1970s, physicist Richard Feynman turned lunch with a friend into a math problem—how to optimize dish selection over multiple meals—but his handwritten notes remained a mystery for decades. Here we present the fully deciphered problem and solution, prove its optimality, generalize it to related problems, and compare the results to human behavior. The optimal policy specifies decreasing thresholds for switching from exploring new dishes to exploiting the best, with thresholds varying based on the distribution of the quality of dishes. We connect these results to the existing psychological literature on optimal stopping problems, which has explored close variants on Feynman’s problem, and use our generalization of the solution to explore how the underlying distribution of the quality of the options influences people’s choices. A preregistered experiment with 2,520 participants shows that people adopt thresholds that decrease linearly with the proportion of trials remaining, consistent with the observation of linear thresholds in other optimal stopping problems. However, we show that people tend to explore more than predicted by linear thresholds, and that different distributions of quality result in thresholds with the same slope but different intercepts. These results indicate that people adapt linear thresholds used in optimal stopping tasks in a way that is sensitive to the underlying distribution—a simple strategy that we show is nearly as effective as Feynman’s solution.

Human communication is motivated: people speak, write, and create content with a particular communicative intent in mind. As a result, information that large language models (LLMs) and AI agents process is inherently framed by humans’ intentions and incentives. People are adept at navigating such nuanced information: we routinely identify benevolent or self-serving motives in order to decide what statements to trust. For LLMs to be effective in the real world, they too must critically evaluate content by factoring in the motivations of the source-for instance, weighing the credibility of claims made in a sales pitch. In this paper, we undertake a comprehensive study of whether LLMs have this capacity for motivational vigilance. We first employ controlled experiments from cognitive science to verify that LLMs’ behavior is consistent with rational models of learning from motivated testimony, and find they successfully discount information from biased sources in a human-like manner. We then extend our evaluation to sponsored online adverts, a more naturalistic reflection of LLM agents’ information ecosystems. In these settings, we find that LLMs’ inferences do not track the rational models’ predictions nearly as closely-partly due to additional information that distracts them from vigilance-relevant considerations. However, a simple steering intervention that boosts the salience of intentions and incentives substantially increases the correspondence between LLMs and the rational model. These results suggest that LLMs possess a basic sensitivity to the motivations of others, but generalizing to novel real-world settings will require further improvements to these models.