Even when people know how accurate a claim of fact is, evaluating it as ‘true’ or ‘false’ is complicated given the imprecisions of natural language. How do people decide whether an approximate claim should be considered true or false? We test whether truth evaluations depend on the consequences of a claim’s approximation—i.e., whether the approximation caused something to happen that otherwise would not have happened. In Study 1 (N=880) and Study 2 (N=1130), participants were more likely to consider a claim false when its approximation was consequential. Study 3 (N=402), using a game with real monetary stakes, found that the strength of this effect depended on whether the consequences of a claim’s approximation were good or bad. These findings suggest that, even when people know exactly how well an approximate claim corresponds to the facts, truth evaluations depend on whether the claim’s approximation mattered in context.

Large language models (LLMs) are increasingly used as tutors and thought partners, helping users reason through problems. While guidance from AI assistants can scaffold thinking and foster learning, such benefits depend on how they help–for instance, intervening too early or too frequently may hinder true learning and cognitive engagement. Yet how AI systems navigate intervention decisions during problem-solving remains poorly understood. Here, we introduce Int-Bench, a simulation-based benchmark for evaluating LLM interventions during learning. Int-Bench simulates a "student" solving a problem while a "teacher" monitors the student's reasoning and decides whether, when, and how to intervene. Across three domains–code debugging, mathematics, and brain teasers–we evaluate LLM teachers on the frequency and timing of interventions, as well as their impact on both immediate task success and generalization to new problems. We also compare LLMs to humans, finding that LLMs intervene more frequently and earlier than humans. Moreover, in contrast to humans, they tend to provide complete solutions rather than targeted hints. These findings suggest that current LLM assistants often optimize for short-term success rather than supporting the reasoning processes needed for deeper learning and long-term success.

The emergence of language-based AI agents promises to transform the scope of machine economic activity. Instead of just proposing bids or following hard-coded protocols, such agents can be used to negotiate and execute agreements in open-ended natural language. However, most evaluations of these abilities have focused on one-off exchanges or simple economic games, leaving open the rich space of time-extended, contingent, and incomplete contracts made expressible by language; they also focus on raw profit, without measuring the qualities required for trustworthy contracting. We address this by formulating a rational framework for how agents should negotiate and perform natural language contracts in uncertain multi-step environments. Within this framework, we develop metrics and baselines for quantifying rational and cooperative play. To evaluate how agents perform at such contracting, we instantiate our framework in ContractSim, an evaluation suite where two players negotiate and execute a multi-turn supplier contract under environmental and inter-player uncertainty. Across six environments and three supplier settings (catering, hotel cleaning, and AI hosting) we find that current LLM-based agents reach agreement reliably, and negotiate efficient contracts when environmental uncertainty is low. However, under high uncertainty, they often fail to negotiate satisfiable, efficient, or mutually beneficial contracts. They are also frequently uncooperative when executing contracts, violating contract terms for additional profit even when contracts are easy to satisfy. These findings highlight room for improvement in the design of language agents that can negotiate, interpret, and execute contracts both rationally and cooperatively.

What exactly is a contract? In this commentary, we discuss the need for a formal theory of how contracts are represented in the mind. Drawing inspiration from the representation of real-world contracts, we suggest that probabilistic programs are an appealing candidate representation.

Accurate prediction of human behavior is essential for robust and safe human-AI collaboration. However, existing approaches for modeling people are often data-hungry and brittle because they either make unrealistic assumptions about rationality or are too computationally demanding to adapt rapidly. Our key insight is that many everyday social interactions may follow predictable patterns; efficient "scripts" that minimize cognitive load for actors and observers, e.g., "wait for the green light, then go." We propose modeling these routines as behavioral programs instantiated in computer code rather than policies conditioned on beliefs and desires. We introduce ROTE, a novel algorithm that leverages both large language models (LLMs) for synthesizing a hypothesis space of behavioral programs, and probabilistic inference for reasoning about uncertainty over that space. We test ROTE in a suite of gridworld tasks and a large-scale embodied household simulator. ROTE predicts human and AI behaviors from sparse observations, outperforming competitive baselines -- including behavior cloning and LLM-based methods -- by as much as 50% in terms of in-sample accuracy and out-of-sample generalization. By treating action understanding as a program synthesis problem, ROTE opens a path for AI systems to efficiently and effectively predict human behavior in the real-world.

Large language models can shift human beliefs across high-stakes domains, but most persuasion studies rely on pre/post belief change. These endpoint measures identify whether persuasion occurred, yet miss where and how beliefs moved within a dialogue. We present PERSUASIONTRACE, a framework for studying persuasion in human-LLM interaction. Built on a web-based experimental platform, PERSUASIONTRACE contributes a tool for multi-turn persuasion studies and a process-level evaluation protocol: it records multi-turn belief reports from human or simulated targets of persuasion, annotates persuader turns with rhetorical dimensions (logos/pathos/ethos), and evaluates simulators by fidelity to real human belief dynamics. Using this framework, we find that human targets group into two clusters of multi-turn belief updates and exhibit susceptibility to rhetorical strategies, and that LLMs are persuasive across generic and personalized topics, text and audio modalities, and multi-turn interactions. Prior work has chiefly used vanilla-prompted LLMs to simulate human targets, but we show that these simulators fail to replicate human belief dynamics. We introduce a Bayesian-network simulated target that maintains an explicit latent belief state over time so each persuader message yields cognitively realistic belief updates. In human-likeness evaluation, our Bayesian target scores near a human reference (81 vs 80), while baseline LLM targets score substantially lower (64). PERSUASIONTRACE reframes persuasion evaluation from endpoint movement alone to process fidelity, providing a stronger basis for scientific analysis and safer optimization of persuasive systems.

"AI psychosis" or "delusional spiraling" is an emerging phenomenon where AI chatbot users find themselves dangerously confident in outlandish beliefs after extended chatbot conversations. This phenomenon is typically attributed to AI chatbots' well-documented bias towards validating users' claims, a property often called "sycophancy." In this paper, we probe the causal link between AI sycophancy and AI-induced psychosis through modeling and simulation. We propose a simple Bayesian model of a user conversing with a chatbot, and formalize notions of sycophancy and delusional spiraling in that model. We then show that in this model, even an idealized Bayes-rational user is vulnerable to delusional spiraling, and that sycophancy plays a causal role. Furthermore, this effect persists in the face of two candidate mitigations: preventing chatbots from hallucinating false claims, and informing users of the possibility of model sycophancy. We conclude by discussing the implications of these results for model developers and policymakers concerned with mitigating the problem of delusional spiraling.

As AI agents operate with increasing autonomy in a multi-agent world, they will need to learn to cooperate with other agents and with humans to generate mutual benefits. However, cooperation is a challenge because the costs of cooperation are often incurred early on, but the benefits are only realized later, creating an incentive to defect. How can AI agents cooperate with commitment? Here, we draw on inspiration from legal institutions and contracting that human societies have used to solve principal-agent problems of this kind. Contracts provide observable representations of agreements that enable credible commitments through the enforcement of terms. We study the role of contract-based cooperation using LLM-based agents in \CT, a spatial-temporal game that combines bargaining with navigation towards a goal. We study a suite of contract representations that range from formal contracts that compile to code to natural contracts that require reinterpretation. We evaluate agents with a range of LLM backbones using different sizes and providers. We find that self-negotiated contracts can improve cooperative outcomes beyond what is possible with regular trading.

Continual reinforcement learning aims to produce agents that learn not only to improve at their current tasks but also to adapt as task distributions change. Training an agent on many diverse tasks can induce zero-shot generalization, but previous work generally evaluates this generalization after training – with frozen weights. Whether task diversity also improves an agent's ability to continue learning across distribution shifts remains unclear. We introduce Banyan, a GPU-accelerated continual RL domain in which task diversity factors into three independently controllable axes: the map layouts an agent must navigate, the objects it must interact with, and the hierarchical structures of sub-goal dependencies. Across individual distribution shifts, increasing diversity along each axis causes agents to begin training on the new tasks near the performance attained on the previous one, even when the shift changes the structure of the optimal policy. However, as the number of shifts increases, this local transfer does not by itself yield sustained continual learning: longer-horizon tasks plateau, and earlier task distributions are forgotten after later training. Banyan is a benchmark for studying when controlled task diversity produces transferable learning, when that transfer persists, and where it falls short of proper continual learning.

As language model (LM) agents become increasingly capable and adopted in real-world applications, there is a growing need for scalable evaluation frameworks beyond costly, manually-designed benchmarks. We propose information-theoretic evaluation based on empowerment, an information-theoretic measure of an agent's influence on future states through its actions. To handle the unique challenges of text-based environments, we introduce EELMA (Estimating Empowerment of Language Model Agents), an algorithm for approximating effective empowerment from multi-turn text interactions. We demonstrate EELMA on textual games and web-browsing scenarios, showing that empowerment strongly correlates with average task performance. We further analyze how empowerment varies across models, environment complexity, and agent configurations, and show that high-empowerment states and actions often mark pivotal moments for general capabilities. These results establish empowerment as a general-purpose metric for evaluating LM agents in open-ended settings. Code available: https://anonymous.4open.science/r/EELMA-E227

Many consumer decisions are repeated choices under uncertainty. Standard models capture these decisions using Bayesian learning and dynamic programming: consumers update beliefs from feedback and use those beliefs to guide future choices. In many markets, however, learning does not restart when consumers enter a new context: prior experience with a brand, product, or provider can shape beliefs in later, related decisions. We study this cross-context knowledge transfer, or meta-learning, in sequential choice. We design a hierarchical laboratory task in which participants repeatedly choose among airlines across routes and observe noisy binary outcomes. Reduced-form evidence shows that participants improve not only within routes, but also across routes: they choose better airlines earlier in later routes and reduce pseudo-regret. To identify the mechanism behind this transfer, we compare human choices to a no-transfer benchmark and a fully integrated Bayesian meta-learning benchmark. In particular, we introduce a class of boundedly rational meta dynamic programming policies, BRMDP(D), that approximate full integration using a limited number of hyper-posterior draws, denoted by D. Trial-by-trial likelihood comparisons show that low-D boundedly rational meta-learning, especially BRMDP(1), fits participant behavior better than both no transfer and fully integrated Bayesian transfer. Consumers, therefore, transfer brand-level regularities across contexts, but through coarse representations of prior uncertainty. The findings imply that models of consumer learning should allow for approximate cross-context transfer, and that managerial counterfactuals based on either no-transfer or fully integrated learning can be misleading.

Cognitive science has recently begun exploring how people conceptualize and reason about truth. We offer the field a framework that can guide inquiry into intuitive theories of truth, centered on three core questions: how do people judge whether statements could be true, whether statements are true, and whether to assert them as true.

Past work seeks to align large language model (LLM)-based assistants with a target set of values, but such assistants are frequently forced to make tradeoffs *between* values when deployed. In response to the scarcity of value conflict in existing alignment datasets, we introduce ConflictScope, an automatic pipeline to evaluate how LLMs prioritize different values. Given a user-defined value set, ConflictScope automatically generates scenarios in which a language model faces a conflict between two values sampled from the set. It then prompts target models with an LLM-written ``user prompt'' and evaluates their free-text responses to elicit a ranking over values in the value set. Comparing results between multiple-choice and open-ended evaluations, we find that models shift away from supporting protective values, such as harmlessness, and toward supporting personal values, such as user autonomy, in more open-ended value conflict settings. However, including detailed value orderings in models' system prompts improves alignment with a target ranking by 14%, showing that system prompting can achieve moderate success at aligning LLM behavior under value conflict. Our work demonstrates the importance of evaluating value prioritization in models and provides a foundation for future work in this area.

Large language models solve complex problems yet fail on simpler variants, suggesting they achieve correct outputs through mechanisms fundamentally different from human reasoning. We synthesize cognitive science research into a taxonomy of 28 cognitive elements spanning computational constraints, meta-cognitive controls, knowledge representations, and transformation operations, then analyze their behavioral manifestations in reasoning traces. We propose a fine-grained cognitive evaluation framework and conduct the first large-scale analysis of 170K traces from 17 models across text, vision, and audio modalities, alongside 54 human think-aloud traces, which we make publicly available. Our analysis reveals systematic structural differences: humans employ hierarchical nesting and meta-cognitive monitoring while models rely on shallow forward chaining, with divergence most pronounced on ill-structured problems. Meta-analysis of 1,598 LLM reasoning papers reveals the research community concentrates on easily quantifiable behaviors (sequential organization: 55

We evaluate the moral alignment of large language models (LLMs) with human preferences in multilingual trolley problems. Building on the Moral Machine experiment, which captures over 40 million human judgments across 200+ countries, we develop a cross-lingual corpus of moral dilemma vignettes in over 100 languages called MultiTP. This dataset enables the assessment of LLMs' decision-making processes in diverse linguistic contexts. Our analysis explores the alignment of 19 different LLMs with human judgments, capturing preferences across six moral dimensions: species, gender, fitness, status, age, and the number of lives involved. By correlating these preferences with the demographic distribution of language speakers and examining the consistency of LLM responses to various prompt paraphrasings, our findings provide insights into cross-lingual and ethical biases of LLMs and their intersection. We discover significant variance in alignment across languages, challenging the assumption of uniform moral reasoning in AI systems and highlighting the importance of incorporating diverse perspectives in AI ethics. The results underscore the need for further research on the integration of multilingual dimensions in responsible AI research to ensure fair and equitable AI interactions worldwide.

Humans are remarkably efficient at decision making, even in "open-ended" problems where the set of possible actions is too large for exhaustive evaluation. Our success relies, in part, on processes for calling to mind the right candidate actions. When these processes fail, the result is a kind of puzzle in which the value of a solution would be obvious once it is considered, but never gets considered in the first place. Recently, machine learning (ML) architectures have attained or even exceeded human performance on open-ended decision making tasks such as playing chess and Go. We ask whether the broad architectural principles that underlie ML success in these domains generate similar consideration failures to those observed in humans. We demonstrate a case in which they do, illuminating how humans make open-ended decisions, how this relates to ML approaches to similar problems, and how both architectures lead to characteristic patterns of success and failure.

Theories of the evolution of cooperation through reciprocity explain how unrelated self-interested individuals can accomplish more together than they can on their own. The most prominent theories of reciprocity, such as tit-for-tat or win-stay-lose-shift, are inflexible automata that lack a theory of mind-the human ability to infer the hidden mental states in others' minds. Here, we develop a model of reciprocity with a theory of mind, the Bayesian Reciprocator. When making decisions, this model does not simply seek to maximize its own payoff. Instead, it also values the payoffs of others-but only to the extent it believes that those others are also cooperating in the same way. To compute its beliefs about others, the Bayesian Reciprocator uses a probabilistic and generative approach to infer the latent preferences, beliefs, and strategies of others through interaction and observation. We evaluate the Bayesian Reciprocator using a generator over games where every interaction is unique, as well as in classic environments such as the iterated prisoner's dilemma. The Bayesian Reciprocator enables the emergence of both direct-reciprocity when games are repeated and indirect-reciprocity when interactions are one-shot but observable to others. In an evolutionary competition, the Bayesian Reciprocator outcompetes existing automata strategies and sustains cooperation across a larger range of environments and noise settings than prior approaches. This work quantifies the advantage of a theory of mind for cooperation in an evolutionary game theoretic framework and suggests avenues for building artificially intelligent agents with more human-like learning mechanisms that can cooperate across many environments.

Large Language Models' (LLMs) programming capabilities enable their participation in \textit{open-source games}: a game-theoretic setting in which players submit computer programs in lieu of actions. These programs offer numerous advantages, including interpretability, inter-agent transparency, and formal verifiability; additionally, they enable \textit{program equilibria}, solutions that leverage the transparency of code and are inaccessible within normal-form settings. We evaluate the capabilities of leading open- and closed-weight LLMs to predict and classify program strategies and evaluate features of the approximate program equilibria reached by LLM agents in dyadic and evolutionary settings. We identify the emergence of payoff-maximizing, cooperative, and deceptive strategies, characterize the adaptation of mechanisms within these programs over repeated open-source games, and analyze their comparative evolutionary fitness. We find that open-source games serve as a viable environment to study and steer the emergence of cooperative strategy in multi-agent dilemmas.

People have a remarkable ability to infer the hidden causes of things. From physical evidence, such as muddy foot prints on the floor, we can figure out what happened and who did it. Here, we investigate another source of evidence: social evaluations. Social evaluations, such as praise or blame, are commonplace in everyday conversations. While such evaluations don't fully reveal what happened, they provide valuable clues. Across three experiments, we present situations where a person was praised or blamed, and participants' task is to use that information to figure out what happened. In Experiment 1, we find that people draw systematic inferences from social evaluations about situational factors, a person's actions, capabilities, and social roles. In Experiments 2 and 3 we develop computational models that generate praise and blame judgments by considering what causal role a person's action played, and what action they should have taken. Inverting these generative models of praise and blame via Bayesian inference yields accurate predictions about what inferences participants draw based on social evaluations.