About

Nathaniel Daw is the Huo Professor in Computational and Theoretical Neuroscience at the Department of Psychology within Princeton University, affiliated with the Princeton Neuroscience Institute. His lab studies how people and animals learn from trial and error, rewards, and punishments to make decisions, integrating computational, neural, and behavioral perspectives. The research focuses on understanding how subjects manage computationally demanding decision situations, such as choice under uncertainty or in tasks requiring sequential decisions like spatial navigation or strategic games such as chess. Daw's work draws on algorithms from machine learning to develop detailed, quantitative hypotheses about how the brain approaches these problems. Current projects include investigating how the brain controls its own decision-making processes, such as making higher-level decisions about when to deliberate or act, and exploring how these processes relate to issues of self-control and psychiatric disorders involving compulsion. His research aims to elucidate the neural mechanisms underlying decision-making and self-regulation, contributing to a deeper understanding of cognitive functions and mental health.

Research topics

• Computer Science
• Artificial Intelligence
• Biology
• Machine Learning
• Psychology
• Neuroscience
• Econometrics
• Cognitive science
• Cognitive psychology
• Economics

Selected publications

• Planning in the Brain: It's Not What You Think It Is

  Annual Review of Neuroscience · 2026-04-16

  articleSenior author

  The neuroscience of planning has long been analogized to search algorithms in artificial intelligence (AI), which simulate future actions to guide immediate choices. We argue that advances in both neuroscience and AI suggest that planning is better understood to encompass a broader class of computations where mental simulation supports learning, often well before a decision is needed. We review three neurocomputational mechanisms that illustrate this shift. First, hippocampal replay resembles search but also often occurs prospectively or offline, likely training downstream circuits rather than directly guiding choice. Second, temporally abstract representations, such as grid cells, can enable planning without iterative search. Third, metalearning may shape how prefrontal dynamics implement task-specific planning strategies, echoing how AI systems learn to adapt across contexts. This view recasts the brain's planning machinery as a family of learning processes that leverage simulations to build representations and strategies, with forward search as one special case.

  PublisherDOI

• AI-Discovered Cognitive Models Reveal Novel Insights into Human and Animal Learning

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-05-21

  articleOpen access

  Scientific models are widely used across the natural sciences as an interface between scientific theories and empirical data. Such models play a key role, for example, in the study of human and animal learning, where they express algorithmic hypotheses and relate them to psychology and neuroscience data. These models are traditionally handcrafted by expert researchers based on existing theory or new insights. Such handcrafted models, however, are now known to fall short of capturing the full richness of behavior, even in their narrow domains. An alternative data-driven approach has emerged, seeking to discover new insights by fitting and interpreting flexible models. However, these tools require substantial human effort to derive insight from data, and it has been unclear how to discover new ideas from data efficiently. Here, we present DataDIVER, a general approach for automatically discovering computational models from data, and demonstrate that these models surface novel mechanistic insights into human and animal learning. Our approach delivers models that take the form of short computer programs, which are optimized both to fit data well and to be simple. These programs explicitly connect with existing theoretical frameworks and are readily understandable by human scientists. They can also be used to make novel predictions, some of which we show are borne out in re-analysis of existing data. General-purpose tools for surfacing new ideas from data, especially in combination with the large datasets that are increasingly available in many fields, stand to dramatically accelerate scientific discovery.

  PublisherDOI

• Children leverage predictive representations for flexible, value-guided choice

  2025-10-07

  preprintOpen access

  By building a mental model of how the world works and using it to forecast the outcomes of different actions, a learner can make flexible choices in changing environments. However, while children and adolescents readily acquire structured knowledge about their environments, relative to adults, they tend to demonstrate weaker signatures of leveraging this knowledge to plan actions. One explanation for these developmental differences is that using a mental model to prospectively simulate potential choices and their outcomes is computationally costly, taxing cognitive control and working memory mechanisms that continue to develop into adulthood. Here, we ask whether children might effectively leverage structured knowledge to make flexible choices by relying on two alternative strategies that do not require costly mental simulation at choice time. First, through offline replanning, models can be queried before the time of choice to generate possible scenarios and update the values of potential actions. Second, an abstracted predictive model, known as a Successor Representation, can be built and harnessed to enable simplified computation of long-run reward values of candidate actions, without requiring iterative simulation of multiple time steps. To assess whether children, adolescents, and adults aged 7 - 23 years similarly harness these learning strategies, we ran three experiments. In Experiments 1 and 2, we used a reward revaluation task in which we manipulated the opportunity for offline replanning during rest, and found that children flexibly updated their behavior by leveraging structured knowledge in an adult-like manner. Surprisingly, across age, rest did not mediate flexible replanning, raising the possibility that participants may have behaved adaptively by harnessing predictive representations online. In Experiment 3, we directly tested whether children use predictive representations. Here, we observed early-emerging use of the SR, providing a mechanistic account of how children use structured knowledge to guide choice without detailed model-based simulation.

  PublisherOA PDFDOI

• Improving the reliability of the Pavlovian go/no-go task for computational psychiatry research

  2025-11-06

  preprintOpen access

  Background: The Pavlovian go/no-go task is commonly used to measure individual differences in Pavlovian biases and their interaction with instrumental learning. However, prior research has found suboptimal reliability for computational model-based performance measures for this task, limiting its usefulness in individual-differences research. These studies did not make use of several strategies previously shown to enhance task-measure reliability (e.g., task gamification, hierarchical Bayesian modeling for model estimation). Here we investigated if such approaches could improve the task’s reliability. Methods: Across two experiments, we recruited two independent samples of adult participants (N=103, N=110) to complete a novel, gamified version of the Pavlovian go/no-go task multiple times over several weeks. We used hierarchical Bayesian modeling to derive reinforcement learning model-based indices of participants' task performance, and additionally to estimate the reliability of these measures. Results: In Experiment 1, we observed considerable and unexpected practice effects, with most participants reaching near-ceiling levels of performance with repeat testing. Consequently, the test-retest reliability of some model parameters was unacceptable (range: 0.379–0.973). In Experiment 2, participants completed a modified version of the task designed to lessen these practice effects. We observed greatly reduced practice effects and improved estimates of the test-retest reliability (range: 0.696–0.989). Conclusion: The results demonstrate that model-based measures of performance on the Pavlovian go/no-go task can reach levels of reliability sufficient for use in individual- differences research. However, additional investigation is necessary to validate the modified version of the task in other populations and settings.

  PublisherDOI

• Developmental change in structure learning reflects a shift from recency-based to relational prediction

  2025-03-06

  preprintOpen access

  Children are adept statistical learners, capable of parsing streams of structured input into meaningful units, but the cognitive processes they engage during learning may differ from those of adults. To date, however, it is unclear how learners of different ages predict upcoming experience when navigating environments with complex structure, as well as how changes in predictive learning mechanisms influence structured knowledge acquisition. To address this question, we tested 106 children, adolescents, and adults, ages 8 - 22 years, on a predictive learning task, in which they experienced sequences of stimuli with a higher-order temporal structure. After an initial learning phase, participants’ explicit knowledge of the relations between stimuli was probed via two additional task measures. We used a recently introduced computational model to characterize participants’ response times during learning, and found that all participants relied on simple, recency-based prediction, anticipating that they would encounter stimuli they recently encountered in the past. With increasing age, however, participants demonstrated greater evidence of additionally relying on a more sophisticated learning mechanism, which captured a predictive representation of the conditional relations between stimuli. Though predictive learning changed with age, we found only weak evidence that these changes related to the acquisition of explicit knowledge of the environment. Our results suggest that the learning mechanisms through which people parse continuous streams of experience change with age, influencing their predictions about upcoming events.

  PublisherOA PDFDOI

• Building compositional tasks with shared neural subspaces

  Nature · 2025-11-26 · 5 citations

  articleOpen access

  Abstract Cognition is highly flexible—we perform many different tasks 1 and continually adapt our behaviour to changing demands 2,3 . Artificial neural networks trained to perform multiple tasks will reuse representations 4 and computational components 5 across tasks. By composing tasks from these subcomponents, an agent can flexibly switch between tasks and rapidly learn new tasks 6,7 . Yet, whether such compositionality is found in the brain is unclear. Here we show the same subspaces of neural activity represent task-relevant information across multiple tasks, with each task flexibly engaging these subspaces in a task-specific manner. We trained monkeys to switch between three compositionally related tasks. In neural recordings, we found that task-relevant information about stimulus features and motor actions were represented in subspaces of neural activity that were shared across tasks. When monkeys performed a task, neural representations in the relevant shared sensory subspace were transformed to the relevant shared motor subspace. Monkeys adapted to changes in the task by iteratively updating their internal belief about the current task and then, based on this belief, flexibly engaging the shared sensory and motor subspaces relevant to the task. In summary, our findings suggest that the brain can flexibly perform multiple tasks by compositionally combining task-relevant neural representations.

  PublisherDOI

• Leveraging large language models to estimate clinically relevant psychological constructs in psychotherapy transcripts

  medRxiv · 2025-03-05 · 1 citations

  preprintOpen accessSenior author

  Abstract Developing precise, innocuous markers of psychopathology and the processes that foster effective treatment would greatly advance the field’s ability to detect and intervene on psychopathology. However, a central challenge in this area is that both assessment and treatment are conducted primarily in natural language, a medium that makes quantitative measurement difficult. Although recent advances have been made, much existing research in this area has been limited by reliance on previous-generation psycholinguistic tools. Here we build on previous work that identified a linguistic measure of “psychological distancing” (that is, viewing a negative situation as separated from oneself) in client language, which was associated with improved emotion regulation in laboratory settings and treatment progress in real-world therapeutic transcripts (Nook et al., 2017, Nook et al., 2022). However, this formulation was based on context-insensitive word count-based measures of distancing (pronoun person and verb tense), which limits the ability to detect more abstract expressions of psychological distance, such as counterfactual or conditional statements. This approach also leaves open many questions about how therapists’ — likely subtler — language can effectively guide clients toward increased psychological distance. We address these gaps by introducing the use of appropriately prompted large language models (LLMs) to measure linguistic distance, and we compare these results to those obtained using traditional word-counting techniques. Our results show that LLMs offer a more nuanced and context-sensitive approach to assessing language, significantly enhancing our ability to model the relations between linguistic distance and symptoms. Moreover, this approach enables us to expand the scope of analysis beyond client language to shed insight into how therapists’ language relates to client outcomes. Specifically, the LLM was able to detect ways in which a therapist’s language encouraged a client to adopt distanced perspectives—rather than simply detecting the therapist themselves being distanced. This measure also reliably tracked the severity of patient symptoms, highlighting the potential of LLM-powered linguistic analysis to deepen our understanding of therapeutic processes.

  PublisherOA PDFDOI

• Between planning and map building: Prioritizing replay when future goals are uncertain

  Neuron · 2025-10-27 · 2 citations

  articleSenior author
  PublisherDOI

• Humans rationally balance detailed and temporally abstract world models

  Communications Psychology · 2025-01-04 · 5 citations

  articleOpen accessSenior author

  How do people model the world's dynamics to guide mental simulation and evaluate choices? One prominent approach, the Successor Representation (SR), takes advantage of temporal abstraction of future states: by aggregating trajectory predictions over multiple timesteps, the brain can avoid the costs of iterative, multi-step mental simulation. Human behavior broadly shows signatures of such temporal abstraction, but finer-grained characterization of individuals' strategies and their dynamic adjustment remains an open question. We developed a task to measure SR usage during dynamic, trial-by-trial learning. Using this approach, we find that participants exhibit a mix of SR and model-based learning strategies that varies across individuals. Further, by dynamically manipulating the task contingencies within-subject to favor or disfavor temporal abstraction, we observe evidence of resource-rational reliance on the SR, which decreases when future states are less predictable. Our work adds to a growing body of research showing that the brain arbitrates between approximate decision strategies. The current study extends these ideas from simple habits into usage of more sophisticated approximate predictive models, and demonstrates that individuals dynamically adapt these in response to the predictability of their environment.

  PublisherOA PDFDOI

• A Rational Information Gathering Account of Infant Habituation

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-01-09

  preprintOpen accessSenior author

  Abstract Gaze is one of the primary experimental measures for studying cognitive development, especially in preverbal infants. However, the field is only beginning to develop a principled explanatory framework for making sense of the various factors affecting gaze. We approach this issue by addressing infant gaze from first principles, using rational information gathering. In particular, we revisit the influential descriptive account of Hunter and Ames (1988) (H&A), which posits a set of regularities argued to govern how gaze preference for a stimulus changes with experience and other factors. When the H&A’s model is reconsidered from the perspective of rational information gathering (as recently also proposed by other authors), one feature of the model emerges as surprising: that preference for a stimulus is not monotonic with exposure. This claim, which has at least some empirical support, is in contrast to most statistical measures of informativeness, which strictly decline with experience. We present a normative, computational theory of visual exploration that rationalizes this and other features of the classic account. Our account suggests that H&A’s signature nonmonotonic pattern is a direct manifestation of a ubiquitous principle of the value of information in sequential tasks, other consequences of which have recently been observed in a range of settings including deliberation, exploration, curiosity, and boredom. This is that the value of information gathering, putatively driving gaze, depends on the interplay of a stimulus’ informativeness (called Gain , the focus of other rationally motivated accounts) with a second factor (called Need ) reflecting the estimated chance that information will be used in the future. This computational decomposition draws new connections between infant gaze and other cases of exploration, and offers novel, quantitative interpretations and predictions about the factors that may impact infant exploratory attention.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R01MH087882

  NIH · $1.8M · 2015

• CRCNS: Computational and neural mechanisms of memory-guided decisions

  NIH · $1.7M · 2014–2020

• CRCNS Research Proposal: Collaborative Research: Prioritization of memory reactivation for decision-making

  NSF · $391k · 2018–2022

• CRCNS: Representational foundations of adaptive behavior in natural and artificial

  NIH · $384k · 2015–2018

• CRCNS: Representational foundations of adaptive behavior in natural and artificial

  NIH · $799k · 2015–2019

Frequent coauthors

• Peter Dayan

  Max Planck Institute for Biological Cybernetics

  45 shared

• Daphna Shohamy

  Columbia University

  31 shared

• David S. Touretzky

  Carnegie Mellon University

  30 shared

• Yael Niv

  Princeton University

  29 shared

• Stephen M. Fleming

  University College London

  28 shared

• Ben Seymour

  University of Oxford

  26 shared

• Raymond J. Dolan

  National Hospital for Neurology and Neurosurgery

  22 shared

• Marcelo G. Mattar

  New York University

  22 shared