Nathaniel Daw

Timeline

1. 2003

   PhD at Carnegie Mellon

   In August 2003, Daw completes a Ph.D. in computer science at Carnegie Mellon, advised by David Touretzky. The thesis, “Reinforcement Learning Models of the Dopamine System and their Behavioral Implications”, extends temporal-difference theories of dopamine neurons and explains dopamine responses that had previously seemed anomalous.

2. 2005–2006

   Gatsby Unit decision papers

   Postdoctoral research at UCL’s Gatsby Computational Neuroscience Unit yields a 2005 Nature Neuroscience paper with Yael Niv and Peter Dayan, which proposes arbitrating between prefrontal and dorsolateral striatal decision systems according to uncertainty. A 2006 Nature paper with John P. O’Doherty, Peter Dayan, Ben Seymour and Raymond Dolan finds the frontopolar cortex preferentially active during exploratory decisions in humans.

3. 2011–2012

   Model-based choice at NYU

   At New York University’s Center for Neural Science and Department of Psychology, Daw coauthors the 2011 Neuron paper “Model-based influences on humans’ choices and striatal prediction errors” with Samuel Gershman, Ben Seymour, Peter Dayan and Raymond Dolan. Ventral striatal signals in the study reflect both model-free and model-based predictions, in proportions matching those that best explained choice behavior. In 2012, the Society for NeuroEconomics names Daw an Early Career Award winner.

4. 2022

   A NeurIPS Outstanding Paper

   Daw is one of ten authors, including Sreejan Kumar and Thomas L. Griffiths, of “Using natural language and program abstractions to instill human inductive biases in machines”, which wins an Outstanding Paper Award at NeurIPS 2022. Co-training meta-learning agents on predictions of natural-language task descriptions and induced programs guides them toward more human-like inductive biases.

5. 2024

   Natural and Artificial Minds launch

   In 2024, Princeton launches Natural and Artificial Minds, a research initiative within the Princeton Laboratory for Artificial Intelligence that brings cognitive science and AI researchers together. At the launch event, Daw presents work on how modern AI can help explain how the human brain solves difficult tasks, particularly decision-making under uncertainty.

6. March 2026

   Confidence signals in language models

   In March 2026, an arXiv paper by Dharshan Kumaran, Daw, Simon Osindero, Petar Veličković and Viorica Patraucean asks whether language models use their own confidence to decide whether to answer or abstain. Activation steering of the models’ internal confidence signals shifts abstention rates, and Nature Machine Intelligence publishes the paper in 2026. Daw’s affiliations on the paper are Google DeepMind and Princeton University.

7. April 2026

   Automated discovery of cognitive models

   In April 2026, Daw gives the talk “Automated discovery of interpretable cognitive models” in the Siegel Family Quest for Intelligence seminar series at MIT. It presents two projects that use techniques from modern AI, hybrid neuro-symbolic networks and LLM-driven program synthesis, to discover new models of learning automatically while preserving interpretability.

8. As of October 2026

   Huo Professor at Princeton

   As Huo Professor in Theoretical and Computational Neuroscience, Daw holds a post in the Princeton Neuroscience Institute and the Department of Psychology at Princeton University, after serving on the faculty at New York University. The Daw Lab, with Daw as principal investigator, studies computational approaches to reinforcement learning and decision making, using computational models in the design of experiments and the analysis of behavioral and neural data.