Natasha Jaques

Natasha Jaques is a reinforcement learning researcher who studies how AI agents learn from people and from each other. She completed a PhD at the MIT Media Lab in 2019, where she developed techniques for fine-tuning language models with reinforcement learning and human feedback, and OpenAI's series of work on RLHF built on them. Her 2018 paper on social influence rewarded agents for having causal influence over other agents' actions, and her 2020 PAIRED paper with Sergey Levine and others proposed Unsupervised Environment Design. After a visiting postdoc at UC Berkeley and a senior research role at Google Brain, she started the Social RL Lab at the University of Washington in 2024, and in 2025 she became an AI2050 Early Career Fellow. She is an Assistant Professor at the University of Washington's Allen School and a Senior Research Scientist at Google DeepMind.

Now at: University of Washington and Google DeepMind
Badges: Bridge

Timeline

2012–2019

Degrees from Regina to MIT

She earns a BSc in Computer Science and a BA in Psychology at the University of Regina in 2012, then an MSc in Computer Science at the University of British Columbia in 2014. In 2019, she completes a PhD at the MIT Media Lab, where she belongs to the Affective Computing group.
2018–2019

Social influence as a reward

In 2018, she is first author of "Social Influence as Intrinsic Motivation for Multi-Agent Deep Reinforcement Learning", written with Angeliki Lazaridou, Joel Z. Leibo, Nando de Freitas and others. The paper rewards agents for having causal influence over other agents' actions and reports better coordination and communication in social dilemma environments. In 2019, she receives a Best Paper Honourable Mention at ICML.
2019

Offline learning from human preferences

In 2019, she is first author of "Way Off-Policy Batch Deep Reinforcement Learning of Implicit Human Preferences in Dialog", written with Asma Ghandeharioun, Rosalind Picard and others. The algorithms learn offline from a fixed batch of human interaction data and use KL-control to penalize divergence from a pre-trained prior. OpenAI's series of work on Reinforcement Learning from Human Feedback builds on the techniques she develops during her MIT PhD for fine-tuning language models with RL and learning from human feedback.
2020

PAIRED environment design

In 2020, she coauthors "Emergent Complexity and Zero-shot Transfer via Unsupervised Environment Design" with Michael Dennis, Eugene Vinitsky, Alexandre Bayen, Stuart Russell, Andrew Critch and Sergey Levine. The paper proposes Unsupervised Environment Design and introduces PAIRED, in which an adversary generates environments that maximize the regret between a protagonist agent and an antagonist agent.
2021

An AAAC dissertation award

She wins the Association for the Advancement of Affective Computing's Outstanding PhD Dissertation Award in 2021 for her thesis "Social and Affective Machine Learning". The award is presented at the International Conference on Affective Computing and Intelligent Interaction (ACII).
January 2024

Founding the Social RL Lab

She starts the Social RL Lab at the University of Washington's Paul G. Allen School of Computer Science & Engineering in January 2024. She arrives from Google Brain, where she is a senior research scientist, after time as a Visiting Postdoctoral Scholar at UC Berkeley, in Sergey Levine's group. At Google Brain she builds methods for adversarial environment generation to improve the robustness of RL agents.
2025

AI2050 Early Career Fellow

She is named a 2025 AI2050 Early Career Fellow, with a project on training language models that evolve and co-adapt to each other to simulate how people adapt to interacting with AI. Two papers from her lab win workshop awards in 2025: "Modeling Others' Minds as Code" receives a Best Paper award at the NeurIPS 2025 LAW Workshop, and "Chasing Moving Targets with Online Self-Play Reinforcement Learning for Safer Language Models" wins the Outstanding Paper Award at the COLM 2025 AIA workshop.
As of October 2026

Allen School and DeepMind

She is an Assistant Professor at the University of Washington's Paul G. Allen School of Computer Science & Engineering, where she leads the Social RL Lab, and a Senior Research Scientist at Google DeepMind. The lab, founded in 2024, studies reinforcement learning fine-tuning of LLMs, multi-agent reinforcement learning, learning from human feedback and human-AI interaction.