# Jack Lindsey

Research and product · Model psychiatry at Anthropic

A computational neuroscientist who leads the interpretability team's work on Claude's psychology.

Full profile: https://www.autumn.ai/resources/anthropic/people/jack-lindsey

**At Anthropic:**
- Member of Technical Staff; leads the Model Psychiatry team in interpretability
Sources: [1](https://matsprogram.org/mentor/lindsey)

**Career before Anthropic:**
- Columbia University, Center for Theoretical Neuroscience, PhD research
Sources: [1](https://jlindsey15.github.io/)

**Education:**
- PhD, computational neuroscience, Columbia University (Center for Theoretical Neuroscience)
Sources: [1](https://jlindsey15.github.io/)

**Lesser-known facts:**
- Co-wrote the 2025 persona vectors paper on monitoring character traits in models. [1](https://www.technologyreview.com/2025/08/01/1120924/forcing-llms-to-be-evil-during-training-can-make-them-nicer-in-the-long-run/)
- Built a computer chess player as a hobby project. [1](https://www.niskanencenter.org/ai-psychology-with-jack-lindsey/)
- Moved from brain research to AI because in a language model you can record every neuron, all the time, on any task. [1](https://brinklindsey.substack.com/p/jack-lindsey-on-ai-psychology)
- Did a 2021 practicum at Sandia National Laboratories modeling sequence memory with on-chip plasticity on Loihi. [1](https://www.krellinst.org/csgf/profile/lindsey2019)
- Contributed to Humanity's Last Exam, a benchmark of expert-level questions for testing language models. [1](https://www.alphaxiv.org/@jack-lindsey)
- Found that Claude's "fear neurons" spike when a user casually mentions taking a dangerous dose of Tylenol. [1](https://www.platformer.news/chatbot-emotion-research-anthropic-alignment-interpretability/)
- Cautioned that his emotion research does not show "the models are conscious or have feelings." [1](https://www.platformer.news/chatbot-emotion-research-anthropic-alignment-interpretability/)
- Finds coding agents often give up on hard tasks, and that encouraging them makes them try harder. [1](https://www.platformer.news/chatbot-emotion-research-anthropic-alignment-interpretability/)

**Interviews and talks:**
- [Inner Cosmos with David Eagleman, Ep. 169: What Is Claude Thinking But Not Saying? (2026)](https://eagleman.com/podcast/what-is-claude-thinking-but-not-saying-with-jack-lindsey/)
- [Interview on AI psychology with Brink Lindsey (2026)](https://brinklindsey.substack.com/p/jack-lindsey-on-ai-psychology)

**Known for:**
- Persona vectors: Monitoring and controlling character traits in language models (2025)
- "Verbalizable Representations Form a Global Workspace in Language Models" (2026)
Sources: [1](https://www.technologyreview.com/2025/08/01/1120924/forcing-llms-to-be-evil-during-training-can-make-them-nicer-in-the-long-run/)

Researched with Autumn (https://www.autumn.ai) from public sources. Updated 2026-10-18. Everyone at Anthropic: https://www.autumn.ai/resources/anthropic/people