PeopleofSuperintelligence

Researcher

Esin Durmus

As of October 2026, Esin Durmus works at Anthropic, where her research covers how language models express opinions and how persuasive they are. In 2019 she was a fifth-year PhD candidate at Cornell, and in 2020 she co-authored FEQA, a question-answering metric for faithfulness in summarization. The OpinionsQA dataset, introduced in 2023, compared model opinions with those of 60 US demographic groups. In October 2023 she ran the OpinionQA evaluation for Anthropic's Collective Constitutional AI study, and in 2024 she led a study of model persuasiveness and co-authored Scaling Monosemanticity. In 2025 she co-authored a Nature Human Behaviour paper on AI and democracy.

Now at
Anthropic
Works
64
Citations
4,347
h-index
20

Publication numbers from OpenAlex, as of 15 October 2026.

Timeline

  1. By 2019

    A Cornell PhD candidate

    In 2019 she is a fifth-year PhD candidate in Cornell's Natural Language Processing group, advised by Claire Cardie.

  2. 2020

    FEQA for summary faithfulness

    She co-authors FEQA, a question-answering metric for faithfulness in abstractive summarization, which appears at ACL 2020.

  3. March 2023

    The OpinionsQA dataset

    She co-authors Whose Opinions Do Language Models Reflect?, which creates the OpinionsQA dataset in 2023 to compare model opinions with those of 60 US demographic groups.

  4. October 2023

    Collective Constitutional AI

    She runs the OpinionQA evaluation and helps frame and design the experiments for Anthropic's Collective Constitutional AI study, published in 2023.

  5. April 2024

    Measuring model persuasiveness

    She leads the research on how persuasive language models are, designing and running the experiments for an Anthropic study published in 2024.

  6. May 2024

    Scaling Monosemanticity

    She co-authors Scaling Monosemanticity in 2024, a paper that extracts interpretable features from Claude 3 Sonnet.

  7. October 2025

    AI's impact on democracy

    She co-authors The Impact of Advanced AI Systems on Democracy, a Nature Human Behaviour paper published in 2025.

  8. As of October 2026

    Working at Anthropic

    She works at Anthropic, and a Nature Human Behaviour paper published in 2025 lists her with that affiliation.

Sources and credits

Photo: We have no published photo of Esin Durmus yet. Send us a free-licensed photo.

  1. Esin Durmus – Rising Stars 2019 publish.illinois.edu. Checked 15 October 2026.
  2. FEQA: A Question Answering Evaluation Framework for Faithfulness Assessment in Abstractive Summarization - ACL Anthology aclanthology.org. Checked 15 October 2026.
  3. [2303.17548] Whose Opinions Do Language Models Reflect? arxiv.org. Checked 15 October 2026.
  4. Collective Constitutional AI \ Anthropic anthropic.com. Checked 15 October 2026.
  5. Measuring the persuasiveness of language models \ Anthropic anthropic.com. Checked 15 October 2026.
  6. Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet transformer-circuits.pub. Checked 15 October 2026.
  7. The Impact of Advanced AI Systems on Democracy | Harvard Kennedy School hks.harvard.edu. Checked 15 October 2026.
  8. The impact of advanced AI systems on democracy | Nature Human Behaviour nature.com. Checked 15 October 2026.