Liane Lovitt
Liane Lovitt is a public policy manager at Anthropic and an industry representative at Yale's Governing (with) AI conference. She co-authored papers on training helpful and harmless assistants, on models that judge their own answers, and on Constitutional AI. In October 2023 she jointly led Collective Constitutional AI, a study whose public input process involved about 1,000 Americans in drafting a constitution for an AI system. In 2024 she co-wrote a blog post on measuring the persuasiveness of language models and co-authored Clio, a platform for studying real-world AI use.
Timeline
-
April 2022
Helpful and harmless assistants
She co-authors Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback in 2022, a paper that applies preference modeling and reinforcement learning from human feedback to finetune language models.
-
July 2022
Models that judge their own answers
She co-authors Language Models (Mostly) Know What They Know, a 2022 study of whether models can evaluate their own claims and predict which questions they will answer correctly.
-
December 2022
Constitutional AI for harmlessness
In 2022 she co-authors Constitutional AI: Harmlessness from AI Feedback, a method that trains a harmless assistant through self-improvement guided by a list of rules or principles.
-
October 2023
A constitution from public input
She jointly leads Collective Constitutional AI with Deep Ganguli, Saffron Huang and Divya Siddarth in 2023. Its public input process involved about 1,000 Americans in drafting a constitution for an AI system.
-
April 2024
Measuring model persuasiveness
Measuring the Persuasiveness of Language Models, an Anthropic blog post from 2024, is co-written by her and Esin Durmus.
-
December 2024
Clio, a usage-insights platform
In 2024 she co-authors Clio: Privacy-Preserving Insights into Real-World AI Use, a platform that uses AI assistants to analyze aggregated usage patterns across millions of conversations.
-
As of October 2026
Public policy at Anthropic
She is a public policy manager at Anthropic and an industry representative at Yale's Governing (with) AI conference.
Sources and credits
Photo: We have no published photo of Liane Lovitt yet. Send us a free-licensed photo.
- [2204.05862] Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback
- [2207.05221] Language Models (Mostly) Know What They Know
- [2212.08073] Constitutional AI: Harmlessness from AI Feedback
- Collective Constitutional AI: Aligning a language model with public input
- Measuring the persuasiveness of language models
- [2412.13678] Clio: Privacy-Preserving Insights into Real-World AI Use
- Participant List – ISPS Governing X Series