Joanne Jang
Product leader shaping model behavior and novel AI interfaces
San Francisco Bay Area, California
Summary
Product leader who turns frontier research into polished, widely used products: Joanne led product work on DALL·E (helping move DALL·E 2 from research to a beta product and shaping safeguards), and she has driven productized features across ChatGPT and the API (chat mode, voice, personalization/memory). Her background (Stanford CS/applied math, prior work on Google Assistant) supports a blend of technical product and research collaboration. newyorker+2
Founded and scaled OpenAI's Model Behavior discipline focused on personality, steerability, and safe interactions: Joanne helped codify model behavior principles (the Model Spec), worked with research and safety teams to reduce sycophancy and improve helpful, boundary-aware responses, and coordinated product/policy tradeoffs when shipping personality and personalization previews. theverge+2
Research-minded thinker about human–AI relationships and ethics: she publishes and speaks publicly about perceived vs. ontological consciousness, emotional impacts of AI, and how design choices (tone, memory, personalization) shape user well‑being—demonstrating a product+policy approach that emphasizes social science and user research alongside technical work. substack+2
Work
Education
Projects
Writing
Some thoughts on human-AI relationships
June 1, 2025A long-form essay exploring why people form emotional attachments to AI, distinctions between ontological and perceived consciousness, and how product/research choices (Model Spec) aim to design warmth without implying selfhood.
Introducing the Model Spec (LinkedIn post / public discussions)
May 1, 2024Public discussions and posts introducing and explaining the Model Spec framework for debating model behavior (objectives, rules, defaults) and describing policy/implementation tradeoffs.