Stanford Institute for Human-Centered Artificial Intelligence (HAI)

What it is

Stanford HAI, the Institute for Human-Centered Artificial Intelligence, is a research institute at Stanford University devoted to artificial intelligence and its effects on people. It was launched on 18 March 2019 and is co-directed by John Etchemendy, a professor of philosophy and former Stanford provost, and Fei-Fei Li, a professor of computer science and former director of the Stanford AI Lab^[1]. Its stated mission is to advance AI research, education, policy and practice to improve the human condition^[1]. The institute's about page puts the same aim more briefly: to advance humanity through AI and data science^[5].

The word "human-centered" in the name signals the framing the institute works from. Questions about AI are treated not only as engineering problems but as questions about how the technology affects people, institutions and public policy. Much of what HAI produces is written for readers outside engineering: policy material, public reports and evaluations intended for people who will never train a model themselves. The leadership pairing reflects the same combination of concerns, with a philosopher and former university administrator sharing the directorship with a computer scientist^[1].

Where it came from

HAI was announced in March 2019 as a Stanford institute rather than a single departmental lab, and Fei-Fei Li is a founding co-director^[4]. Her later recognition includes the Queen Elizabeth Prize for Engineering, received in 2025^[5].

The institute is deliberately cross-disciplinary: it draws faculty and students from all seven of Stanford's schools, namely business, education, engineering, humanities and sciences, law, medicine, and sustainability^[5]. That spread matters for how the work is done. An economist, a lawyer, a physician and a computer scientist can sit inside the same programme, so a question such as how a model should be evaluated in a clinical setting is not left to engineers alone.

How the work is organised

HAI groups its activity into three pillars: accelerating discovery, transforming education, and guiding and shaping AI for society^[5]. The first covers research, including work on the models themselves; the second covers teaching; the third covers the policy and public-facing side, where the institute tries to inform how AI is governed and used.

The Center for Research on Foundation Models

On 18 August 2021, HAI announced the Center for Research on Foundation Models (CRFM), directed by Percy Liang and bringing together more than 175 researchers across more than 10 Stanford departments^[2]. The centre's 200-page report, "On the Opportunities and Risks of Foundation Models", was written by more than 100 Stanford scholars and described models trained on broad data at scale and adaptable to many downstream tasks, citing BERT, GPT-3, CLIP and Codex as examples^[2].

The idea behind the term is straightforward. Instead of building a separate system for every job, one model is trained once on a very large body of data and then adapted to many narrower jobs, called downstream tasks. The model serves as a base that other work is built on, which is where the word "foundation" comes from. Because the same base can end up in many different products, the report's subject was not only capability but also the risks that come with wide reuse^[2].

The Holistic Evaluation of Language Models (HELM) project came out of CRFM in 2022^[5]. Its name points at the ambition: to look at language models across many dimensions of behaviour rather than reduce them to a single score, so that users can see where a model is strong and where it fails.

The AI Index

HAI publishes the AI Index, an independent programme that gathers data on artificial intelligence for policymakers, researchers, journalists, executives and the public^[3]. Its report archive begins with the 2017 edition, and the latest is the 2026 AI Index Report^[3]. The point of such a programme is measurement: rather than relying on impressions about how fast the field is moving, readers get collected figures and trends they can check for themselves, which is useful to a journalist on deadline and to a government official drafting rules alike^[3].

Where it sits

Taken together, the institute combines three strands. Through CRFM and HELM it does research on the models that now drive the field and on how to evaluate them; through the AI Index it produces public measurement; and through its third pillar it works on policy and on how AI is used in society^[5]. Its cross-school structure and its pairing of technical and humanistic leadership place it among the places where AI is discussed as a social question as much as a technical one^[1][5].