Why we’re needed
For most of the last century, the person accountable for a hard call also assembled the evidence for it. An operator read the panels. A clinician examined the patient. Judgment was the whole job.
Now the inputs talk back — the system flags what matters, surfaces the likely scenario, recommends when to trust it. Someone still has to decide, and someone is still accountable for what happens next.
The risks have not changed. The decision-making process has.
Meanwhile, the cost of saying “no” to AI isn’t just damaging across industries, it’s existential. Nobody is measuring what that shift does to a person’s judgment. Deployment decisions are being made now, at scale, in settings where the consequences compound for decades — and whether AI strengthens or erodes the judgment of the people using it is being answered by default rather than by evidence.
We are not here to slow AI down. We are here to change what gets measured — so that human judgment, expertise, and agency are treated as outcomes of AI deployment, not assumptions behind it.
Assessment is not a brake on this technology. It is how trust becomes warranted rather than assumed, and how agency survives the systems built to assist it.
What we do
We build the measurements that show what an AI system does to the person using it — and we put them in the hands of the people who build, buy, and deploy AI.
Our founding instrument, the HAAT Index, asks a question no existing framework asks: does this system leave the people using it better at being human? It assesses three things.
-
01
Trust calibration
Do people rely on the system appropriately, or over- or under-trust it?
-
02
Decision quality
Does it genuinely improve judgment, critical thinking, and sensemaking?
-
03
Fit for purpose
Does it improve the wellbeing of real operators under real pressure?
We are starting in energy and safety-critical operations, where the consequences are most immediate and the measurements can be validated against real conditions — then carrying the same architecture to the next domain, and the next.
We publish what we find, openly — methodology, results, and limits — whether or not it flatters the system under assessment. It goes out in the format the decision actually takes: a procurement specification, a deployment checklist, a benchmark, not only a paper.
We publish our findings, not files on research subjects: work inside a specific organization is de-identified or aggregated, and any attribution is agreed to in writing up front. Whether we publish is never negotiable.
First releases in progress: the HAAT Index methodology, an initial assessment, and notes on what we learn building it.
How we do it
The expertise to judge AI’s effect on people already exists — scattered across practitioners, researchers, and operators, most of it outside the institutions that usually get asked. HAATlab is the machinery built to reach it.
-
01
Open challenges
Questions put to anyone who can answer them — the first will stress-test the Index against thousands of practitioners.
-
02
Expert circles
The work is held to a documented, auditable process: credibility through process, not pedigree.
-
03
A lean core
No more than five full-time staff, which keeps the work fast and cheap enough to repeat in every next domain.
Our independence is structural. We do not lobby. We do not assess systems in which we hold a financial stake. We never sit inside the organization whose tools we are assessing. And we do not accept funding conditioned on a finding — what is paid for is a documented review process, never an outcome. We publish everything: methodology, results, and limits.
Founding support does not buy one framework for one industry. It builds the capacity to do this again — faster, more cheaply, and more credibly each time.
Who we are
Built by practitioners in cross-sector collaboration, bridging public policy, tech, academia, and civil society. Experts in AI governance, policy, evaluation, and organizational design.
-
Max Scott
Board Co-Chair, AI & Methodology
Globally recognized leader on responsible AI and emerging technology governance. At Microsoft’s Office of Responsible AI, led teams developing and deploying governance frameworks for sensitive AI technologies; co-chaired UNESCO Business Council for the Ethics of AI; former U.S. State Department. Member, Stimson Center’s Alfred Lee Loomis Innovation Council. MSc, International Relations, London School of Economics.
-
Elliot Fleming
Board Co-Chair, Development & Operations
Formerly Chief of Staff, The Brookings Institution, and Operations at Microsoft’s Office of Responsible AI. Worked for both Democrats and Republicans on Capitol Hill. Led domestic and international campaigns focusing on organizational design and coalition building. MPA, George Washington University; Areas of Focus, Nonprofit Management, National Security and Foreign Policy.
-
Corey Broschak
Board Vice-Chair, Strategy and Vision
Brookings Institution Director of Strategic Initiatives, Senior Advisor to the President. Former Deputy Director and global resilience policy advisor, U.S. Department of Defense. Works where institutional decisions are made under real consequence. MA, International Affairs, American University.
-
Tori Westerhoff
Board Treasurer, AI & Cognitive Science
Leads Microsoft’s AI Red Team Ops, overseeing pre-launch security and safety testing of high-risk AI systems, and leads frontier AI harm evaluation implementation and emerging risk research. Previously Deloitte, consulting on national security strategy. Neuroscience, Yale; MBA, Wharton. 2026 Top 100 Women in AI.
We are forming an advisory council of leaders committed to HAATlab’s long-term success. Put a name forward →
Four ways to take part.
-
Fund the launch
Founding gifts cover the operating footprint while the first assessments are built. Email the Executive Director directly.
-
Join the advisory council
A small working group rather than an honorary board. Put a name forward — your own or someone else’s.
-
Partner on a pilot
Bring us a setting where people are accountable for decisions an AI system now shapes.
-
Join the expert network
Practitioners, researchers, and operators who want to be asked when a question falls in their field.
Fund the launch
Founding support builds the first assessments.
HAATlab is funded by philanthropic grants and founding donors. That funding validates the methodology and covers the initial operating footprint while the first assessments are built. We do not accept funding conditioned on a finding, and we publish results regardless of what they show.
Conversations about founding support go directly to the Executive Director. There is no form for this one.
Join the advisory council
A small working group, not an honorary board.
The council is forming now. We are looking for people who have made consequential calls under real conditions and who want to shape how this work develops. Put a name forward — your own or someone else’s.
Partner on a pilot
Bring us a setting where the decision matters.
A pilot means applying the HAAT Index in a real setting where people are accountable for decisions an AI system now shapes. We work from scenario-based decision tasks, system logs, interviews, and workflow measures, gathered with the people doing the work rather than about them. We are the external validator: we do not sit inside the organization whose tools are under assessment, and we publish the methodology, the results, and the limits.
If you have a setting in mind, write to the Executive Director with a sentence or two about the decisions being made and the systems involved.
Join the expert network
The expertise already exists. We are building the way to reach it.
Practitioners, researchers, and operators who want to be asked when a question falls in their field. Joining puts you on the list for open challenges and expert circles. It is not a commitment to any particular piece of work.