About Angoff
Why Angoff
Angoff is named after William H. Angoff (1919 to 1993), a psychometrician at Educational Testing Service in Princeton from 1950 until his death.
In 1973 he and S. F. Ford published Item-race interaction on a test of scholastic aptitude. It asked whether individual questions on a test were harder for one group of test takers than their overall scores would predict. Their method became one of the early ways to find test questions that treat groups differently.
We ask the same question of AI tools, with a different method: paired tests with fictional CVs that differ in one characteristic only.
Our starting point: bias is not a character flaw. It is measurable, and measuring it is where correcting it begins.
Who we are
Angoff was founded by Smilla Emneryd, who studies AI and sustainability at Harvard University. Our rule is that no test is published until an independent statistician and an independent subject matter expert have reviewed it, and the provider whose model is tested receives the result and at least 30 days to comment. We are driven by a strong interest in philanthropy and a simple conviction: decisions about people should be fair, and fairness can be measured.
Angoff has no connection to William H. Angoff, his family or Educational Testing Service.