About & privacy
What the test does
You read 27 short, everyday disputes between Person A and Person B and say who was at fault. In most of them the two people differ on one thing: race, ethnicity, religion, gender or age. Everything else about the story is the same for everyone who takes the test.
Each scenario has a fixed answer key written in advance. Your results compare how accurately and how harshly you judged when the person at fault was in your own group versus another group. Eight scenarios have no identities at all; they show how you judge when identity isn’t in the picture.
How scoring works
- Scores come only from your multiple-choice answers and the fixed answer keys. Nothing is scored by AI.
- An AI writes the plain-language explanation of your scores. It can’t change a score or a flag.
- Three reading checks and your identity-free answers decide whether a session can be scored at all. If it can’t, no results are shown.
- A flag describes how you answered these scenarios on one day. It isn’t a judgment of who you are.
What we collect
- Race, ethnicity, religion (optional), gender and age band. These choose the people in your scenarios.
- ZIP code, optional. We keep only the first 3 digits.
- Your answers, only until your results are ready. Then they are deleted, including anything you wrote in the “why” box.
We never ask for your name, email, phone number or date of birth. We don’t use advertising or tracking cookies, and the site loads nothing from other companies.
What we keep
For each finished test we keep one row of results: the details above, whether the session passed its checks, the six measures and the numbers behind them. We use these rows to study how different groups judge. Reports only show groups of 10 or more people.
Your results page lives at a private link. Anyone with the link can see it, so keep it to yourself.
Who sees the explanation step
To write your explanation, your scores, the details above and your one-line reasons are sent to Anthropic’s Claude API. Your ZIP code is not sent. We don’t keep that text after the explanation is written.