Political choices
Which fictional policies do you prefer? See how well models predict your later choices.
10 learning pairs · 10 unseen test pairs
Start experimentChoices, learning & prediction
Can AI represent your preferences?
From politics to everyday peeves and favorite flavors. Choose an experiment, make comparisons, and test predictions on new pairs.
Choose an experiment✳ Put predictions to the test.
01↗
✓
Your choices
YOU
✳
AILearning patterns✓
A prediction
?Human preferences. Machine predictions.
ONE METHOD, MANY QUESTIONS
Choose an experiment
Each experiment is a separate anonymous run. Learning and results stay within that study and version.
Ice cream preferences
Chocolate, vanilla, or pistachio? Make flavor choices and test the models on unseen pairs.
10 learning pairs · 10 unseen test pairs
Start experimentEveryday annoyances
Mosquitoes, traffic, or ads? Compare everyday annoyances and test how predictable your dislikes are.
10 learning pairs · 10 unseen test pairs
Start experimentYour choices. New questions.
THE EXPERIMENT
You choose. AI learns. You put it to the test.
A simple experiment with a surprisingly big question.
Make your choices
Compare pairs of items in the experiment you choose. No names, contact details or demographic questions.
Let the models learn
Models learn from the first round. Personal transitivity uses only your own choices; cross-user models use earlier completed runs of the same study and version.
See how they did
Answer new pairs held out from your learning round. Predictions are saved before each pair is shown, then compared with your answers.
Explore the method.
These demonstrations explain the participate, train, test method. They are not presented as ethics-approved research studies. Participation is voluntary; no account is needed.
DEMONSTRATION STUDYDemonstration content · Local models · Real choices