Before you begin

What this study is about. We study how people use recommendations from an AI system when making repeated decisions, and how the way the AI presents itself affects those decisions.

What you will do. You will complete 60 short decision trials in two blocks (about 12–15 minutes). On each trial you give an answer and a confidence rating, see an AI recommendation, then give a final answer and confidence. There are no right or wrong ways to use the AI.

The AI. The recommendation system in this study is a research tool whose accuracy is controlled by the experimenters. It is right most of the time but not always. You will be told how it behaved in the debriefing.

What we record. Your answers, confidence ratings, response times and the condition you were assigned to. We do not record your name, email or IP address. Two optional questions ask for an age band and how often you use AI assistants; you can skip them. If you have a participant code from a recruitment platform we store only a salted hash of it.

Your rights. Participation is voluntary. You can close the page at any time. At the end you receive a deletion key; using it removes all of your data from the research database. Anonymised, aggregated data may be shared publicly for research.

Risks and benefits. There are no known risks beyond mild fatigue. There is no direct benefit; the study helps design AI systems that people rely on more appropriately.

Contact. Questions about the study or your data: see the About page.