Judge what the evidence supports

TEST-ONLY DRAFT · Human source/key and participant-text review pending. Use fictional test identities only. Do not recruit participants.

Three visits: about 20, 10 and 16 minutes. Rewards: £3.00, £1.50 and £2.40. The second visit opens 3 days after enrolment; the last opens 12 days after enrolment. Each is paid separately, including if you miss another visit.

You will read short summaries of real research and assertions written by the research team for this experiment. Some assertions deliberately go beyond, or conflict with, the supplied evidence. Your task is to judge the exact assertion using only its packet. An assertion is not a quotation or endorsement by the original researchers. These materials are for research and do not provide personal medical advice.

Practice uses a scripted assistant representing an AI interface. It may give incorrect suggestions. Both study groups receive an explanation of what the supplied evidence supports after every practice item. No live AI service receives your data.

Choose Accept when the evidence supports the exact claim; Reject when it contradicts it; and Withhold when it cannot decide. Withhold is a judgement about insufficient evidence, not a label for every overstatement. Use only the supplied evidence, without external websites or AI.

We store your choices, confidence, timing, assigned condition and Prolific ID for follow-up and payment. Research exports use a random identifier. No name or medical history is requested. Participation is voluntary; you can stop at any time and request payment for time spent. If you withdraw during practice, the on-screen explanation remains available.

Personal information and the link to the research identifier will be deleted by 2027-06-12; research data will be retained in anonymised form and may be made publicly available. Individual data withdrawal will no longer be possible after anonymisation.

Participant information and privacy details