‘Do you trust it or do you check?’: how children aged 8 to 10 verify an AI’s suggestions whilst playing

A biology-based reasoning game compares three types of AI agent. When presented as a partner, the agent encourages children to ask it questions; when presented as a tool, it encourages them to look for evidence.

Children are increasingly using AI agents to learn, while the fluid and engaging responses of these systems may be inaccurate or incomplete. A team of researchers has designed a reasoning game based on biological knowledge in which AI slips in suggestions, to observe not the confidence declared after the fact but the verification behaviors in the situation: acceptance, search for clues, questions, corrections, continuation of use.

Three styles of agent were compared: partner, tool, and an agent alternating the two. An initial evaluation with children aged 8 to 10, cross-referencing interaction diaries, questionnaires and interviews, suggests that style influences reaction: partner mode favours direct questions, tool mode searches for evidence. Children generally consider AI useful but fallible, with questions of varying depth. The authors draw design ideas to encourage critical and evidence-based use.