A new study tested whether a simple interface change could help older adults be more careful with AI chatbots. The researchers created a "safety UI bundle" that included generic source labels and a nudge reminding users to verify information. They compared this to a standard chat interface in a randomized experiment with 200 older Chinese adults aged 60 and older, recruited from community sites, clinic waiting areas, and WeChat groups.
People using the safety interface reported a higher intention to verify the AI's answers (average 4.72 vs. 4.41 on a 7-point scale). They also showed better trust calibration, meaning they were less likely to over-trust the AI. However, the change did not significantly affect how much they relied on the AI, and their actual verification behavior (clicking to expand source info) was higher but not statistically significant.
The safety interface did not hurt comprehension, usability, or cognitive load. Perceived trustworthiness was slightly lower, but that may be a positive sign of healthy skepticism. No safety concerns were reported, but the study did not track adverse events.
This was a small, short-term experiment using screenshots, not a real-world test. The intervention was tested as a bundle, so we can't say which specific element worked. Also, the results may not apply to other age groups or cultures. Still, the findings suggest that simple design tweaks could help older adults use AI more safely.
