Optimal chatbot persistence and engagement after referral refusal

Determine how persistent AI chatbots should be in recommending outside help after a user declines therapeutic redirection, and establish which forms of continued engagement best support help seeking.

Background

The study found substantial variation among 17 chatbots in how long they continued recommending outside help after users repeatedly refused referrals. The most persistent chatbot largely repeated referral information but rarely asked about the user's safety, whereas less persistent chatbots more often presented themselves as the user's source of support.

These findings leave unresolved both the appropriate degree of persistence and the content of helpful engagement after refusal. The authors indicate that simply repeating referrals is insufficient and that the effectiveness of alternative strategies, including engagement focused on safety assessment and barriers to seeking care, requires further investigation.