Background Adolescents face a variety of potential harms in the online environment, including exposure to distressing illegal material, cyberbullying, image-based abuse, and “sextortion.” Various agencies provide on-demand helpline and information services for children and adolescents to support them with navigating online (and offline) harms. Objective This study examined whether a chat-based conversational agent (chatbot) might be a useful additional tool for meeting the needs of adolescents at risk from online harms. We developed a prototype chatbot—including both conversational and menu-driven user options—and evaluated users’ trust in the chatbot. In this context, trust relates to perceptions of the chatbot’s usability and the value of the information and support it provides. Methods Participants (n=224; mean age 16.8 years) interacted with the chatbots and evaluated them in terms of user trust: perceived usability and utility (ie, relevance of support resources provided). Results Most participants (conversational chatbot: 141/224, 63% and menu-driven chatbot: 142/224, 63%) showed a willingness to click on the chatbots’ recommended support links. Participants with higher trust in the chatbots were more likely to click the links for recommended support services (with extreme evidence for large effects: δ=0.73, 95% credible interval CrI 0.46-1.00 and δ=0.78, 95% CrI 0.50-1.07, for the conversational and menu-driven chatbots, respectively; Bayes factor BF10>50,000), and participants who clicked the links, compared with those who did not, reported higher rates of positive attitudes toward their decision (with extreme evidence for large effects: δ=0.87, 95% CrI 0.58-1.15 and δ=0.84, 95% CrI 0.54-1.12, for the conversational and menu-driven chatbots, respectively; BF10>3,000,000). The conversational and menu-driven chatbots differed little in perceived trust or effectiveness. Conclusions Chatbots represent a promising additional tool to help adolescents access mental health–related support services and navigate online harms. However, establishing trust is critical.
Charles et al. (Thu,) studied this question.