Privacy and Respectful Discourse in AI Chatbots

Chatbots’ ability to engage in fluid and complex discussions creates new challenges in privacy engineering. Starting with a brief overview of general-purpose AI chatbots, we present a case study of information disclosure in such bots by conducting a thematic analysis of 37 conversational logs from a publicly available general-purpose chatbot (Cleverbot). Our work provides developers and privacy researchers with an analysis of chatbot relationship-building that leads to information disclosure topics from our sample. We discuss (i) information disclosure mechanisms around the type of information users are willing to share and (ii) the types of sensitive data that can be gathered in an open-ended conversation. Our findings provide general insights into conversational expectations and disclosures of users, indicating a need for privacy-preserving design in chatbots through improved data handling. We close with recommendations for integrating privacy-preserving AI principles into conversational chatbot design and providing examples of how they can work in practice.