• Being a doctoral researcher at the Research Unit

Power to the people(?) – (How) can users of ComAI applications participate in governance processes?

PhD project of Kim Lisa Ermler

“ChatGPT is my best friend. No joke. (…) I ask it for advice, vent to it, brainstorm ideas, even make big life decisions with it. Sometimes it honestly feels like it knows me better than people around me.”

Over the past three years, chatbots such as OpenAI’s ChatGPT have gained massive popularity. From the study of human-computer interaction, we know that people tend to treat computers as if they were social beings, despite knowing this is not the case. As the above comment from a Reddit user illustrates, users can form friendships with chatbots that feel as strong as, if not stronger than, those they build with other people. What we do not know yet is how chatbot providers govern these relationships through the means of content moderation and how this governance influences users’ experience.

Thus, the dissertation investigates how providers of Communicative Artificial Intelligence (ComAI) govern user-chatbot relationships. Based on the observation that users increasingly treat chatbots as companions, friends, or mentors, the research examines how ComAI providers regulate these relationships through the means of content moderation. Four sub-studies analyze how four major ComAI providers — OpenAI, Anthropic, Google and Meta — formulate, implement and are perceived in their governance of user–chatbot relationships. Study 1 conducts a longitudinal analysis of corporate policies including terms of service, community guidelines, and privacy policies, to identify and trace the evolution of key themes of content moderation. Study 2 takes an in-depth look at how providers conceptualize user–chatbot relationships in these policies. Study 3 examines traces of content moderation in real user-chatbot conversations using data donations, assessing how policy rules align with actual chatbot responses. Study 4 applies the concept of ‘folk theories’ to understand how users interpret the governance of their relationships with chatbots. The research combines qualitative and quantitative methods across the four studies. The overarching goal is to uncover how ComAI providers govern human-chatbot relationships and how this governance shapes user experience, bridging gaps in platform governance and human-computer interaction research.

Contact

• Kim Lisa Ermler