Loneliness is a growing and global issue that affects approximately a third of individuals in industrialized countries. Given the rapid increase in popularity of companion AI chatbot services such as Replika and Character.ai, it has become pertinent to investigate how their usage may influence an individual’s mental state, sense of loneliness and human relationships in both positive and negative ways. Users of Replika, stated by the company to be over 10 million, turn to their artificial companion to cope with loneliness and address their mental health. Character.ai, with over 20 million users, is another companion chatbot service established late 2022, where a user can role-play using a constructed persona and interact with chatbots based on fictional characters. In this service, the most popular mental health character by far – and the sixth most popular character as of January 2024 – is Psychologist, with over 70 million messages shared with the bot over a year, demonstrating that people are already turning to AI for mental help.
We propose to investigate how users of companion chatbots are influenced through their interactions with these conversational AI systems. Particularly, we plan to investigate how conversational AI systems change a user’s subjective experience of loneliness, their quality and quantity of socialization (with real people), and other aspects of their mentality, empathy, self-esteem, and level of social anxiety. Previous work has established that the ability to empathize is inversely correlated with loneliness over a lifetime, and role-play been shown to increase empathy. The ability of a user to role-play as fictional personas with AI agents – as encouraged by some of these services, including Character.ai, might have unique impact on users’ mental and social wellbeing, but has not been explored to date in the research literature, despite the large numbers of users who may be affected. Conversational AI systems now make it possible to convincingly simulate social agents with particular beliefs and agendas, thereby making diverse virtual social experiences more accessible. It is critical to understand how this technology can influence socialization and mental health so that we might be able to mitigate its negative effects and make the most of its positive effects. In this project, we intend to evaluate how the use of companion chatbots influences an individual’s socialization and loneliness.