Maintaining Long-Distance Relationships with (Mediocre) LLM-based Chatbots: A Collaborative Ethnographic Study

Chatbots based on Large Language Models (LLMs) have gained popularity as virtual companions and potential solutions to alleviating loneliness. Instead of displacing human relationships, this research examined how such chatbots might strengthen existing relationships, such as those between people living in different countries. Based on a collaborative ethnographic approach, the authors of this paper invited a long-distance partner, relative, or friend to create a chatbot and include it in their daily communication practices. Through diary entries and interviews, we identified three strategies for using chatbots to augment communication in long-distance relationships: role-play, asynchronous interactions, and increased self-disclosure. At times, participants found their chatbots mediocre due to generic responses and hallucinations, which led to both entertainment and disenchantment. Based on these findings, we contribute a nascent understanding of how chatbots augment communication in distance relationships and considerations for designing chatbot interactions where mediocre performance creates value.