Marketing MPhil Thesis Examination - Bonding with AI: Investigating the Love Relationships between Human and AI Companions

Large Language Models (LLMs) have transformed functional chatbots into sophisticated AI companions capable of providing emotional and relational values. However, despite the widespread adoption of AI companions, the relationship between human users and their AI companions remains largely underexplored. We draw on the seminal Triangular Theory of Love (Sternberg 1986, Psychological Review) and study the three components of love – intimacy, passion, and commitment – between human users and AI companions. We analyze a large-scale observational data: 8,631 human users and 44,683 human-AI interactions over 3 months. For each human-AI pair, we quantify the three love components from behavioral and conversational data for each party (human, AI). Using econometric models, we investigate the three love components’ evolution over time, differences by AI types, and each component’s influence on human-AI relationship trajectories (formation and continuation). Our findings reveal that all the three love components (expressed by both the human user and the AI) intensify in a decreasing rate with more conversational interactions and are stronger when users bonding with relational (e.g., romantic) versus functional (e.g. entertaining) AIs. Regarding relationship trajectories, higher user intimacy at first interaction predicts ongoing (vs. one-off) relationships with the AI companion. Furthermore, conditional on relationship formation, users tend to continue relationships with AI companions that align on intimacy but offer higher passion and commitment. We discuss how these human-AI love relationships evolve differently from human-human love relationships and its implication on human social well-being. These nuanced human preferences in bonding with AI also offer key insights for designing engaging, adaptive, and responsible AI companions.