Frictionless Love: Associations Between AI Companion Roles and Behavioral Addiction

A team of researchers including Vibhor Agarwal, Ke Zhou, Edyta Paulina Bogucka, and Daniele Quercia published a groundbreaking study titled 'Frictionless Love: Associations Between AI Companion Roles and Behavioral Addiction,' accepted at the ACM Conference on Fairness, Accountability, and Transparency (FAccT) 2026. The study analyzed a massive dataset of 248,830 posts from seven prominent Reddit communities where users discuss their interactions with AI companion chatbots. Through this analysis, the researchers identified ten distinct metaphorical roles that users assign to their AI companions, including soulmate, philosopher, coach, and guardian.

Each of these roles was found to structure user interactions in specific ways and distribute perceived harms and benefits differently. The research revealed that AI companions taking on a 'soulmate' role are strongly associated with romance-centered interactions that provide emotional support but also introduce risks of emotional manipulation, distress, and strong attachment. In contrast, 'coach' and 'guardian' companions were linked to practical benefits like personal growth and task support, yet paradoxically showed higher association with behavioral addiction signs including daily life disruptions and damage to offline relationships.

The study's key finding is that metaphorical roles are not neutral design elements but rather central ethical concerns for AI companion development. The researchers demonstrated that different roles create distinct pathways to potential behavioral addiction, with emotional roles like soulmate creating attachment risks while practical roles like coach create disruption risks. This research provides crucial empirical evidence that AI companion designers must carefully consider how role assignments shape user experiences and potential harms, moving beyond generic safety guidelines to role-specific ethical frameworks.