ASIDE's AI pixel-art replays help couples resolve text conflicts

ASIDE, introduced by Xinyi Zhang and colleagues, addresses a core problem in AI-mediated conflict reflection: existing tools typically rely on one person's account, while dyadic tools fail to expose interpretation gaps. The system follows a Dyadic Spectator Reflection (DSR) sequence—externalize independently, then encounter together. From a past chat conflict, ASIDE generates a pixel-art theatrical replay featuring AI-generated inner-state hypotheses about each partner's thoughts and feelings. Each participant first confirms or revises hypotheses about their own state and separately edits their interpretation of the other person's state. Once both finish, they jointly view the co-annotated scene and examine Divergence Cards, which pair a confirmed account with the partner's reading at specific conversational beats.

The exploratory study involved 10 couples revisiting real text-based conflicts via ASIDE. Participants described the theatrical replay as enabling them to physically and emotionally step out of their roles, transforming from entrenched participants into co-observers. The pixel-art aesthetic and AI hypotheses lowered defensiveness and made interpretation gaps visible without being confrontational. The authors contribute DSR as a reusable interaction structure for dyadic reflection, and their findings suggest that AI-generated interpretive scaffolding can foster shared understanding in intimate relationships. This work opens pathways for integrating similar reflection mechanics into consumer messaging apps, couple therapy tools, and human-AI collaborative systems that mediate interpersonal communication.