AI teammates are increasingly proposed to support teams during conflict, yet builders and researchers still lack a compact vocabulary for describing how such systems intervene. We introduce a taxonomy for AI teammate conflict support that links conflict type and trigger cues to mediator strategy, stance, and intervention goals. We developed a baseline from 23 previous studies and refined it through matched speculative design workshops in Germany and Japan that generated ten conflict scenarios. The taxonomy supports structured analysis of conflict episode segments through a fixed sequence from conflict type and trigger cues to strategy, stance, intervention goals, and optional team and authority descriptors. We also provide coding guidance for recurrent ambiguities and a worked example that shows how the taxonomy can inform analysis and design reflection. This summative taxonomy offers initial cross-site refinement across two settings. It is intended for early analysis and design exploration, and future work should test the integrated taxonomy on real or realistic traces.
Chen et al. (Fri,) studied this question.