Role ambiguity for a Black-accented agent destabilized trust, culminating in breakdown after the agent inaccurately tracked group decisions¶

Claim · All claims
Evidence · 1 study · qualitative r3 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? In the Black-accent case study, Phoenix's role oscillated between peripheral and authoritative, and inaccurate tracking of the group's ranking triggered visible frustration and disengagement. → Ravi 2026

Evidence¶

Ravi 2026¶

Ravi, P., Stevens, C., Hurt, B., Hanks, B., Lin, G., and Anderson, E. (2026). Exploring How Agent Voice Accents Shape Human-AI Collaboration in K-12 Group Learning. https://arxiv.org/abs/2606.12805 doi:10.48550/arxiv.2606.12805

q2 · i? · qualitative · r3

Interaction Analysis case study of Group 1 in the Black-accent condition. The group probed Phoenix's role during the icebreaker, later treated it as a validating authority, and after persistent ranking errors questioned its competence, lowered expectations, and disengaged.

"However, Pheonix's use as validating authority ultimately led to a trust breakdown when it inaccurately tracked the group's decisions."

Discussion¶

Related Claims¶

• Degree of anthropomorphism attributed to the agent was shaped by both interaction modality and, within audio conditions, voice accent — related
• Role expectations for a group-facing AI agent are anchored in early encounters rather than renegotiated throughout collaboration — related
• Agent accent condition showed no statistically significant differences in teachers' ratings of the agent's proficiency, etiquette, or personality — related
• An Indian-accented agent was integrated as a trustworthy conversational peer through repeated, contextually useful contributions — related
• A British-accented agent was framed as a background technological helper and othered rather than integrated as a peer — related
• Teachers drawing Black- and Indian-accented agents predominantly depicted them as human, while British-accented agents were drawn as technology — related
• In CLL, teachers perceive a trust erosion cycle, AI dominance in groups, collaborative hollowing, and a counselor's dilemma — related
• Participants rated the arithmetic-error peer agent more harshly than the conceptual-error peer agent, calibrating trust by error observability rather than severity — related
• Students anthropomorphized the chatbot and built relationships with it through personal and social questions — related
• Error timing and stakes shape trust in AI systems: later errors may be more damaging, early errors may matter more, and trust recovers quickly after good post-error performance, with high-stakes errors causing abrupt collapse — related