Gamifying Compassion: Mitigating Dialect Prejudice Through An AI-Driven Serious Game

Paper Title

Gamifying Compassion: Mitigating Dialect Prejudice Through An AI-Driven Serious Game

Publication Info

• Topic area: Addressing dialect prejudice using AI-driven serious games.
• Keywords: Dialect bias, accent bias, serious games, AI-mediated learning, empathy, intergroup contact, dialogue modification, bias mitigation, HCI, psychological safety.

Background and Problem

• Problem / challenge: Dialect and accent biases are pervasive, often unconscious, and normalized, leading to discrimination in social, educational, and professional contexts. Existing interventions primarily focus on documenting disparities or reactive fixes, leaving a gap in proactive, experiential learning tools.
• Significance: Addressing dialect prejudice is critical for fostering inclusivity and reducing systemic inequities in communication, hiring, education, and healthcare.
• Motivation and related work: Prior studies have documented accent bias in voice technologies, hiring practices, and social interactions, but rarely provide proactive, low-stakes environments for users to practice bias mitigation. Existing empathy-focused HCI tools often lack socially situated practice spaces where users can experiment with responses and observe their impact.

Solution

• Proposed approach: CompassioMate, an AI-driven serious game designed to mitigate dialect prejudice by fostering empathy, perspective-taking, and communication skills through interactive gameplay.
• Novelty:
  1. Integration of authentic dialect audio, region-mapping play, and bias-type diagnosis into a serious game framework.
  2. Use of an AI-driven dialogue modification system to provide adaptive, personalized feedback and facilitate bias repair.
  3. Design of branching narratives that make social consequences visible while maintaining psychological safety.
• Procedure and key techniques:
  • Players listen to dialect audio and identify its geographic origin on a map.
  • Engage in scenarios depicting dialect prejudice, identify bias triggers, and analyze underlying causes.
  • Rewrite dialogue to resolve conflicts, receiving AI-generated feedback and scoring based on empathy, proactiveness, constructiveness, and adaptability.
  • Progress through three levels: Recognition & Identification, Analysis & Understanding, and Intervention & Prevention.

Results

• Concrete findings:
  • Participants progressed from vague discomfort to precise analytical labeling of bias.
  • Dialogue-rewriting mechanics enabled players to develop actionable mitigation strategies, with AI feedback fostering iterative refinement.
  • Players reported increased confidence in compassionate communication and stronger empathy for individuals facing dialect discrimination.
• Advantage over baselines:
  • Unlike static or reactive tools, CompassioMate provides a dynamic, low-stakes environment for practicing bias mitigation through iterative, personalized feedback.
  • AI-mediated guidance ensures psychological safety, enabling candid reflection and skill development.
• Experiments / evaluation:
  • Conducted a three-week field study with 20 university students aged 18–24.
  • Used think-aloud protocols, semi-structured interviews, and verbal protocol analysis to assess engagement, learning outcomes, and user experience.
  • Achieved an inter-rater reliability of 0.86 (Cohen’s Kappa) for qualitative coding.
• Limitations and future work:
  • Lack of quantitative pre- and post-intervention measures to assess long-term behavioral change.
  • Homogeneous participant pool (university students) limits generalizability to broader demographics.
  • Scenarios are culturally specific to Chinese dialect politics, which may reduce transferability to other contexts.

Summary

CompassioMate is an AI-driven serious game designed to mitigate dialect prejudice by fostering empathy, perspective-taking, and communication skills. Through interactive gameplay, players identify bias triggers, analyze causes, and rewrite dialogue to resolve conflicts, receiving adaptive AI feedback. A three-week field study demonstrated that participants progressed from vague discomfort to precise bias labeling and actionable strategies, with AI guidance ensuring psychological safety. While effective in promoting initial attitude shifts, the study highlights the need for future research on long-term behavioral change and broader demographic applicability. This work contributes to HCI by providing a reusable design pattern for dialect-aware empathy games and insights into AI-mediated facilitation of sensitive topics.