Case 429: A Murder Mystery Game for Teaching Bias in Generative AI

As generative AI tools become increasingly embedded in education and everyday decision-making, there is a pressing need to foster critical AI literacy—skills that go beyond technical knowledge to include the ability to recognize bias, question automation, and reason under uncertainty. We present a narrative-based mystery game designed to teach players how to detect and interrogate bias in AI-generated outputs. Set in the aftermath of a fictional murder, the game casts individual players as detectives who must solve the case using both human evidence and summaries generated by an AI assistant. These summaries contain embedded forms of algorithmic bias, including representation bias, automation bias, and proxy discrimination. Players interact with the system by cross-verifying summaries with source texts and highlighting statements as accurate, misrepresentation, or hallucination. Grounded in principles from constructivist learning and metacognition, the game encourages reflective judgment, situated learning, and active critique. We describe the game's learning design, narrative structure, interaction mechanics within this paper along with insights gathered from early playtests suggesting that players not only enjoy the mystery format but also begin to question AI outputs more critically. This work contributes to the growing field of AI literacy through games, offering a design framework for embedding ethical reasoning in narrative play.