The Reversed Turing Test

Abstract

This paper introduces The Reversed Turing Test, an innovative game that reimagines Alan Turing’s iconic question, "Can machines think?" by reversing the evaluation process. Players interact with a large language model (LLM) impersonating Turing, who seeks to determine if the player is human or AI. Set in the 1940s and 1950s historical context, the game combines immersive storytelling with period-accurate design. Developed using Unity and generative AI tools, it leverages LLM-driven dialogue to explore AI capabilities and limitations, such as coherence and interpretative challenges. By embedding AI’s imperfections into gameplay, the project fosters an understanding of LLM constraints while encouraging reflections on human-AI dynamics. Through its accessible yet thought-provoking design, the game contributes to the discourse on AI, which is often dominated by moralistic critiques or oversimplified portrayals that fail to capture AI’s complexities. This engaging approach contributes to the broader discourse on AI’s evolving role.