Abduct, Deduct, Induct, Respawn! A Retroductive Perspective on Games

Games are more than systems to master; they are engines of iterative thought. Through play, users engage in a retroductive loop: an unexpected event triggers hypothesis formation (abduction), leading to action planning and execution (deduction) and learning from feedback (induction). While it is central to player experience, current game AI typically optimizes for challenge balancing or content generation, rarely reflecting back on the player's evolving mental model. We propose treating both the player and the game as coupled retroductive agents that continuously observe, hypothesize, act, and learn across nested timescales. We formalize the retroductive processes of each agent and model their co-evolutionary dialogue as an exchange of inferences and feedback, positioning the game not as a reactive system but as a thinking partner. Building on this formalism, we articulate the vision of retroductive play: Gameplay as an epistemic negotiation in which the system deliberately probes, challenges, and reshapes the player's mental model while refining its own, judging progress by shifts in understanding rather than optimized performance. Retroductive play invites players-and the game that serves them-to respawn their ways of thinking.