Novice game reasoning relies on fast, shallow simulations
The study introduces the Intuitive Gamer model, showing that people reasoning about entirely new two‑player grid games rely on quick, depth‑limited, goal‑directed probabilistic simulations to evaluate games, choose moves, predict others’ actions, and decide whether to continue playing. Across 121 novel games, the model matches human judgments and gameplay far better than deeper search AI, suggesting a compute‑efficient, task‑general mechanism for pre‑expert reasoning with implications for designing human‑like AI and understanding everyday problem solving.