In the Anthropocene, human activities dominate planetary change, yet the strategic motivations behind this unsustainable behavior remain poorly understood. Game theory has been used to analyze the strategic structure of the Anthropocene; however, it lacks empirical grounding. Here, we bridge this gap by inferring the strategic dynamics underlying global CO₂ emissions. We model emissions as the outcome of stylized games, then we use Bayesian probabilistic programming to compare ten model configurations against historical data.</p> <p>Our results indicate that the climate problem is more likely to be a free-riding problem, not a coordination failure. The inferred parameters capture a deep inequality in emissions, resonating with the Global North–South narrative and differentiated responsibilities. The projections align with RCP 6.0 (approximately 2.0–3.7°C warming by 2100; best estimate ~2.8°C).</p> <p>Our results suggest that intervention toward transforming the payoff structure must be ordered: the first priority is controlling the dominant emitter through reducing gains from defection through enforceable mechanisms and technological transformation; the second is addressing fear-driven defection by the minor emitter through transfers that support vulnerable actors. Methodologically, we demonstrate that Bayesian probabilistic programming can bridge the gap between stylized game theoretic models and empirical data, offering a principled framework for inferring strategic incentives from observed trajectories while quantifying parameter uncertainty.