Computation to Choose a Future: Planetary Stewardship in the Age of AI

Maria Pérez-Ortiz 

The accelerating transformations of the Anthropocene demand governance systems capable of anticipating and steering complex, nonlinear Earth-system dynamics. Existing models optimize for likely trajectories rather than exploring a broader set of futures. This commentary introduces the concept of Computational Foresight (CF): an integrative framework combining artificial intelligence, simulation, and complex-systems modeling to augment human anticipation and collective reasoning about the future. CF organizes foresight into five interlinked functions forming a continuous, reflexive cycle for anticipatory governance. It draws on advances in machine learning, reinforcement learning, causal inference, and generative modeling to detect emerging signals, map feedbacks, and test “what if” interventions within virtual environments. CF thus shifts computation from prediction to possibility mapping, treating uncertainty as a resource for learning. The paper outlines the key technical and ethical frontiers of this transition: validation under deep uncertainty, reasoning in open-ended domains, alignment with human values, prevention of algorithmic closure, pluralistic model integration, and genuine human–AI collaboration. CF is proposed as a new layer of civic and scientific infrastructure for Earth system stewardship, aiming to enable societies not merely to forecast the future but to co-design it through transparent, participatory, and adaptive intelligence.