Planet-Centered AI (PCAI) proposed to replace human-centric paradigms

Maria Perez-Ortiz's latest paper, accepted at ICML 2026, makes a bold case for abandoning purely human-centric AI design. The author argues that contemporary AI paradigms — narrowly focused on individual user preferences, corporate objectives, or even general alignment with human values — are structurally inadequate for addressing global challenges like climate change, biodiversity loss, and geopolitical instability. Drawing on systems thinking, Perez-Ortiz proposes Planet-Centered AI (PCAI) as a research agenda that treats Earth as an interconnected whole where humans are one component. The paper diagnoses recurring limitations in existing frameworks: they ignore feedback loops across sectors, assume stable environments (non-stationarity), and fail to account for deep uncertainty — all of which become catastrophic under current planetary conditions.

PCAI fundamentally reshapes the AI lifecycle. It demands problem formulation that aligns with global sustainability goals, model design that incorporates system-level interactions, evaluation metrics oriented toward long-term trajectories rather than short-term accuracy, and deployment models that enable continuous monitoring of planetary impacts. The author advances a falsifiable claim: AI systems optimized without explicit consideration of systemic consequences are more likely to exacerbate instability than mitigate it. This paper serves as both a critique and a call to action, offering concrete directions for researchers and practitioners to build AI that serves not just human interests but the health of the entire planet.