The Regenerative Compute Protocol: Using Agricultural Ecosystem Recovery Signals to Govern the Energy Consumption and Workload of Local Artificial Intelligence Systems
Description
Artificial intelligence is commonly evaluated through accuracy, speed, economic value, safety and computational efficiency. These measures are incomplete when an AI-enabled intervention affects interconnected human and ecological systems.
This paper proposes the Regenerative Intelligence Index (RII), a transparent framework for measuring whether an AI-enabled intervention produces verified net restoration or contributes to ecological and human depletion. The framework evaluates eight domains: climate and energy, water, soil and land, biodiversity, circular material use, human capability, distributive benefit and governance integrity.
RII distinguishes the environmental footprint of developing and operating an AI system from its immediate application effects and wider system-level consequences. Indicators are converted into bounded change scores, adjusted for evidence quality and uncertainty and aggregated only after non-compensable safeguards have been assessed. These safeguards prevent serious human-rights violations, ecological damage or governance failures from being concealed by positive results elsewhere.
A hypothetical agricultural decision-support example demonstrates the proposed calculation, sensitivity analysis and reporting structure. The paper also establishes a measurement protocol, validation program, classification system and anti-greenwashing requirements.
This publication presents a conceptual framework and prospective research methodology. It does not report completed experimental results or claim that the proposed index has already been empirically validated. The next stage is independent, multi-site testing comparing AI-enabled interventions with credible non-AI alternatives and existing practices.
Methods (English)
The Regenerative Intelligence Index produces a result containing a composite score, eight-domain outcome vector, safeguard status, uncertainty profile and declared assessment boundary. The proposed framework requires a defined intervention, credible counterfactual, functional unit, spatial and temporal boundaries, preregistered indicators, evidence-quality adjustments, sensitivity testing and independent verification for high-impact claims.
Other (English)
Most AI systems are judged by how fast, accurate or profitable they are. The Regenerative Intelligence Index asks a harder question: after accounting for the entire intervention, are people and the environment actually better equipped to function and recover than they were before? The framework is designed to expose hidden trade-offs and prevent a benefit in one area from being used to excuse serious damage elsewhere.