AI Best Practices for Industrial Ecology

This project comprises two parallel research efforts examining the state of artificial intelligence and machine learning in industrial ecology — a field that increasingly relies on computational methods to model material flows, supply chains, and environmental impacts.

The first paper, led by Rankin and colleagues, focuses on reproducibility. An assessment of ML applications in IE finds that the vast majority of studies do not provide sufficient methodological detail to replicate their findings — a serious concern for a body of evidence that informs both research and policy. The work calls for the adoption of community-wide best practices around data sharing, model documentation, and reporting standards.

The second paper, led by Taghdisian and colleagues, takes a broader evaluative approach, asking not just whether ML is done reproducibly but whether it is being used well at all. Drawing on a systematic review, it assesses how and when ML methods are applied across IE research, where those applications are methodologically sound, and where they are mismatched to the questions being asked.

Together, the two papers offer a comprehensive diagnostic of ML practice in industrial ecology — one focused on transparency, the other on appropriateness — with the shared goal of improving the quality and credibility of computational evidence in environmental research.