TOKYO, April 23 — Engineers at Japan’s National Institute for Materials Science (NIMS) have developed a system designed to make the discovery of new materials more transparent and reproducible, according to a study published in Science and Technology of Advanced Materials: Methods.

The system, called Pinax, records the entire process of materials development, including machine learning workflows and decision-making steps. Lead author Satoshi Minamoto said the tool formalizes both successful and unsuccessful trial-and-error processes, enhancing reproducibility, accountability, and knowledge sharing while maintaining strict data governance.

Machine learning models are increasingly used in materials research, but their reasoning processes are often opaque. Pinax aims to address this by visualizing the steps behind predictions, allowing other researchers to review, verify, and build upon them.

“This demonstrates how transparent AI systems can transform scientific discovery into a more reliable, efficient, and socially responsible endeavor,” Minamoto said.

The team tested pinax in two case studies: predicting steel properties and using transfer learning to estimate the thermal conductivity of polymers. The system linked performance predictions to specific data and model aspects, making complex workflows traceable.