Reflections on the future of machine learning for materials research

TL;DR

This paper advocates for a strategic shift in applying machine learning in materials research, emphasizing model integration, bias awareness, scientific understanding, and ethical considerations to foster innovative and equitable scientific progress.

Contribution

It proposes a move from data quantity focus to model-centric approaches, emphasizing integration with theoretical models, bias management, and ethical practices in ML for materials science.

Findings

Encourages model-oriented ML over data-driven approaches.

Highlights importance of dataset bias analysis.

Calls for ethical and equitable ML practices.

Abstract

Applied machine learning (ML) has rapidly spread throughout the physical sciences; in fact, ML-based data analysis and experimental decision-making has become commonplace. We suggest a shift in the conversation from proving that ML can be used to evaluating how to equitably and effectively implement ML for science.We advocate a shift from a "more data, more compute" mentality to a model-oriented approach that prioritizes using machine learning to support the ecosystem of computational models and experimental measurements.We also recommend an open conversation about dataset bias to stabilize productive research through careful model interrogation and deliberate exploitation of known biases. Further, we encourage the community to develop ML methods that connect experiments with theoretical models to increase scientific understanding rather than incrementally optimizing materials. Moreover we envision a future of radical materials innovations enabled by computational creativity tools combined with online visualization and analysis tools that support active outside-the-box thinking inside the scientific knowledge feedback loop. Finally, as a community we must acknowledge ethical issues that can arise from blindly following machine learning predictions and the issues of social equity that will arise if data, code, and computational resources are not readily available to all.