Human intuition fuels AI-driven quantum materials discovery

quantum-materialsmachine-learningmaterials-sciencehuman-ai-collaboration

Abstraction: Expert-curated ML model encodes human intuition for quantum materials discovery

Key points:

Cornell's Eun-Ah Kim and Princeton's Leslie Schoop developed ME-AI (Materials Expert-Artificial Intelligence), a model that "bottles" expert intuition into descriptors for predicting material functional properties
Published in Communications Materials (DOI: 10.1038/s43246-025-00928-7); tested on 879 candidate materials to predict a specific desirable characteristic
ME-AI reproduced Schoop's own expert reasoning — she recognized her own thought process in the model's conclusions
Model generalized beyond the training set, predicting similar desirable materials in a different compound group
Key insight: indiscriminate data collection without expert curation is misleading; the expert decides fundamental features and curates data, then the machine learns from it
Framework designed for targeted discovery as opposed to serendipitous discovery; model for future AI-Materials Institute collaborations

Connections: Cornell University · Eun Ah Kim · Human AI Collaboration · Scientific Discovery