Self-driving labs get smarter as AI learns when to ask humans

A new study from Argonne National Laboratory and the University of Chicago presents an innovative “AI advisor” model designed to enhance collaboration between humans and autonomous laboratories in scientific discovery. Unlike traditional self-driving labs that follow fixed algorithmic plans, this system continuously analyzes experimental data in real time and alerts human researchers when their judgment could improve outcomes. Developed by a team led by Assistant Professor Jie Xu, the AI advisor adapts the experimental strategy dynamically, enabling a cooperative decision-making process that significantly boosts performance. Testing the model in Argonne’s Polybot lab, the researchers applied it to design mixed ion-electron conducting polymers (MIECPs) used in electronic materials. The AI advisor-driven approach yielded a 150% improvement in mixed conducting performance compared to previous methods. Beyond performance gains, the system also helped identify key material properties responsible for the improvement, demonstrating its ability to advance both practical results and scientific understanding. The researchers emphasize that while AI excels at data analysis, human intuition remains crucial when data is sparse