Designing Trustworthy AI Systems for Modeling Critical Minerals Recovery Processes
Authors/Creators
- 1. Rose-Hulman Institute of Technology
-
2.
Lawrence Berkeley National Laboratory
- 3. University of Utah
-
4.
The University of Texas at Austin
Description
As AI assistance lowers the barrier to coding, domain experts with limited programming experience are increasingly engaging with scientific computing software. However, a critical gap persists: evaluating and trusting AI-generated code remains a significant challenge for verification and validation.
Drawing on our experience building the Process Optimization and Modeling for Minerals Sustainability (PrOMMiS) Flowsheet Inspector, a domain-specific tool for code-driven computational software, this poster details our methodology for building user trust. We demonstrate how we enable AI agents and design intuitive UX features to confidently integrate AI assistance into complex computational workflows.
PrOMMiS allows chemical engineers to model critical minerals recovery, but the reliance on programmatic Python frameworks slows down setup and troubleshooting for non-programmers. We illustrate specific features within the PrOMMiS UI that directly address this: enabling users to validate AI-driven changes, improve diagnostic accessibility, and enhance error discoverability through an intuitive interface.