Why Accuracy Isn't Enough: A New Quality Model for Real-World ML Components

AI components, whether traditional machine learning or generative AI, are embedded in many present-day systems. However, despite advances in automation, infrastructure, and tooling for developing AI components, many do not leave the prototype stage or reach production because they fail to meet overall system quality expectations. In our latest podcast from the Carnegie Mellon University Software Engineering Institute, Rachel Brower-Sinning and Robert Edman, both machine learning scientists in the SEI's Tactical Edge and AI-Enabled Systems Initiative, sit down with Grace Lewis, the initiative's lead and an SEI principal researcher, to discuss a quality model for machine learning components to support their proper testing and evaluation. The research team that developed the model also included Alex Derr, Sebastián Echeverría, and Ipek Ozkaya from the SEI, and, from Carnegie Mellon University, Kate Maffey. Colin Beaudoin, a PhD intern at the SEI in the summer of 2025 who is now an assistant professor at Fairfield University, also worked on the model.