Beyond that, however, we should reframe the problem as AI-assisted decisions instead of AI-driven. Just like we can use measures of risk to inform design, regulation, or operations, AI can provide useful insights that assist in decision-making—but it should not be the primary driver of any decisions. If we consider AI as an assistant to a qualified human expert, then we can evaluate AI output as one input in the decision-making process.
When we still rely on a human subject-matter expert to make final decisions, the question shifts to “How can we trust humans to evaluate AI recommendations in the context of all other inputs?” We already trust humans to weigh evidence and make decisions, often in the face of incomplete data or contradictory indications. With proper training to interpret AI recommendations and understand the limitations, human experts can consider AI-generated results (along with explanations and measures of performance) as one more input for review. The true trust comes not from replacing human judgment and accountability but from using AI as one part of a holistic, human-driven decision-making system.
Jamie Coble ([email protected]) is a professor in the Department of Nuclear Engineering and Engineering Physics at the University of Wisconsin–Madison. Her expertise is in applications of data analytics, machine learning, and artificial intelligence to support robust decision-making and operations and maintenance planning in nuclear power plants and other nuclear facilities.
Leave a Reply