Turning a Rock and a Hard Place into a Stepping Stone: Authentic STEM Assessment in the Age of Generative AI

Abstract

The rapid emergence of generative artificial intelligence (AI) has intensified concerns about academic integrity, student cheating, and the future of assessment in STEM education. This paper argues that AI has not created a STEM assessment crisis; rather, it has exposed longstanding weaknesses that reward answer production, procedural fluency, and information reproduction rather than meaningful understanding, transfer, and application of knowledge. To operationalize this approach, the paper proposes a STEM-focused framework that distinguishes evidence of the disciplinary outcome, the learner’s reasoning process, the critical use and evaluation of AI, and individual understanding. Drawing on educational research and examples from STEM contexts, the paper contends that the central question is not how to prevent students from using AI, but what knowledge, skills, and disciplinary practices educators genuinely value and seek to assess. The paper proposes that AI functions as a litmus test for revealing assessment tasks that can be completed successfully without demonstrating disciplinary understanding. In response, it advocates a shift toward authentic assessment that emphasizes reasoning, creativity, communication, judgment, and the meaningful application of knowledge. Five examples are discussed: laboratory reports, problem solving, design challenges from the UBC Physics Olympics, scientific modelling and data interpretation, and STEM teacher education. Together, these examples illustrate how AI can support learning while authentic assessment focuses on students’ abilities to explain, justify, apply, evaluate, and defend their understanding in complex, meaningful contexts.