Rethinking Engineering Assessment in the Age of Generative Artificial Intelligence: Challenges, Risks, and a Responsible Redesign Framework

Abstract

The advent of generative AI (GAI) and in particular, large language models (LLMs) has presented a set of new problems for the purposes of assessment in higher education. The impact on engineering education is especially significant as many frequently used assessment methods (including coding assignments, numerical problem solving, lab reports, design documentation, literature review, and project reports) can be completed or significantly aided by the use of AI tools. The use of GenAI can be helpful as an explanatory aid, a problem-solving tool, feedback and debugging guide, and even as a writing assistant, but it does have several issues to consider related to academic integrity, authenticity of student work, overreliance, technical inaccuracies, privacy concerns, confidentiality concerns, and the loss of foundational engineering skills. In this paper, a conceptual overview of the problems of assessment in engineering education is introduced in the era of GenAI. It does not report primary experimental data, survey data or institutional implementation data. Rather, it draws on existing concerns raised in recent educational guidance and engineering education literature and offers a framework for a redesign of assessments that is responsible. It proposes to focus on Engineers' Judgement rather than only answers, transparent policies for use of AI, alignment with learning outcomes, process-based evidence, oral verification, practical demonstration, reflective disclosure, and an evaluation of the process. The paper proposes that the assessment of engineering students needs to go beyond relying on AI-powered detection systems and blanket policies that prohibit. Rather, assessments should be restructured to reveal student thinking, designing decisions, verification strategies, and practical skills. Finally, the paper offers that the introduction of GenAI could be considered a learning support tool in appropriate cases; the ethical considerations, technical validation, and engineering reasoning, however, must remain with the student and educator.