REDESIGNING ASSESSMENT FOR AI-SUPPORTED PROJECT-BASED LEARNING: EVIDENCE FROM ENGINEERING EDUCATION

The integration of Artificial Intelligence (AI) tools into engineering education is reshaping how students learn, collaborate, and complete project-based tasks. In Project-Based Learning (PBL) environments, where students design, build, and justify complete engineering solutions, AI offers powerful advantages, including enhanced productivity, improved conceptual understanding, and support for ideation and debugging. However, increasing reliance on AI also presents significant challenges for assessment fairness, authenticity, and the evaluation of individual student contribution. This study investigates how engineering students use AI within a PBL course, examines their perceptions of its benefits and risks, and analyzes the implications for traditional assessment structures. Survey results from 38 students reveal that AI is extensively used across all stages of the PBL workflow, yet students acknowledge that fair assessment requires mechanisms ensuring they can explain, justify, and validate AI-assisted work. Based on these exploratory findings, this paper proposes an AI-aware PBL assessment framework grounded in four principles: transparency of AI use, prioritization of understanding over output quality, incorporation of multiple evidence sources for individual contribution, and proportional weighting of tasks susceptible to automation. The framework introduces practical mechanisms such as AI-use declarations, micro-vivas, process-oriented artefacts, and rebalanced rubrics, all of which can be integrated into existing assessment structures. This work contributes a practical CDIO-aligned approach for incorporating AI into engineering education without compromising academic integrity or learning outcomes. By emphasizing accountability, understanding, and responsible AI use, the proposed framework supports the evolving needs of modern engineering programs while maintaining fairness and rigor in PBL assessment.