Toward AI-Resilient Assessment in Engineering Education: Evaluating Student Video Presentations Through Speech and Language Analysis

In the era of generative artificial intelligence (GenAI), traditional assessment practices in engineering education face new challenges, particularly in verifying student authenticity and fostering deep learning. This study proposes a Speech and Language Analysis (SLA) framework for evaluating student-produced video presentations as a resilient and pedagogically meaningful alternative. Drawing on five analytic dimensions-clarity, accuracy, critical thinking, engagement, and originality-the study analyzes final-year engineering students' video submissions using both transcript-based linguistic markers and live speech delivery cues. Results show that structured video presentations promote higher-order thinking, reflective articulation, and expressive communication. The integration of manual rubric scoring with natural language processing techniques provides educators with diagnostic insights and scalable evaluation tools. Findings support the use of video-based coursework as a viable strategy for fostering student ownership, safeguarding academic integrity, and enhancing communication skills in technical disciplines. The study concludes with practical recommendations for embedding SLA-informed video assessments into engineering curricula as part of future-ready, AI-resilient pedagogical design.