Standardizing AI-Assisted Coursework: A Log-Based Two-Track Evaluation Framework for University Engineering Reports

The widespread availability of generative AI tools has made it increasingly impractical to enforce “AI-free” policies in university coursework, especially in engineering and computer science report assignments. This paper proposes a practical two-track assignment policy that separates Non-AI submissions from AI-assisted submissions, rather than treating AI usage as a binary violation. The proposed framework defines clear submission requirements and evaluation criteria for each track. For the AI-assisted track, students are required to disclose tool usage, submit a minimal interaction log (e.g., key prompt–response pairs), and complete a short post-submission verification check to demonstrate conceptual ownership. For the Non-AI track, process-based evidence such as outline drafts and time-bounded writing conditions are emphasized to preserve authenticity. By redefining assessment goals, the framework aligns grading with real-world engineering competencies such as problem formulation, verification, documentation, and responsible tool use. This approach reduces enforcement friction, improves transparency, and provides an operational pathway toward standardizing AI-integrated coursework assessment. The paper also discusses implementation options and institutional considerations for adoption at scale.