Intermediate exams that require students to explain their own code to counter blind use of AI tools¶

Strategy · All strategies
Evidence · no claims cited

Description¶

Assessment combined project work with two intermediate exams designed to verify comprehension rather than tool output. "Assessment was conducted through two intermediate exams. The first, after week five, consisted of presenting a ROS-based solution and answering technical questions." Although AI tools like ChatGPT and CoPilot were allowed, the exams forced students to dive deep into their code and convince the examiners with their knowledge, with the stated aim of avoiding blind use of chatbots.

Design Implications¶

Context¶

Requirements¶

• Exam format in which students present their own solution and answer unscripted technical questions about it

Constraints¶

• Stated for code-heavy robotics projects where AI coding assistants are permitted; the article does not report measured effects of this assessment design

Target Learners¶

• undergraduate students with access to LLM-based coding assistants

Target Learning Goals¶

• genuine comprehension of one's own developed solution rather than unexamined AI-generated code

Related Strategies¶

• Fade structured teaching sessions into independent project work over the semester

Examples¶

-

Key Sources¶

• Tobias Geger, Dominique Briechle and Andreas Rausch. (2026). Bots and Blocks: Presenting a project-based approach for robotics education. https://arxiv.org/abs/2603.14529