USC engineering faculty add AI assignments and oral checks on student learning

USC’s electrical and computer engineering faculty are bringing AI-generated code into assignments while asking students to demonstrate more of their knowledge in person. In an October 15 account, the university detailed changes spanning introductory projects, oral quizzes and video evaluations—pairing practice with AI tools with closer scrutiny of what students actually understand.

The push followed a classroom experiment a year earlier. Stephen Tu, teaching graduate-level machine learning, tested AI models on assignments that included advanced, multipage mathematical proofs. When he showed colleagues an AI solution beside a student’s solution, nobody could identify which was which, according to USC’s account.

Department chair Richard Leahy described that moment as a wake-up call. Faculty discussions over subsequent months went beyond finding alternatives to conventional assignments. They also considered what students should learn, how to teach it and how to motivate them to master fundamentals when AI can supply detailed answers.

A finished answer is not the whole assessment

In Mark Redekopp’s first-year Introduction to Embedded Systems course, EE 109, students now complete short oral quizzes after each lab. Hand-coding exercises also require verbal explanations. These checks ask students to articulate their reasoning, rather than leave instructors to judge only the work they submit.

Other faculty are putting more emphasis on in-person lab work, live data collection, physical measurements and interactive problem-solving. Those settings allow instructors to observe and coach students as they work. Redekopp also sees oral assessment as career preparation: explaining technical reasoning is a skill students will need on engineering teams.

Professor Bhaskar Krishnamachari has introduced another format: students record short videos explaining concepts covered in class. He then uses AI to analyze the submissions for class-wide knowledge gaps and adjusts later lectures to address them. Here, AI helps the instructor decide what needs more teaching, rather than simply helping students produce answers.

Just because students can generate the right answer for a homework or test question doesn’t mean they learned or understand the concepts

Bhaskar Krishnamachari, USC electrical and computer engineering professor

AI stays inside the coursework

The tighter checks accompany explicit instruction in using AI. In EE 250, Distributed Systems and the Internet of Things, Krishnamachari teaches students to prompt AI models to generate working Python applications. He has also used AI himself to move course software between two hardware platforms, from Raspberry Pi to ESP32 microcontrollers.

Professors Rehan Kapadia and Yasser Khan redesigned EE 105, Introduction to Electrical Engineering, around hands-on projects, much of the work centered on AI. A custom knowledge graph—a map connecting course concepts—guides students through the material and toward the foundational knowledge needed for practical experiments.

Kapadia argues that AI tools let undergraduates attempt more complex, substantive projects with advanced hardware than they could five or 10 years ago. Across the department, faculty are also using tools such as 3D visualization and video generation to present course content. Tu now holds dedicated class discussions on recent AI developments.

Faculty describe fundamental knowledge as a condition for effective AI use, not something the tools make unnecessary. Students need it to ask useful questions, spot mistakes and validate a final product. The department is also emphasizing AI’s limitations and the ethical and human effects of using it.

Leahy says feedback from alumni, advisers and research partners has helped shape that approach, with AI literacy becoming part of engineering workflows. Vice chair Michelle Povinelli draws the boundary at responsibility: deciding what deserves to be designed, choosing an approach and assessing its impact remain the engineer’s tasks.