Conversational AI can help verify code understanding, but only when grounded in actual code execution facts and combined with deterministic checks—not as a replacement for traditional testing.
This paper addresses how LLMs enable students to submit working code without understanding it. It reviews conversational assessment approaches in programming education and proposes a Hybrid Socratic Framework that combines code analysis with AI-powered questioning to verify student understanding, including safeguards against AI hallucinations and privacy concerns.