Nasser Giacaman, Valerio Terragni, Paul Denny, Viraj Kumar (2026) — arXiv preprint.
Analyzes how students specify intended behavior in natural language to AI code tools (Copilot) across multiple years, deriving a taxonomy of code-generation specifications expressed through comments. As AI tools shift emphasis from writing code to specifying behavior, the study documents what students actually ask of these systems.
Situates in llm-assisted programming-its and student-experience within cs-education and higher-ed, extending reshaping-cs-education-genai by characterizing the new 'specification' literacy. It informs ai-literacy for coding and the design of tools that scaffold rather than replace student reasoning.
Related Pages
- llm — LLM code-generation tools
- programming-its — Programming intelligent tutors
- student-experience — Student specification behavior
- cs-education — Computing education
- higher-ed — Higher-ed contexts
- reshaping-cs-education-genai — GenAI reshaping CS education