Validate individual AI-use attribution from generated-reference matching

Establish whether generated-reference matching can reliably determine how an individual student submission was produced and identify individual use of generative artificial intelligence.

Background

The study finds that generated-reference matching supports population-level analysis of code convergence but cannot by itself determine the provenance of an individual submission. Individual attribution remains unresolved because the corpus lacks verified submission-level source labels and contemporaneous process evidence. The paper identifies student disclosures, verified process histories, prompts, revisions, intermediate code, and other direct evidence as necessary inputs for future validation.

References

Determining how an individual submission was produced remains a separate question, one that requires student disclosure, verified process histories or other direct evidence.

— Correctness, Convergence, and AI-Generated Code Detection: A Longitudinal Study of Student and Large Language Model Code in Introductory Programming  (2610.00863 - Ye et al., 1 Oct 2026) in Section 7, “Conclusion”