Evidence Alignment and Transfer Boundaries in Ai-Assisted Programming

Abstract

Progress in AI-assisted programming depends on more than accumulating favorable results. This critical synthesis connects comparative analysis of observable coding patterns produced by people and machines with hierarchical debugging that closes the gap between generated code and executable correctness and asks how measurement choices, boundary conditions, and decision costs shape the interpretation of both. A structured reading of two target studies and 12 verified companion references is conducted across five lenses: programmer behavior, error localization, execution feedback, repair hierarchy, evaluation leakage. Emphasis is placed on the provenance of evidence, the comparability of baselines, and the consequences of alternative explanations. The combined literature indicates that methodological gains become actionable only when programmer behavior and error localization are evaluated together and when limits associated with evaluation leakage are explicit. This shifts the emphasis from isolated scores toward traceable chains of evidence and decision relevance. The article concludes with a research agenda built around transparent comparators, targeted stress tests, and evidence records that can be reused without overstating causal or practical reach.