The Substrate Collapse: AI Code Generation Invalidates Authorship-Based Knowledge Metrics

The paper argues that AI code generation fundamentally undermines traditional authorship-based knowledge metrics in software engineering, such as the truck factor and Degree-of-Authorship metric. It contends that these metrics, which historically inferred understanding based on code authorship, are now invalid due to the disconnect between authorship and comprehension when AI-generated code is merged by humans. The authors propose that new metrics grounded in direct evidence of comprehension are necessary to address this measurement challenge, highlighting a critical gap in current methodologies for assessing software development teams' capabilities.