Classification-Induced Convergence: How Categories With Consequences Draw What They Classify Toward Their Boundaries

This conceptual paper introduces Classification-Induced Convergence, the mechanism by which classifications with different consequences draw responsive cases toward the features that secure more favorable treatment. A classification can thereby help change the population it sorts, including through the people who design or present the classified cases. The claim is causal and conditional. Classification alone is insufficient: differential treatment must create a reason to change, and the cases or those who shape them must be able to respond. Convergence need not be deceptive or undermine the classification’s purpose. Its assessment requires evidence that differential treatment contributed to movement toward favored features. The paper relates the mechanism to Hacking’s account of looping effects and examines NHTSA’s description of vehicle design responses to passenger-car and light-truck categories. The case presents an agency assessment in a contested rulemaking. It does not independently measure convergence or isolate the effect of regulatory definitions from other influences, including consumer demand. Part of the Equivalence and Interchangeability series within On Human Modeling, the paper offers questions about category consequences, responsiveness, and who bears the resulting changes. Its framework can inform human–AI research on how people adapt to consequential automated classifications and how those adaptations reshape the cases systems subsequently evaluate.