Purpose-Dependent Selection of Relevant Differences: How the Purpose of a Comparison Decides Which Differences Count

This conceptual paper introduces Purpose-Dependent Selection of Relevant Differences, an evaluative criterion for judging whether cases can be treated as equivalent. The differences a comparison may disregard depend on the purpose the judgment serves. The paper asks whether ignored differences are irrelevant to the stated use. Parties can agree on the facts yet disagree about equivalence because they are judging for different purposes. This does not make equivalence arbitrary or all purposes equally defensible. Where the relevance of an ignored difference is unresolved, equivalence remains an assumption. The account relates the criterion to Kerr’s analysis of reward systems and examines the Cardiac Arrhythmia Suppression Trial. The case distinguishes successful arrhythmia suppression from improved survival: patients assigned encainide or flecainide had higher mortality than those assigned placebo. The discussion does not explain that mortality increase or generalize the findings to every setting. Part of the Equivalence and Interchangeability series within On Human Modeling, the paper offers questions about what comparisons are for, what they ignore, and who bears the consequences. Its framework can inform human–AI research on whether evaluation metrics and judgments of equivalent performance preserve the differences that matter for the intended use.