Underdetermination of Causes by Compatible Traces: When the Same Evidence Fits More Than One History

This conceptual paper introduces Underdetermination of Causes by Compatible Traces, a structural constraint on causal inference from an available record, sign, or finding. When several admissible histories fit a trace, compatibility alone does not uniquely identify which history occurred. The paper distinguishes compatibility from equal probability and unique identification from warranted inference. Evidence can favor one history while alternatives remain possible. Its assessment standard asks how the available evidence bears on competing accounts, including whether proposed alternatives remain possible given what is already known. The account relates the constraint to contrastive underdetermination in philosophy of science and examines the statistical evidence in the Sally Clark case. The discussion separates an invalid calculation about one explanation from the comparison needed to assess rival explanations, and considers undisclosed evidence relevant to that comparison. It does not attribute a particular interpretation of the statistic to the jury. Part of the Generative Operations series within On Human Modeling, the paper offers questions for identifying admissible histories, assessing comparative support, and examining who bears uncertainty. These questions can inform human–AI research on whether causal explanations distinguish what evidence establishes, what it favors, and what remains compatible with it, while making supporting assumptions explicit.