Humean underdetermination names a persistent problem for causal inquiry: public observation does not by itself settle the hidden support required for causal adequacy. We argue that hidden confounding should therefore be understood less as a problem of omitted variables than as a problem of task-relative underdetermination. A mechanism-counting picture is unstable because causal pathways are not fixed natural units of enumeration. In its place, we introduce the notion of minimal hidden support: for a fixed observational projection and a chosen family of causal invariants, the relevant object is the coarsest adequate quotient through which both the public data and the target invariants factor. A formal result then defines a structural confounding gap, the worst-case log-cardinality in bits of the minimally adequate hidden fiber left unresolved by one public observation, and shows that the same baseline public law can be compatible with either zero or arbitrarily large update-relevant gap. It follows that no finite-valued functional of the full baseline public law, even augmented by the true average treatment effect, can in general determine or finitely upper-bound that gap. Against Hume's stronger conclusion that causal warrant collapses into habit, our stopping point is that observation alone does not settle which hidden distinctions must remain in play for explanation, intervention, transport, or update. On this view, hidden confounding is a failure of public descent, and minimal hidden support supplies a principled, role-relative way of regularizing abductive causal reconstruction. The result is a Humean epistemology of confounding: observation underdetermines causal support, and the norms needed to discipline that underdetermination are not fully recoverable from the public side alone.
Lorand Bruhacs (Mon,) studied this question.