The degree to which a stated outcome, relation, or inference would continue to obtain under specified counterfactual changes to background conditions or interventions; evaluated relative to an explicitly defined set of alternative scenarios and assumptions.

Principle

Counterfactual robustness is meaningful only with an explicit specification of which interventions or background variations are treated as relevant; robustness is a relation between a claim and a set of counterfactuals rather than an intrinsic property of the claim alone.

Demonstration

Illustrative scenario: A policy claim—"Program P reduces recidivism"—is tested by comparing predicted outcomes under two counterfactuals: (a) introducing P while holding policing unchanged, and (b) introducing P during an economic downturn. If the reduction persists in (a) but reverses in (b), the claim is robust to some but not all relevant counterfactuals, showing conditional stability.

Misapplication

Treating robustness as absolute by testing only trivial or irrelevant perturbations (e.g., tiny numerical noise) and then claiming general stability; the error is failing to specify or vary the substantive background conditions that matter to the claim.

Consequence

When a claim has high counterfactual robustness relative to a well-justified set of alternatives, it supports stronger causal or policy inference and more reliable transfer across contexts; low robustness indicates sensitivity, limiting generalization and guiding further investigation into mediating conditions.

Reversal

If the underlying mechanisms are non-modular (interventions change the mechanism itself) or counterfactuals are ill-defined because of indeterminate dependencies, then robustness judgments may not obtain or may require reformulation in structural or model-relative terms.

Boundary

Clearly within: causal statements tied to specified interventions and background assumptions. Boundary case: statistical associations tested under some but not all relevant interventions. Clearly outside: descriptive correlations reported without any counterfactual or interventional framing.

Semantic Tension

Tension between breadth (robustness across many, diverse counterfactuals) and relevance (robustness only across counterfactuals judged salient for the claim); maximizing one can reduce the other.

Synthesis

Counterfactual robustness reframes ‘stability’ as a relational, context‑dependent judgment: one must specify which alternative worlds matter and how mechanisms respond there to determine whether a claim is reliably preserved.