Counterfactuals

Counterfactuals are conditional statements about what would be the case if circumstances were different, typically concerning events that did not happen in the actual world. They express subjunctive conditionals such as “If X had occurred, Y would have followed.” Their truth conditions depend on how closely the actual world is related to nearby possible worlds in which the antecedent holds.

In philosophy and linguistics, counterfactuals are analyzed with possible-world semantics. The leading accounts include David Lewis’s similarity-based approach and Robert Stalnaker’s minimum-change framework. These theories aim to explain when a counterfactual is true by comparing the actual world with the nearest worlds where the antecedent is true, and by specifying how much would need to change to accommodate the hypothetical.

Counterfactual reasoning plays a central role in causal analysis. Causal counterfactuals consider what would have happened if a factor had been different, and they underlie approaches such as the potential outcomes framework in statistics. They are used in law, medicine, and public policy to evaluate hypothetical interventions and to reason about responsibility, treatment effects, and policy design.

In artificial intelligence and machine learning, counterfactuals are used to generate explanations that indicate how altering an input would change an outcome, supporting model interpretability. Historically, counterfactual reasoning appears in early philosophical work and was formalized in the 20th century by Lewis, Stalnaker, and others, highlighting its broad influence across disciplines.