DataTerminology.comNEW
A hypothetical scenario representing what would have happened under different treatment or conditions.
DESCRIPTION
Counterfactuals underpin causal inference by comparing outcomes under alternative interventions or exposures. In the potential outcomes framework, counterfactuals define treatment effects as differences between observed and hypothetical outcomes. Structural causal models and do-calculus provide tools to reason about counterfactuals and identify causal effects. Applications include policy evaluation, uplift modeling, and fairness analysis in AI. Estimation techniques rely on assumptions such as ignorability and stable unit treatment value (SUTVA).
ALTERNATIVE DEFINITIONS
Wikipedia
Counterfactuals are conditional statements exploring what would have happened if circumstances had been different.
General philosophical formulation of counterfactuals.
Hernán & Robins (Causal Inference: What If)
Counterfactual outcomes are the outcomes that would be observed under alternative interventions, even if those interventions did not actually occur.
Anchors counterfactuals in the potential outcomes framework for causal inference.
Stanford Encyclopedia of Philosophy
Counterfactuals are subjunctive conditionals concerning what would be the case if certain antecedent conditions were different.
Philosophical perspective on subjunctive conditionals and semantics.
RELATED TERMS
Causal Inference Potential Outcomes Potential Outcomes Framework Structural Causal Model Do-Calculus Average Treatment Effect Directed Acyclic Graph Uplift Modeling A/B Testing
COMPETING TERMS
Factual
KNOWLEDGE AREAS
Data Science AI Mathematics AI Governance
TARGET AUDIENCE
Data Scientist AI Engineer AI Product Manager Chief AI Officer Data Strategy Consultant
EXTERNAL SOURCES
- Wikipedia - Counterfactual
- Harvard T.H. Chan School of Public Health - Open causal inference textbook
- Stanford Encyclopedia of Philosophy - Philosophical treatment of counterfactuals
TOP QUESTIONS
Why are Counterfactuals central to causal inference?
They define what would have happened under alternative treatments, enabling estimation of causal effects from observed data under specific assumptions.
Category: Relevancy
How do Counterfactuals appear in uplift modeling?
Uplift models estimate differential outcomes between treated and untreated states—essentially counterfactual contrasts for individual predictions.
Category: Example
What assumptions relate to Counterfactuals?
Ignorability, SUTVA, and consistency are key. Violations can bias estimated counterfactual effects.
Category: Requirements
What is a common misunderstanding about Counterfactuals?
They are not directly observable; they are inferred using models and assumptions, which must be validated.
Category: Misunderstanding
How do DAGs help with Counterfactual reasoning?
DAGs represent causal structure, clarifying confounding and identifying valid adjustment sets for estimating counterfactual effects.
Category: Application
How are Counterfactuals used in AI fairness?
Counterfactual fairness evaluates whether outcomes would remain unchanged under different sensitive attributes, holding other factors constant.
Category: Topic
What risks arise when modeling Counterfactuals?
Misspecified models or invalid assumptions can lead to misleading policy recommendations and biased inferences.
Category: Risk
Which methods estimate Counterfactuals from data?
Matching, inverse probability weighting, doubly robust methods, and causal forests estimate treatment effects under assumptions.
Category: Alternatives
Where should a practitioner start learning Counterfactuals?
Begin with Hernán & Robins’ textbook, Pearl’s work on SCMs, and causal DAG tutorials for foundational understanding.
Category: Overview