Treatment Effect
Definition
The treatment effect describes what a measure actually changes in the outcome.
The causal difference between outcomes with treatment and without treatment for a unit or population.
Simply explained
It is not only important whether someone buys, but whether the discount caused the purchase. This exact difference is what one wants to determine.
Why it matters in practice
The term helps to select, develop, evaluate AI systems in a well-founded manner and realistically assess their technical risks.
Example from business
A retailer estimates whether a discount causes additional purchases instead of just reaching customers who would buy anyway.
Limits and pitfalls
- The individual counterfactual is unobservable; confounding, selection, and interference violate identification assumptions.
Origin & history
The potential outcomes model was formalized by Rubin from the 1970s onward for causal inference.
Use in mid-sized companies
- Possible application: A retailer estimates whether a discount causes additional purchases rather than just reaching customers who would buy anyway.
Frequently asked questions
What does treatment effect mean?
The causal difference between outcomes under treatment and without treatment for a unit or population. What should be considered? The individual counterfactual is unobservable; confounding, selection, and interference violate identification assumptions.
Sources
- Estimating Causal Effects of Treatments · Rubin, Estimating Causal Effects of Treatments
Source checked on 10.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: