Counterfactual Explanation
Definition
A counterfactual explanation describes which minimal change to an input would lead to a different model outcome.
It does not necessarily explain the internal model logic but shows an alternative case close to the starting point. Plausibility and actually changeable features must be considered.
Simply explained
Like saying: If feature X were different, the model would have decided differently.
Why it matters in practice
Relevant for understandable decision support and checking model boundaries.
Example from business
A model shows that a decision would have been different with lower utilization and higher delivery reliability.
Limits and pitfalls
- Changing immutable features; assuming causality; generating unrealistic combinations; understanding a single explanation as a guarantee.
Origin & history
Wachter, Mittelstadt, and Russell proposed counterfactual explanations in 2017 as an action-oriented form of explaining automated decisions.
Use in mid-sized companies
- Possible application in SMEs: Reviewing decisions with plausible, professionally permissible counterexamples.
Frequently asked questions
Does a counterfactual show cause?
Not automatically. Must the change be realistic? Yes, otherwise the explanation is of little use.
Sources
- Counterfactual Explanations · Wachter et al. (2017), Counterfactual Explanations
Sources checked on 10.10.2026.
Editorially reviewed on: