Index Errans

person in Black box

Chris Russell

Counterfactual explanation

in Black box

Chris Russell, machine-learning researcher and the source of the technical counterfactual method in Counterfactual Explanations Without Opening the Black Box. At the University of Surrey and the Alan Turing Institute at the time, later at Amazon and the Oxford Internet Institute; his work spans computer vision and fairness in machine learning.

Stake§

Academic and reform-oriented, with public funding at the time of the paper; the counterfactual method is his contribution to a legal-technical argument.

Russell provided the machine-learning content of the counterfactual explanation — the smallest change to the inputs, the closest possible world, that would yield a different outcome. For this topic it is a way to give recourse without opening the model, the technical move underneath the legal argument that he and his co-authors make against requiring full transparency.

Works in this corpus§

their concepts on the territory
Counterfactual explanationCounterfactual explanation

1 concept in this scholar's webopen the full territory →

excerpts

explanations can, in principle, be offered without opening the 'black box.' Looking at explanations as a means to help a data subject act rather than merely understand, one could gauge the scope and content of explanations according to the specific goal or action they are intended to support.
— Counterfactual Explanations Without Opening the Black Box: Automated Decisions and the GDPR (2018)

The legal move that matches the technical one: drop the demand to see inside the model, and ask instead what the person needs in order to act. An [[concept:explainability|explanation]] is judged by the recourse it enables, not by the mechanism it reveals.

on Counterfactual explanation, Explainability

You were denied a loan because your annual income was £30,000. If your income had been £45,000, you would have been offered a loan.
— Counterfactual Explanations Without Opening the Black Box: Automated Decisions and the GDPR (2018)

A [[concept:counterfactual-explanation|counterfactual explanation]] in one line: the smallest change to the inputs that would have flipped the decision. It tells the subject what to change without disclosing — or even requiring — the model's internals.

on Counterfactual explanation