Predicting Satisfaction of Counterfactual Explanations from Human Ratings of Explanatory Qualities

Marharyta Domnich, Rasmus Moorits Veski, Julius Välja, Kadi Tulver, Raul Vicente · 2025

Read Predicting Satisfaction of Counterfactual Explanations from Human Ratings of Explanatory Qualities by Marharyta Domnich, Rasmus Moorits Veski, Julius Välja, Kadi Tulver, Raul Vicente (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-13899.txt.

Description

Marharyta Domnich and four co-authors ask what makes a counterfactual explanation satisfying, in this 2025 paper. Such explanations tell a person what small change would flip a decision: earn a little more, and the loan is approved. Standard scores like sparseness miss what people actually like, while user studies do not scale.

The team analyzed ratings from two hundred six participants, who judged explanations on overall satisfaction plus seven qualities: feasibility, coherence, complexity, understandability, completeness, fairness, and trust. Feasibility, whether the suggested change can really be done, and trust that it would work were the strongest predictors of satisfaction, with completeness also counting. Even without those two, the other qualities explained fifty-eight percent of the satisfaction differences. More detail did not by itself annoy users. The authors advise tuning explanations to the user's expertise and setting.

Catalog record

CREATORMarharyta Domnich, Rasmus Moorits Veski, Julius Välja, Kadi Tulver, Raul Vicente
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-13899

Browse the full library · About the Tower of Babel