On Computing Probabilistic Explanations for Decision Trees

View Full Paper

Ask AI

Bookmark

Finding minimal δ-sufficient-reasons is computationally intractable, highlighting challenges in explainable AI.
The study shows that both features and inclusion-minimality of δ-sufficient-reasons are complex problems.
It extends previous findings on the hardness of explanations for boolean circuits to decision trees.
Identifying certain structural restrictions can make the problem of finding δ-sufficient-reasons tractable.

Discussion

No takes yet. Share an insight, caveat, or question.

Formal xai reveals intractability of δ-sufficient-reasons in decision trees, suggesting new structural insights.

Arenas et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark