We propose a simple definition of an explanation for the outcome of a classifier based on concepts from causality. We compare it with previously proposed notions of explanation, and study their complexity. We conduct an experimental evaluation with two real datasets from the financial domain.

Additional Metadata
Persistent URL dx.doi.org/10.1145/3399579.3399865
Conference 4th Workshop on Data Management for End-To-End Machine Learning, DEEM 2020 - In conjunction with the 2020 ACM SIGMOD/PODS Conference
Citation
Bertossi, L, Li, J. (Jordan), Schleich, M. (Maximilian), Suciu, D. (Dan), & Vagena, Z. (Zografoula). (2020). Causality-based Explanation of Classification Outcomes. In Proceedings of the 4th Workshop on Data Management for End-To-End Machine Learning, DEEM 2020 - In conjunction with the 2020 ACM SIGMOD/PODS Conference. doi:10.1145/3399579.3399865