We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Contrastive Explanations for Model Interpretability

Abstract: Contrastive explanations clarify why an event occurred in contrast to another. They are more inherently intuitive to humans to both produce and comprehend. We propose a methodology to produce contrastive explanations for classification models by modifying the representation to disregard non-contrastive information, and modifying model behavior to only be based on contrastive reasoning. Our method is based on projecting model representation to a latent space that captures only the features that are useful (to the model) to differentiate two potential decisions. We demonstrate the value of contrastive explanations by analyzing two different scenarios, using both high-level abstract concept attribution and low-level input token/span attribution, on two widely used text classification tasks. Specifically, we produce explanations for answering: for which label, and against which alternative label, is some aspect of the input useful? And which aspects of the input are useful for and against particular decisions? Overall, our findings shed light on the ability of label-contrastive explanations to provide a more accurate and finer-grained interpretability of a model's decision.
Comments: Accepted to EMNLP 2021 as a long paper
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as: arXiv:2103.01378 [cs.CL]
  (or arXiv:2103.01378v3 [cs.CL] for this version)

Submission history

From: Alon Jacovi [view email]
[v1] Tue, 2 Mar 2021 00:36:45 GMT (5382kb,D)
[v2] Thu, 9 Sep 2021 22:12:10 GMT (5690kb,D)
[v3] Tue, 14 Sep 2021 06:51:43 GMT (5690kb,D)

Link back to: arXiv, form interface, contact.