Towards Unifying Feature Attribution and Counterfactual Explanations: Different Means to the Same End
Open accessThe DOI matches its Crossref record; 32 of the 32 references it cites were found in a registry or an index.
Abstract
Feature attributions and counterfactual explanations are popular approaches to explain a ML model. The former assigns an importance score to each input feature, while the latter provides input examples with minimal changes to alter the model's predictions. To unify these approaches, we provide an interpretation based on the actual causality framework and present two key results in terms of their use. First, we present a method to generate feature attribution explanations from a set of counterfactual examples. These feature attributions convey how important a feature is to changing the classification outcome of a model, especially on whether a subset of features is necessary and/or sufficient for that change, which attribution-based methods are unable to provide. Second, we show how counterfactual examples can be used to evaluate the goodness of an attribution-based explanation in terms of its necessity and sufficiency. As a result, we highlight the complimentary of these two approaches. Our evaluation on three benchmark datasets --- Adult-Income, LendingClub, and German-Credit --- confirms the complimentary. Feature attribution methods like LIME and SHAP and counterfactual explanation methods like Wachter et al. and DiCE often do not agree on feature importance rankings. In addition, by restricting the features that can be modified for generating counterfactual examples, we find that the top-k features from LIME or SHAP are often neither necessary nor sufficient explanations of a model's prediction. Finally, we present a case study of different explanation methods on a real-world hospital triage problem.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Rudin, C. (2019) · Nature Machine Intelligence · doi:10.1038/s42256-019-0048-x · Cited by 10,561 · cited 1×
“explanation method should have, the two main types of explanations, and how different explanation methods compare to each other. There is also important work on building intelligible models by design [6, 28, 40] that we do not discuss here. 2.1 Desirable Properties of an explanation Explanations serve a variety of pur…”
BackgroundInfluentialLipton, Z. C. (2018) · Queue · doi:10.1145/3236386.3241340 · Cited by 5,427
Predicting the future - big data, machine learning, and clinical medicine
Obermeyer, Z., & Emanuel, E. J. (2016) · New England Journal of Medicine · doi:10.1056/nejmp1606181 · Cited by 3,652 · cited 1×
“appropriate medical interventions, and improving patient treatment rates [5, 12, 15, 17, 20, 30, 38, 40].”
BackgroundLearning Important Features Through Propagating Activation Differences
Shrikumar, A., Greenside, P., & Kundaje, A. (2017) · arXiv · doi:10.48550/arxiv.1704.02685 · Cited by 2,761
Causal Inference Using Potential Outcomes: Design, Modeling, Decisions
Rubin, D. B. (2005) · Journal of the American Statistical Association · doi:10.1198/016214504000001880 · Cited by 2,173 · cited 1×
“To provide a formal connection, we introduce the framework of actual causality [14] to the explanation literature that reasons about the causes of a particular event, unlike the more common causal inference setting that estimate the effect of a particular cause [35, 39].”
BackgroundAnchors: High-Precision Model-Agnostic Explanations
Ribeiro, M. T., Singh, S., & Guestrin, C. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.11491 · Cited by 2,162 · cited 1×
“A similar idea is explored in (local) anchor explanations [46].”
BackgroundIntelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30-day Readmission
Caruana, R., Lou, Y., Gehrke, J., Koch, P., Sturm, M., & Elhadad, N. (2015) · Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2783258.2788613 · Cited by 1,685
Explaining machine learning classifiers through diverse counterfactual explanations
Mothilal, R. K., Sharma, A., & Tan, C. (2020) · Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency · doi:10.1145/3351095.3372850 · Cited by 1,233 · cited 5×
“For our empirical evaluation, we use two attribute-based methods, LIME [37] and SHAP [29], and a counterfactual method, DiCE [33].”
BackgroundMethodInfluentialSelf-citationCounterfactual explanations without opening the black box: automated decisions and the GDPR
Wachter, S., Mittelstadt, B., & Russell, C. (2017) · SSRN Electronic Journal · doi:10.2139/ssrn.3063289 · Cited by 1,061
Questioning the AI: Informing Design Practices for Explainable AI User Experiences
Liao, Q. V., Gruen, D., & Miller, S. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376590 · Cited by 925
Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methods
Slack, D., Hilgard, S., Jia, E., Singh, S., & Lakkaraju, H. (2020) · Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society · doi:10.1145/3375627.3375830 · Cited by 900
Moons, K. G. M., Kengne, A. P., Woodward, M., Royston, P., Vergouwe, Y., Altman, D. G., & Grobbee, D. E. (2012) · Heart · doi:10.1136/heartjnl-2011-301246 · Cited by 889
ERASER: A Benchmark to Evaluate Rationalized NLP Models
DeYoung, J., Jain, S., Rajani, N. F., Lehman, E., Xiong, C., Socher, R., & Wallace, B. C. (2020) · Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2020.acl-main.408 · Cited by 561
Intelligible models for classification and regression
Lou, Y., Caruana, R., & Gehrke, J. (2012) · Proceedings of the 18th ACM SIGKDD international conference on Knowledge discovery and data mining · doi:10.1145/2339530.2339556 · Cited by 547 · cited 1×
“explanation method should have, the two main types of explanations, and how different explanation methods compare to each other. There is also important work on building intelligible models by design [6, 28, 40] that we do not discuss here. 2.1 Desirable Properties of an explanation Explanations serve a variety of pur…”
BackgroundInfluentialDesautels, T., Calvert, J., Hoffman, J., Jay, M., Kerem, Y., Shieh, L., Shimabukuro, D., Chettipally, U., Feldman, M. D., Barton, C., et al. (2016) · JMIR Medical Informatics · doi:10.2196/medinform.5909 · Cited by 509 · cited 1×
“ng patients who are likely to get admitted during emergency visits helps hospitals to better allocate their resources, provide appropriate medical interventions, and improving patient treatment rates [4, 8, 11, 13, 16, 25, 32, 34]. Given the importance of the decision, it is critical that the predictions from an ML mo…”
BackgroundLevin, S., Toerper, M., Hamrock, E., Hinson, J. S., Barnes, S., Gardner, H., Dugas, A., Linton, B., Kirsch, T., & Kelen, G. (2018) · Annals of Emergency Medicine · doi:10.1016/j.annemergmed.2017.08.005 · Cited by 424 · cited 1×
“Predicting patients who are likely to get admitted during emergency visits helps hospitals to better allocate their resources, provide appropriate medical interventions, and improving patient treatment rates [4, 8, 11, 13, 16, 25, 32, 34].”
BackgroundLai, V., & Tan, C. (2019) · Proceedings of the Conference on Fairness, Accountability, and Transparency · doi:10.1145/3287560.3287590 · Cited by 377 · cited 1×
“Explanations of a ML model’s prediction have found many uses, including to understand the most important features [29, 37], discover any unintended bias [42], debug the model [24, 37], increase trust [23, 27], and provide recourse suggestions for unfavorable predictions [33, 41, 49, 50]. There are two main kinds of ex…”
BackgroundSelf-citationPredicting hospital admission at emergency department triage using machine learning
Hong, W. S., Haimovich, A. D., & Taylor, R. A. (2018) · PLOS ONE · doi:10.1371/journal.pone.0201016 · Cited by 326 · cited 3×
“[15] who uses a variety of ML models including XGBoost and deep neural networks to predict hospital admission at emergency department (ED) from “triage” and demographic information, and other data collected during previous ED visits.”
BackgroundInfluentialFACE: Feasible and Actionable Counterfactual Explanations
Poyiadzi, R., Sokol, K., Santos-Rodriguez, R., De Bie, T., & Flach, P. (2020) · Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society · doi:10.1145/3375627.3375850 · Cited by 317
Horng, S., Sontag, D. A., Halpern, Y., Jernite, Y., Shapiro, N. I., & Nathanson, L. A. (2017) · PLOS ONE · doi:10.1371/journal.pone.0174708 · Cited by 310
On the Robustness of Interpretability Methods
Alvarez-Melis, D., & Jaakkola, T. S. (2018) · arXiv · doi:10.48550/arxiv.1806.08049 · Cited by 310
Algorithmic recourse: From counterfactual explanations to interventions
Karimi, A.-H., Schölkopf, B., & Valera, I. (2021) · Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3442188.3445899 · Cited by 274
Multi-Objective Counterfactual Explanations
Dandl, S., Molnar, C., Binder, M., & Bischl, B. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-58112-1_31 · Cited by 271
Explainability fact sheets: a framework for systematic assessment of explainable approaches
Sokol, K., & Flach, P. (2020) · Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency · doi:10.1145/3351095.3372870 · Cited by 264
Problems with Shapley-value-based explanations as feature importance measures
Kumar, I. E., Venkatasubramanian, S., Scheidegger, C., & Friedler, S. (2020) · arXiv · doi:10.48550/arxiv.2002.11097 · Cited by 228
Sharma, S., Henderson, J., & Ghosh, J. (2020) · Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society · doi:10.1145/3375627.3375812 · Cited by 226
Efficient Search for Diverse Coherent Explanations
Russell, C. (2019) · Proceedings of the Conference on Fairness, Accountability, and Transparency · doi:10.1145/3287560.3287569 · Cited by 218 · cited 2×
“Explanations of a ML model’s prediction have found many uses, including to understand the most important features [29, 37], discover any unintended bias [42], debug the model [24, 37], increase trust [23, 27], and provide recourse suggestions for unfavorable predictions [33, 41, 49, 50].”
BackgroundAn Electronic Emergency Triage System to Improve Patient Distribution by Critical Outcomes
Dugas, A. F., Kirsch, T. D., Toerper, M., Korley, F., Yenokyan, G., France, D., Hager, D., & Levin, S. (2016) · The Journal of Emergency Medicine · doi:10.1016/j.jemermed.2016.02.026 · Cited by 132 · cited 1×
“ng patients who are likely to get admitted during emergency visits helps hospitals to better allocate their resources, provide appropriate medical interventions, and improving patient treatment rates [4, 8, 11, 13, 16, 25, 32, 34]. Given the importance of the decision, it is critical that the predictions from an ML mo…”
BackgroundArya, R., Wei, G., McCoy, J. V., Crane, J., Ohman-Strickland, P., & Eisenstein, R. M. (2013) · Academic Emergency Medicine · doi:10.1111/acem.12249 · Cited by 78
Algorithmic Recourse: from Counterfactual Explanations to Interventions
Karimi, A.-H., Schölkopf, B., & Valera, I. (2020) · arXiv · doi:10.48550/arxiv.2002.06278 · Cited by 76
Sharma, S., Henderson, J., & Ghosh, J. (2019) · arXiv · doi:10.48550/arxiv.1905.07857 · Cited by 28
Haimovich, J. S., Venkatesh, A. K., Shojaee, A., Coppi, A., Warner, F., Li, S.-X., & Krumholz, H. M. (2017) · PLOS ONE · doi:10.1371/journal.pone.0172049 · Cited by 15 · cited 1×
“Predicting patients who are likely to get admitted during emergency visits helps hospitals to better allocate their resources, provide appropriate medical interventions, and improving patient treatment rates [4, 8, 11, 13, 16, 25, 32, 34].”
Background
Cited by
Counterfactual explanations and how to find them: literature review and benchmarking
Guidotti, R. (2024) · Data Mining and Knowledge Discovery · doi:10.1007/s10618-022-00831-6 · Cited by 471
A Survey of Algorithmic Recourse: Contrastive Explanations and Consequential Recommendations
Karimi, A.-H., Barthe, G., Schölkopf, B., & Valera, I. (2023) · ACM Computing Surveys · doi:10.1145/3527848 · Cited by 301
Counterfactual Explanations and Algorithmic Recourses for Machine Learning: A Review
Verma, S., Boonsanong, V., Hoang, M., Hines, K. E., Dickerson, J. P., & Shah, C. (2020) · arXiv · doi:10.48550/arxiv.2010.10596 · Cited by 267
Conceptual challenges for interpretable machine learning
Watson, D. S. (2022) · Synthese · doi:10.1007/s11229-022-03485-5 · Cited by 87
Alfeo, A. L., Zippo, A. G., Catrambone, V., Cimino, M. G. C. A., Toschi, N., & Valenza, G. (2023) · Computer Methods and Programs in Biomedicine · doi:10.1016/j.cmpb.2023.107550 · Cited by 38
Metadata
- Authors
- Kommiya Mothilal, R., Mahajan, D., Tan, C., & Sharma, A.
- Venue
- Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society
- Publisher
- ACM
- Published
- 2021-07-21
- Pages
- 652-663
- ISSN
- 3065-8365 (ISSN-L)
- OpenAlex
- W3105449709
- Cited by
- 76
- References
- 32
- FWCI
- 5.68
- Type
- conference paper
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.