Preserving Causal Constraints in Counterfactual Explanations for Machine Learning Classifiers

By Divyat Mahajan, Chenhao Tan, Amit Sharma. 2019

240 citations recorded by the source corpus.

To construct interpretable explanations that are consistent with the original ML model, counterfactual examples---showing how the model's output changes with small perturbations to the input---have been proposed. This paper extends the work in counterfactual explanations by addressing the challenge of feasibility of such examples. For explanations of ML models in critical domains such as healthcare and finance, counterfactual examples are useful for an end-user only to the extent that perturbation of feature inputs is feasible in the real world. We formulate the problem of feasibility as preserving causal relationships among input features and present a method that uses (partial) structural causal models to generate actionable counterfactuals. When feasibility constraints cannot be easily expressed, we consider an alternative mechanism where people can label generated CF examples on feasibility: whether it is feasible to intervene and realize the candidate CF example from the original input. To learn from this labelled feasibility data, we propose a modified variational auto encoder loss for generating CF examples that optimizes for feasibility as people interact with its output. Our experiments on Bayesian networks and the widely used ''Adult-Income'' dataset show that our proposed methods can generate counterfactual explanations that better satisfy feasibility constraints than existing methods.. Code repository can be accessed here: \textit{this https URL}

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

On the Fairness of Causal Algorithmic Recourse (2020, background): “…a decision-making system, to overcome their adverse situation (Joshi et al. 2019; Ustun, Spangher, and Liu 2019; Sharma, Henderson, and Ghosh 2019; Mahajan, Tan, and Sharma 2019; Mothilal, Sharma, and Tan 2020; Venkatasubramanian and Alfano 2020; Karimi et al. 2020c,b; Karimi, Schölkopf, and…”
Algorithmic recourse under imperfect causal knowledge: a probabilistic approach (2020, background): “…x CE for an individual x F is given by a solution to the following optimisation problem: where dist ( · , · ) is a similarity metric on X , and additional constraints may be added to reﬂect plausibility, feasibility, or diversity of the obtained counterfactual explanations [15, 16, 24, 25, 31, 35].”
A survey of algorithmic recourse: definitions, formulations, solutions, and prospects (2020, background): “Relatedly, rather than solving recourse for each individual independently, some works pursue a model-based approach, whereby a mapping from factual to counterfactual instances is learned through gradient optimization [117, 143].”
Two Causal Principles for Improving Visual Dialog (2019, background): “Recently, some works [24, 6, 23, 34, 40, 45] introduced causal inference into machine learning, trying to endow models the abilities of pursuing the causeeffect.”
Counterfactual Explanations and Algorithmic Recourses for Machine Learning: A Review (2020, unknown): “[210] evaluated their algorithm on this metric.”
Algorithmic Recourse: from Counterfactual Explanations to Interventions (2020, background): “Finally, in a world where only additive interventions are allowed, the work of Mahajan et al. (2019), introduced independently of ours, may assist in ﬁnding causally consistent counterfactual explanations.”
A Survey on the Explainability of Supervised Machine Learning (2020, methodology): “The approach developed by Mahajan et al. (2019) presents a method that uses (partial) structural causal models to generate actionable counterfactuals.”
Counterfactual Explanations for Machine Learning: A Review (2020, methodology): “factuals (generally diverse by some metric of diversity) for a single input datapoint, therefore they require to be run once per input to get several counterfactuals [26, 47, 57, 64, 82, 85, 95, 98].”
Towards Robust and Reliable Algorithmic Recourse (2021, background): “For instance, it would be interesting to extend our framework to incorporate causal constraints (Karimi et al., 2020b; Mahajan et al., 2019).”
Deep Neural Networks and Tabular Data: A Survey (2021, background): “[182] demonstrate how to encode various feasibility constraints into such models.”