{
  "id": "2106.02597",
  "version": "v1",
  "published": "2021-06-04T16:54:36.000Z",
  "updated": "2021-06-04T16:54:36.000Z",
  "title": "Model-agnostic and Scalable Counterfactual Explanations via Reinforcement Learning",
  "authors": [
    "Robert-Florian Samoilescu",
    "Arnaud Van Looveren",
    "Janis Klaise"
  ],
  "comment": "18 pages",
  "categories": [
    "cs.LG",
    "stat.ML"
  ],
  "abstract": "Counterfactual instances are a powerful tool to obtain valuable insights into automated decision processes, describing the necessary minimal changes in the input space to alter the prediction towards a desired target. Most previous approaches require a separate, computationally expensive optimization procedure per instance, making them impractical for both large amounts of data and high-dimensional data. Moreover, these methods are often restricted to certain subclasses of machine learning models (e.g. differentiable or tree-based models). In this work, we propose a deep reinforcement learning approach that transforms the optimization procedure into an end-to-end learnable process, allowing us to generate batches of counterfactual instances in a single forward pass. Our experiments on real-world data show that our method i) is model-agnostic (does not assume differentiability), relying only on feedback from model predictions; ii) allows for generating target-conditional counterfactual instances; iii) allows for flexible feature range constraints for numerical and categorical attributes, including the immutability of protected features (e.g. gender, race); iv) is easily extended to other data modalities such as images.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2021-06-04T16:54:36.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "scalable counterfactual explanations",
      "model-agnostic",
      "optimization procedure",
      "necessary minimal changes",
      "deep reinforcement learning approach"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 18,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}