# Scientific Discovery by Generating Counterfactuals using Image Translation

arXiv: https://arxiv.org/abs/2007.05500 · HTML: https://mlsanity.com/paper/2007.05500/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 10 Jul 2020.

## Built on by (most cited first, 11 of 11)

- [Explaining in Style: Training a GAN to explain a classifier in StyleSpace](https://mlsanity.com/paper/2104.13369/) (2021)
- [Transparency of Deep Neural Networks for Medical Image Analysis: A Review of Interpretability Methods](https://mlsanity.com/paper/2111.02398/) (2021)
- [ICAM-reg: Interpretable Classification and Regression with Feature Attribution for Mapping Neurological Phenotypes in Individual Scans](https://mlsanity.com/paper/2103.02561/) (2021)
- [Counterfactual Explanation of Brain Activity Classifiers using Image-to-Image Transfer by Generative Adversarial Network](https://mlsanity.com/paper/2110.14927/) (2021)
- [Using generative AI to investigate medical imagery models and datasets](https://mlsanity.com/paper/2306.00985/) (2023)
- [GLANCE: Global to Local Architecture-Neutral Concept-based Explanations](https://mlsanity.com/paper/2207.01917/) (2022)
- [Leveraging Conditional Generative Models in a General Explanation Framework of Classifier Decisions](https://mlsanity.com/paper/2106.10947/) (2021)
- [Discriminative Attribution from Counterfactuals](https://mlsanity.com/paper/2109.13412/) (2021)
- [Explaining Image Classification with Visual Debates](https://mlsanity.com/paper/2210.09015/) (2022)
- [Counterfactual Image Generation for adversarially robust and interpretable Classifiers](https://mlsanity.com/paper/2310.00761/) (2023)
- [Object-Centric Case-Based Reasoning via Argumentation](https://mlsanity.com/paper/2510.00185/) (2025)

## Abstract (arXiv)

Model explanation techniques play a critical role in understanding the source of a model's performance and making its decisions transparent. Here we investigate if explanation techniques can also be used as a mechanism for scientific discovery. We make three contributions: first, we propose a framework to convert predictions from explanation techniques to a mechanism of discovery. Second, we show how generative models in combination with black-box predictors can be used to generate hypotheses (without human priors) that can be critically examined. Third, with these techniques we study classification models for retinal images predicting Diabetic Macular Edema (DME), where recent work showed that a CNN trained on these images is likely learning novel features in the image. We demonstrate that the proposed framework is able to explain the underlying scientific mechanism, thus bridging the gap between the model's performance and human understanding.