Existing algorithms for explaining the output of image classifiers use different definitions of explanations and a variety of techniques to find them. However, none of the existing tools use a principled approach based on formal definitions of cause and explanation. In this paper we present a novel black-box approach to computing explanations grounded in the theory of actual causality. We prove relevant theoretical results and present an algorithm for computing approximate explanations based on these definitions. We prove termination of our algorithm and discuss its complexity and the amount of approximation compared to the precise definition. We implemented the framework in a tool, ReX, and we present experimental results and a comparison with state-of-the-art tools. We demonstrate that ReX is the most efficient black-box tool and produces the smallest explanations, in addition to outperforming other black-box tools on standard quality measures.

Existing explanation tools for image classifiers usually give only a single explanation for an image’s classification. For many images, however, image classifiers accept more than one explanation for the image label. These explanations are useful for analyzing the decision process of the classifier and for detecting errors. Thus, restricting the number of explanations to just one severely limits insight into the behavior of the classifier. In this paper, we describe an algorithm and a tool, MultiReX, for computing multiple explanations as the output of a black-box image classifier for a given image. Our algorithm uses a principled approach based on actual causality. We analyze its theoretical complexity and evaluate MultiReX against the state-of-the-art across three different models and three different datasets. We find that MultiReX finds more explanations and that these explanations are of higher quality.

Using AI models in healthcare is gaining popularity. To improve clinician confidence in the results of automated triage and to provide further information about the suggested diagnosis, an explanation produced by a separate post-hoc explainability tool often accompanies the classification of an AI model. If no abnormalities are detected, however, it is not clear what an explanation should be. A human clinician might be able to describe certain salient features of tumors that are not in scan, but existing Explainable AI tools cannot do that, as they cannot point to features that are absent from the input. In this paper, we present a definition of and algorithm for providing explanations of absence; that is, explanations of negative classifications in the context of healthcare AI. Our approach is rooted in the concept of explanations in actual causality. It uses the model as a black-box and is hence portable and works with proprietary models. Moreover, the computation is done in the preprocessing stage, based on the model and the dataset. During the execution, the algorithm only projects the precomputed explanation template on the current image. We implemented this approach in a tool, nito, and trialed it on a number of medical datasets to demonstrate its utility on the classification of solid tumors. We discuss the differences between the theoretical approach and the implementation in the domain of classifying solid tumors and address the additional complications posed by this domain. Finally, we discuss the assumptions we make in our algorithm and its possible extensions to explanations of absence for general image classifiers.

Strings are universal containers: they are flexible to use, abundant in code, and difficult to test. String-controlled programs are programs that make branching decisions based on string input. Automatically generating valid test inputs for these programs considering only character sequences rather than any underlying string-encoded structures, can be prohibitively expensive. We present June, a tool that enables Java developers to expose any present latent string structure to test generation tools. June is an annotation-driven testability transformation and an extensible library, JuneLib, of structured string definitions. The core JuneLib definitions are empirically derived and provide templates for all structured strings in our test set. June takes lightly annotated source code and injects code that permits an automated test generator (ATG) to focus on the creation of mutable substrings inside a structured string. Using June costs the developer little, with an average of 2.1 annotations per string-controlled class. June uses standard Java build tools and therefore deploys seamlessly within a Java project. By feeding string structure information to an ATG tool, June dramatically reduces wasted effort; branches are effortlessly covered that would otherwise be extremely difficult, or impossible, to cover. This waste reduction both increases and speeds coverage. EvoSuite, for example, achieves the same coverage on June-ed classes in 1 minute, on average, as it does in 9 minutes on the un-June-ed class. These gains increase over time. On our corpus, June-ing a program compresses 24 hours of execution time into ca. 2 hours. We show that many ATG tools can reuse the same June-ed code: a few June annotations, a one-off cost, benefit many different testing regimes.