Explaining deep learning models is essential for clinical integration of medical image analysis systems. A good explanation highlights if a model depends on spurious features that undermines generalization and harms a subset of patients or, conversely, may present novel biological insights. Although techniques like GradCAM can identify influential features, they are measurement tools that do not themselves form an explanation. We propose a human-machine-VLM interaction system tailored to explaining classifiers in computational pathology, including multi-instance learning for whole-slide images. Our proof of concept comprises (1) an AI-integrated slide viewer to run sliding-window experiments to test claims of an explanation, and (2) quantification of an explanation's predictiveness using general-purpose vision-language models. The results demonstrate that this allows us to qualitatively test claims of explanations and can quantifiably distinguish competing explanations. This offers a practical path from explainable AI to explained AI in digital pathology and beyond. Code and prompts are available at https://github.com/nki-ai/x2x.

Mechanistic Interpretability (MI) aims to understand neural networks through causal explanations. Though MI has many explanation-generating methods, progress has been limited by the lack of a universal approach to evaluating explanations. Here we analyse the fundamental question "What makes a good explanation?" We introduce a pluralist Explanatory Virtues Framework drawing on four perspectives from the Philosophy of Science - the Bayesian, Kuhnian, Deutschian, and Nomological - to systematically evaluate and improve explanations in MI. We find that Compact Proofs consider many explanatory virtues and are hence a promising approach. Fruitful research directions implied by our framework include (1) clearly defining explanatory simplicity, (2) focusing on unifying explanations and (3) deriving universal principles for neural networks. Improved MI methods enhance our ability to monitor, predict, and steer AI systems.

Proposals of artificial intelligence (AI) solutions based on increasingly complex and accurate predictive models are becoming ubiquitous across many disciplines. As the complexity of these models grows, transparency and users' understanding often diminish. This suggests that accurate prediction alone is insufficient for making an AI-based solution truly useful. In the development of healthcare systems, this introduces new issues related to accountability and safety. Understanding how and why an AI system makes a recommendation may require complex explanations of its inner workings and reasoning processes. Although research on explainable AI (XAI) has significantly increased in recent years and there is high demand for XAI in medicine, defining what constitutes a good explanation remains ad hoc, and providing adequate explanations continues to be challenging. To fully realize the potential of AI, it is critical to address two fundamental questions about explanations for safety-critical AI applications, such as health-AI: (1) What is an explanation in health-AI? and (2) What are the attributes of a good explanation in health-AI? In this study, we examined published literature and gathered expert opinions through a two-round Delphi study. The research outputs include (1) a definition of what constitutes an explanation in health-AI and (2) a comprehensive list of attributes that characterize a good explanation in health-AI.

Among the existing approaches for visual-based Recommender System (RS) explainability, utilizing user-uploaded item images as efficient, trustable explanations is a promising option. However, current models following this paradigm assume that, for any user, all images uploaded by other users can be considered negative training examples (i.e. bad explanatory images), an inadvertedly naive labelling assumption that contradicts the rationale of the approach. This work proposes a new explainer training pipeline by leveraging Positive-Unlabelled (PU) Learning techniques to train image-based explainer with refined subsets of reliable negative examples for each user selected through a novel user-personalized, two-step, similarity-based PU Learning algorithm. Computational experiments show this PU-based approach outperforms the state-of-the-art non-PU method in six popular real-world datasets, proving that an improvement of visual-based RS explainability can be achieved by maximizing training data quality rather than increasing model complexity.

Often, a good explanation for a program's unexpected behavior is a bug in the programmer's code. But sometimes, an even better explanation is a bug in the programmer's mental model of the language they are using. Instead of merely debugging our current code ("giving the programmer a fish"), what if our tools could directly debug our mental models ("teaching the programmer to fish")? In this paper, we apply ideas from computational cognitive science to do exactly that. Given a perplexing program, we use program synthesis techniques to automatically infer potential misconceptions that might cause the user to be surprised by the program's behavior. By analyzing these misconceptions, we provide succinct, useful explanations of the program's behavior. Our methods can even be inverted to synthesize pedagogical example programs for diagnosing and correcting misconceptions in students.

Instead of providing an explanation networks (ANNs), which are effective at approximating of a phenomenon, models trained this way present complex functions but often lack transparency us with yet another phenomenon that needs an explanation and explanatory power. It highlights the [Wiegreffe and Pinter, 2019; Jain and Wallace, 2019]. 'problem of induction'--the philosophical issue Thus, despite the recent surge in the field of'explainable that past observations may not necessarily predict AI' [Doshi-Velez and Kim, 2017], which attempts to provide future events, a challenge that ML models face some insight in to the generalizations made by trained models, when encountering new, unseen data. The paper argues it may be the case that the underlying problem of induction for the importance of not just making predictions and a lack of good explanations will remain so long as but also providing good explanations, a feature we use machine induction as the primary path in AI. that current models often fail to deliver.

YOLOv8 was launched on January 10th, 2023. And as of this moment, this is the state-of-the-art model for classification, detection, and segmentation tasks in the computer vision world. The model outperforms all known models both in terms of accuracy and execution time. The ultralytics team did a really good job in making this model easier to use compared to all the previous YOLO models -- you don't even have to clone the git repository anymore! In this post, I created a very simple example of all you need to do to train YOLOv8 on your data, specifically for a segmentation task.

Why Machines fail to reason? I find three of these aspects: An awareness of causal relationships, Good explanations for these causal relationships, The ability to make use of the former in order to achieve motivated goals. Causality One well-known mantra of #statistics is "correlation does not imply causation". It turns out that while there are well-defined statistical measures of correlation, causality is notoriously difficult to define, formalize, or quantify. Good explanations Good explanations are also the reason why we are so convinced that this reasoning can be vastly generalized: you know that you do not want to hide under a tree during a thunderstorm.

An important part of a data scientist's role is to explain model predictions. Often, the person receiving the explanation will be non-technical. If you start talking about cost functions, hyperparameters or p-values you will be met with blank stares. We need to translate these technical concepts into layman's terms. This process can be more challenging than building the model itself. We will explore how you can give human-friendly explanations. We will do this by discussing some key characteristics of a good explanation. The focus will be on explaining individual predictions.

Testing the best: There's only one way to figure that out: ask some users. So that's what researchers from Harvard and Google Brain did, in a series of studies. Test subjects looked at different combinations of inputs, outputs, and explanations around a machine learning algorithm that was designed to learn the dietary habits or medical conditions of an alien (Yes, seriously--alien life was chosen to avoid the test subject's own biases creeping in). Users then scored the different combinations. Keep it short: Longer explanations were found to be more difficult to parse than shorter ones--though breaking up the same amount of text into many short lines was somehow better than making people read a few longer lines.