The field of explainable artificial intelligence emerged in response to the growing need for more transparent and reliable models. However, using raw features to provide explanations has been discussed in several works lately, advocating for more user-understandable explanations. To address this issue, a wide range of papers proposing Concept-based eXplainable Artificial Intelligence (C-XAI) methods have been published in recent years. Nevertheless, a unified categorization and precise field definition are still missing. This paper fills the gap by offering a thorough review of C-XAI approaches. We identify and define different concepts and explanation types. We propose a taxonomy comprising nine categories and guidelines for selecting a suitable category based on the application context. Additionally, we discuss common evaluation strategies including metrics, human evaluations, and datasets employed, aiming to assist the development of future methods. Overall, we believe this survey will assist researchers, practitioners, and domain experts in enhancing their understanding and contributing to the progress of this innovative field.

Graph drawing techniques have been developed in the last few years with the purpose of producing esthetically pleasing node-link layouts. Recently, the employment of differentiable loss functions has paved the road to the massive usage of gradient descent and related optimization algorithms. In this article, we propose a novel framework for the development of Graph Neural Drawers (GNDs), machines that rely on neural computation for constructing efficient and complex maps. GND is Graph Neural Networks (GNNs) whose learning process can be driven by any provided loss function, such as the ones commonly employed in Graph Drawing. Moreover, we prove that this mechanism can be guided by loss functions computed by means of feedforward neural networks, on the basis of supervision hints that express beauty properties, like the minimization of crossing edges. In this context, we show that GNNs can nicely be enriched by positional features to deal also with unlabeled vertexes. We provide a proof-of-concept by constructing a loss function for the edge crossing and provide quantitative and qualitative comparisons among different GNN models working under the proposed framework.

Adversarial attacks on machine learning-based classifiers, along with defense mechanisms, have been widely studied in the context of single-label classification problems. In this paper, we shift the attention to multi-label classification, where the availability of domain knowledge on the relationships among the considered classes may offer a natural way to spot incoherent predictions, i.e., predictions associated to adversarial examples lying outside of the training data distribution. We explore this intuition in a framework in which first-order logic knowledge is converted into constraints and injected into a semi-supervised learning problem. Within this setting, the constrained classifier learns to fulfill the domain knowledge over the marginal distribution, and can naturally reject samples with incoherent predictions. Even though our method does not exploit any knowledge of attacks during training, our experimental analysis surprisingly unveils that domain-knowledge constraints can help detect adversarial examples effectively, especially if such constraints are not known to the attacker. We show how to implement an adaptive attack exploiting knowledge of the constraints and, in a specifically-designed setting, we provide experimental comparisons with popular state-of-the-art attacks. We believe that our approach may provide a significant step towards designing more robust multi-label classifiers.

Resistance spot welding is widely adopted in manufacturing and is characterized by high reliability and simple automation in the production line. The detection of defective welds is a difficult task that requires either destructive or expensive and slow non-destructive testing (e.g., ultrasound). The robots performing the welding automatically collect contextual and process-specific data. In this paper, we test whether these data can be used to predict defective welds. To do so, we use a dataset collected in a real industrial plant that describes welding-related data labeled with ultrasonic quality checks. We use these data to develop several pipelines based on shallow and deep learning machine learning algorithms and test the performance of these pipelines in predicting defective welds. Our results show that, despite the development of different pipelines and complex models, the machine-learning-based defect detection algorithms achieve limited performance. Using a qualitative analysis of model predictions, we show that correct predictions are often a consequence of inherent biases and intrinsic limitations in the data. We therefore conclude that the automatically collected data have limitations that hamper fault detection in a running production plant.