Research Papers

While a substantial amount of work has recently been devoted to enhance the performance of computational Authorship Identification (AId) systems, little to no attention has been paid to endowing AId systems with the ability to explain the reasons behind their predictions. This lacking substantially hinders the practical employment of AId methodologies, since the predictions returned by such systems are hardly useful unless they are supported with suitable explanations. In this paper, we explore the applicability of existing general-purpose eXplainable Artificial Intelligence (XAI) techniques to AId, with a special focus on explanations addressed to scholars working in cultural heritage. In particular, we assess the relative merits of three different types of XAI techniques (feature ranking, probing, factuals and counterfactual selection) on three different AId tasks (authorship attribution, authorship verification, same-authorship verification) by running experiments on real AId data. Our analysis shows that, while these techniques make important first steps towards explainable Authorship Identification, more work remains to be done in order to provide tools that can be profitably integrated in the workflows of scholars.

Despite the essential need for comprehensive considerations in responsible AI, factors like robustness, fairness, and causality are often studied in isolation. Adversarial perturbation, used to identify vulnerabilities in models, and individual fairness, aiming for equitable treatment of similar individuals, despite initial differences, both depend on metrics to generate comparable input data instances. Previous attempts to define such joint metrics often lack general assumptions about data or structural causal models and were unable to reflect counterfactual proximity. To address this, our paper introduces a causal fair metric formulated based on causal structures encompassing sensitive attributes and protected causal perturbation. To enhance the practicality of our metric, we propose metric learning as a method for metric estimation and deployment in real-world problems in the absence of structural causal models. We also demonstrate the application of our novel metric in classifiers. Empirical evaluation of real-world and synthetic datasets illustrates the effectiveness of our proposed metric in achieving an accurate classifier with fairness, resilience to adversarial perturbations, and a nuanced understanding of causal relationships.

Introducing the Attribution Stability Indicator: a Measure for Time Series XAI Attributions

Udo Schlegel, Daniel A. Keim

Given the increasing amount and general complexity of time series data in domains such as finance, weather forecasting, and healthcare, there is a growing need for state-of-the-art performance models that can provide interpretable insights into underlying patterns and relationships. Attribution techniques enable the extraction of explanations from time series models to gain insights but are hard to evaluate for their robustness and trustworthiness. We propose the Attribution Stability Indicator (ASI), a measure to incorporate robustness and trustworthiness as properties of attribution techniques for time series into account. We extend a perturbation analysis with correlations of the original time series to the perturbed instance and the attributions to include wanted properties in the measure. We demonstrate the wanted properties based on an analysis of the attributions in a dimension-reduced space and the ASI scores distribution over three whole time series classification datasets.

eXplainable Artificial Intelligence (XAI) has emerged as an essential requirement when dealing with mission-critical applications, ensuring transparency and interpretability of the employed black box AI models. The significance of XAI spans various domains, from healthcare to finance, where understanding the decision-making process of deep learning algorithms is essential. Most AI-based computer vision models are often black boxes; hence, providing explainability of deep neural networks in image processing is crucial for their wide adoption and deployment in medical image analysis, autonomous driving, and remote sensing applications. Recently, several XAI methods for image classification tasks have been introduced. On the contrary, image segmentation has received comparatively less attention in the context of explainability, although it is a fundamental task in computer vision applications, especially in remote sensing. Only some research proposes gradient-based XAI algorithms for image segmentation. This paper adapts the recent gradient-free Sobol XAI method for semantic segmentation. To measure the performance of the Sobol method for segmentation, we propose a quantitative XAI evaluation method based on a learnable noise model. The main objective of this model is to induce noise on the explanation maps, where higher induced noise signifies low accuracy and vice versa. A benchmark analysis is conducted to evaluate and compare performance of three XAI methods, including Seg-Grad-CAM, Seg-Grad-CAM++ and Seg-Sobol using the proposed noise-based evaluation technique. This constitutes the first attempt to run and evaluate XAI methods using high-resolution satellite images.

Towards Feasible Counterfactual Explanations: A Taxonomy Guided Template-based NLG Method

Pedram Salimi, Nirmalie Wiratunga, David Corsar +1

Counterfactual Explanations (cf-XAI) describe the smallest changes in feature values necessary to change an outcome from one class to another. However, many cf-XAI methods neglect the feasibility of those changes. In this paper, we introduce a novel approach for presenting cf-XAI in natural language (Natural-XAI), giving careful consideration to actionable and comprehensible aspects while remaining cognizant of immutability and ethical concerns. We present three contributions to this endeavor. Firstly, through a user study, we identify two types of themes present in cf-XAI composed by humans: content-related, focusing on how features and their values are included from both the counterfactual and the query perspectives; and structure-related, focusing on the structure and terminology used for describing necessary value changes. Secondly, we introduce a feature actionability taxonomy with four clearly defined categories, to streamline the explanation presentation process. Using insights from the user study and our taxonomy, we created a generalisable template-based natural language generation (NLG) method compatible with existing explainers like DICE, NICE, and DisCERN, to produce counterfactuals that address the aforementioned limitations of existing approaches. Finally, we conducted a second user study to assess the performance of our taxonomy-guided NLG templates on three domains. Our findings show that the taxonomy-guided Natural-XAI approach (n-XAI^T) received higher user ratings across all dimensions, with significantly improved results in the majority of the domains assessed for articulation, acceptability, feasibility, and sensitivity dimensions.

Artificial Intelligence techniques can be used to classify a patient's physical activities and predict vital signs for remote patient monitoring. Regression analysis based on non-linear models like deep learning models has limited explainability due to its black-box nature. This can require decision-makers to make blind leaps of faith based on non-linear model results, especially in healthcare applications. In non-invasive monitoring, patient data from tracking sensors and their predisposing clinical attributes act as input features for predicting future vital signs. Explaining the contributions of various features to the overall output of the monitoring application is critical for a clinician's decision-making. In this study, an Explainable AI for Quantitative analysis (QXAI) framework is proposed with post-hoc model explainability and intrinsic explainability for regression and classification tasks in a supervised learning approach. This was achieved by utilizing the Shapley values concept and incorporating attention mechanisms in deep learning models. We adopted the artificial neural networks (ANN) and attention-based Bidirectional LSTM (BiLSTM) models for the prediction of heart rate and classification of physical activities based on sensor data. The deep learning models achieved state-of-the-art results in both prediction and classification tasks. Global explanation and local explanation were conducted on input data to understand the feature contribution of various patient data. The proposed QXAI framework was evaluated using PPG-DaLiA data to predict heart rate and mobile health (MHEALTH) data to classify physical activities based on sensor data. Monte Carlo approximation was applied to the framework to overcome the time complexity and high computation power requirements required for Shapley value calculations.

Unsupervised discovery of Interpretable Visual Concepts

Caroline Mazini Rodrigues, Nicolas Boutry, Laurent Najman

Providing interpretability of deep-learning models to non-experts, while fundamental for a responsible real-world usage, is challenging. Attribution maps from xAI techniques, such as Integrated Gradients, are a typical example of a visualization technique containing a high level of information, but with difficult interpretation. In this paper, we propose two methods, Maximum Activation Groups Extraction (MAGE) and Multiscale Interpretable Visualization (Ms-IV), to explain the model's decision, enhancing global interpretability. MAGE finds, for a given CNN, combinations of features which, globally, form a semantic meaning, that we call concepts. We group these similar feature patterns by clustering in ``concepts'', that we visualize through Ms-IV. This last method is inspired by Occlusion and Sensitivity analysis (incorporating causality), and uses a novel metric, called Class-aware Order Correlation (CaOC), to globally evaluate the most important image regions according to the model's decision space. We compare our approach to xAI methods such as LIME and Integrated Gradients. Experimental results evince the Ms-IV higher localization and faithfulness values. Finally, qualitative evaluation of combined MAGE and Ms-IV demonstrates humans' ability to agree, based on the visualization, with the decision of clusters' concepts; and, to detect, among a given set of networks, the existence of bias.

The generation of natural language text from data series gained renewed interest among AI research goals. Not surprisingly, the few proposals in the state of the art are based on training some system, in order to produce a text that describes and that is coherent to the data provided as input. Main challenges of such approaches are the proper identification of "what" to say (the key descriptive elements to be addressed in the data) and "how" to say: the correspondence and accuracy between data and text, the presence of contradictions/redundancy in the text, the control of the amount of synthesis. This paper presents a framework that is compliant with xAI requirements. In particular we model ASP/Python programs that enable an explicit control of accuracy errors and amount of synthesis, with proven optimal solutions. The text description is hierarchically organized, in a top-down structure where text is enriched with further details, according to logic rules. The generation of natural language descriptions' structure is also managed by logic rules.

TRIVEA: Transparent Ranking Interpretation using Visual Explanation of Black-Box Algorithmic Rankers

Jun Yuan, Kaustav Bhattacharjee, Akm Zahirul Islam +1

Ranking schemes drive many real-world decisions, like, where to study, whom to hire, what to buy, etc. Many of these decisions often come with high consequences. For example, a university can be deemed less prestigious if not featured in a top-k list, and consumers might not even explore products that do not get recommended to buyers. At the heart of most of these decisions are opaque ranking schemes, which dictate the ordering of data entities, but their internal logic is inaccessible or proprietary. Drawing inferences about the ranking differences is like a guessing game to the stakeholders, like, the rankees (i.e., the entities who are ranked, like product companies) and the decision-makers (i.e., who use the rankings, like buyers). In this paper, we aim to enable transparency in ranking interpretation by using algorithmic rankers that learn from available data and by enabling human reasoning about the learned ranking differences using explainable AI (XAI) methods. To realize this aim, we leverage the exploration-explanation paradigm of human-data interaction to let human stakeholders explore subsets and groupings of complex multi-attribute ranking data using visual explanations of model fit and attribute influence on rankings. We realize this explanation paradigm for transparent ranking interpretation in TRIVEA, a visual analytic system that is fueled by: i) visualizations of model fit derived from algorithmic rankers that learn the associations between attributes and rankings from available data and ii) visual explanations derived from XAI methods that help abstract important patterns, like, the relative influence of attributes in different ranking ranges. Using TRIVEA, end users not trained in data science have the agency to transparently reason about the global and local behavior of the rankings without the need to open black-box ranking models and develop confidence in the resulting attribute-based inferences. We demonstrate the efficacy of TRIVEA using multiple usage scenarios and subjective feedback from researchers with diverse domain expertise. Keywords: Visual Analytics, Learning-to-Rank, Explainable ML, Ranking

Amidst the rapid expansion of Machine Learning (ML) and Large Language Models (LLMs), understanding the semantics within their mechanisms is vital. Causal analyses define semantics, while gradient-based methods are essential to eXplainable AI (XAI), interpreting the model's 'black box'. Integrating these, we investigate how a model's mechanisms reveal its causal effect on evidence-based decision-making. Research indicates intersectionality - the combined impact of an individual's demographics - can be framed as an Average Treatment Effect (ATE). This paper demonstrates that hateful meme detection can be viewed as an ATE estimation using intersectionality principles, and summarized gradient-based attention scores highlight distinct behaviors of three Transformer models. We further reveal that LLM Llama-2 can discern the intersectional aspects of the detection through in-context learning and that the learning process could be explained via meta-gradient, a secondary form of gradient. In conclusion, this work furthers the dialogue on Causality and XAI. Our code is available online (see External Resources section).

The Rashōmon effect poses challenges for deriving reliable knowledge from machine learning models. This study examined the influence of sample size on explanations from models in a Rashōmon set using SHAP. Experiments on 5 public datasets showed that explanations gradually converged as the sample size increased. Explanations from <128 samples exhibited high variability, limiting reliable knowledge extraction. However, agreement between models improved with more data, allowing for consensus. Bagging ensembles often had higher agreement. The results provide guidance on sufficient data to trust explanations. Variability at low samples suggests that conclusions may be unreliable without validation. Further work is needed with more model types, data domains, and explanation methods. Testing convergence in neural networks and with model-specific explanation methods would be impactful. The approaches explored here point towards principled techniques for eliciting knowledge from ambiguous models.

Feature attribution methods have become a staple method to disentangle the complex behavior of black box models. Despite their success, some scholars have argued that such methods suffer from a serious flaw: they do not allow a reliable interpretation in terms of human concepts. Simply put, visualizing an array of feature contributions is not enough for humans to conclude something about a model's internal representations, and confirmation bias can trick users into false beliefs about model behavior. We argue that a structured approach is required to test whether our hypotheses on the model are confirmed by the feature attributions. This is what we call the "semantic match" between human concepts and (sub-symbolic) explanations. Building on the conceptual framework put forward in Cinà et al. [2023], we propose a structured approach to evaluate semantic match in practice. We showcase the procedure in a suite of experiments spanning tabular and image data, and show how the assessment of semantic match can give insight into both desirable (e.g., focusing on an object relevant for prediction) and undesirable model behaviors (e.g., focusing on a spurious correlation). We couple our experimental results with an analysis on the metrics to measure semantic match, and argue that this approach constitutes the first step towards resolving the issue of confirmation bias in XAI.