Decisions made by legal adjudicators and administrative decision-makers often found upon a reservoir of stored experiences, from which is drawn a tacit body of expert knowledge. Such expertise may be implicit and opaque, even to the decision-makers themselves, and generates obstacles when implementing AI for automated decision-making tasks within the legal field, since, to the extent that AI-powered decision-making tools must found upon a stock of domain expertise, opacities may proliferate. This raises particular issues within the legal domain, which requires a high level of accountability, thus transparency. This requires enhanced explainability, which entails that a heterogeneous body of stakeholders understand the mechanism underlying the algorithm to the extent that an explanation can be furnished. However, the “black-box” nature of some AI variants, such as deep learning, remains unresolved, and many machine decisions therefore remain poorly understood. This survey paper, based upon a unique interdisciplinary collaboration between legal and AI experts, provides a review of the explainability spectrum, as informed by a systematic survey of relevant research papers, and categorises the results. The article establishes a novel taxonomy, linking the differing forms of legal inference at play within particular legal sub-domains to specific forms of algorithmic decision-making. The diverse categories demonstrate different dimensions in explainable AI (XAI) research. Thus, the survey departs from the preceding monolithic approach to legal reasoning and decision-making by incorporating heterogeneity in legal logics: a feature which requires elaboration, and should be accounted for when designing AI-driven decision-making systems for the legal field. It is thereby hoped that administrative decision-makers, court adjudicators, researchers, and practitioners can gain unique insights into explainability, and utilise the survey as the basis for further research within the field.

Explainable AI (XAI) aids in deciphering 'black-box' models. While several methods have been proposed and evaluated primarily in the image domain, the exploration of explainability in the text domain remains a growing research area. In this paper, we delve into the applicability of XAI methods for the text domain. In this context, the 'Similarity Difference and Uniqueness' (SIDU) XAI method, recognized for its superior capability in localizing entire salient regions in image-based classification is extended to textual data. The extended method, SIDU-TXT, utilizes feature activation maps from 'black-box' models to generate heatmaps at a granular, word-based level, thereby providing explanations that highlight contextually significant textual elements crucial for model predictions. Given the absence of a unified standard for assessing XAI methods, this study applies a holistic three-tiered comprehensive evaluation framework: Functionally-Grounded, Human-Grounded and Application-Grounded, to assess the effectiveness of the proposed SIDU-TXT across various experiments. We find that, in sentiment analysis task of a movie review dataset, SIDU-TXT excels in both functionally and human-grounded evaluations, demonstrating superior performance through quantitative and qualitative analyses compared to benchmarks like Grad-CAM and LIME. In the application-grounded evaluation within the sensitive and complex legal domain of asylum decision-making, SIDU-TXT and Grad-CAM demonstrate comparable performances, each with its own set of strengths and weaknesses. However, both methods fall short of entirely fulfilling the sophisticated criteria of expert expectations, highlighting the imperative need for additional research in XAI methods suitable for such domains.

Modèles d'attention visuelle pour l'analyse de scènes dynamiques: De nombreuses applications de la vision par ordinateur requièrent la détection, la localisation et le suivi de régions ou d’objets d’intérêt dans une image ou une séquence d’images. De nombreux modèles d’attention visuelle, inspirés de la vision humaine, qui détectent de manière automatique les régions d’intérêt dans une image ou une vidéo, ont récemment été développés et utilisés avec succès dans différentes applications. Néanmoins, la plupart des approches existantes sont limitées à l’analyse de scènes statiques et très peu de méthodes exploitent la nature temporelle des séquences d’images.L'objectif principal de ce travail de thèse est donc l'étude de modèles d'attention visuelle pour l'analyse de scènes dynamiques complexes. Une carte de saliance est habituellement obtenue par la fusion d'une carte statitque (saliance spatiale dans une image) d'une part, et d'une carte dynamique (salience temporelle entre une série d'image) d'autre part. Dans notre travail, nous modélisons les changements dynamiques par un opérateur de texture LBP-TOP (Local Binary Patterns) et nous utilisons l'information couleur pour l'aspect spatial.Les deux cartes de saliances sont calculées en utilisant une formulation discriminante inspirée du système visuel humain, et fuionnées de manière appropriée en une carte de saliance spatio-temporelle.De nombreuses expériences avec des bases de données publiques, montrent que notre approche obteint des résulats meilleurs ou comparables avec les approches de la littérature.