Arjumand Younus

The upsurge of policies and guidelines that aim to ensure Artificial Intelligence (AI) systems are safe and trustworthy has led to a fragmented landscape of AI governance. The European Union (EU) is a key actor in the development of such policies and guidelines. Its High-Level Expert Group (HLEG) issued an influential set of guidelines for trustworthy AI, followed in 2024 by the adoption of the EU AI Act. While the EU policies and guidelines are expected to be aligned, they may differ in their scope, areas of emphasis, degrees of normativity, and priorities in relation to AI. To gain a broad understanding of AI governance from the EU perspective, we leverage qualitative thematic analysis approaches to uncover prevalent themes in key EU documents, including the AI Act and the HLEG Ethics Guidelines. We further employ quantitative topic modelling approaches, specifically through the use of the BERTopic model, to enhance the results and increase the document sample to include EU AI policy documents published post-2018. We present a novel perspective on EU policies, tracking the evolution of its approach to addressing AI governance.

The effectiveness of supervised machine learning models is heavily influenced by the quality of training data, which is often shaped by human annotators. Subjective NLP tasks such as hate speech detection, toxicity identification, and sexism classification frequently exhibit annotator disagreement due to differences in individual perspectives. This study investigates annotator disagreement in sexism detection using English tweets from the EXIST 2023 competition. To systematically analyse disagreement, tweets are categorised based on annotator consensus levels, examining how annotator demographics and linguistic features contribute to labelling inconsistencies. We interpret disagreement patterns using Shapley Additive Explanations (SHAP) and assess the consistency of SHAP-derived feature importance rankings via Spearman Rank Correlation. Our findings demonstrate that both annotator demographics and tweet characteristics significantly shape disagreement, reinforcing the need for perspectivist approaches in NLP by showing that annotator disagreement is not just noise but a meaningful signal that should be incorporated into dataset construction. Please be advised that this work contains examples of offensive content

Explainable AI (XAI) is increasingly used in clinical machine learning, yet quantitative evaluation of explanation quality is often reported inconsistently across methods and datasets. We present a reproducible, metric-driven framework for evaluating XAI methods on healthcare tabular data. The framework consolidates six established, family-specific metrics, fidelity, simplicity, consistency, robustness, precision, and coverage, into explicit equations; pairs them with a pre-specified focal-model protocol; and releases open-source code with a method–metric applicability map. We evaluate LIME, SHAP, Anchors, EBM, and TabNet across four public healthcare tabular datasets. Post-hoc explainers are applied to a single selected Random Forest focal predictor to control model-induced variability, whereas EBM and TabNet are assessed through their native interpretability mechanisms. Global explanation summaries are reported descriptively only. The results show that SHAP/TreeSHAP provides exact score reconstruction for the Random Forest setting, while LIME produces simpler but lower-fidelity explanations with greater instance-level variability. LIME and SHAP show the strongest rank agreement among the evaluated pairs, although agreement varies across datasets. TabNet often yields compact native explanations, but these must be interpreted alongside its dataset-specific predictive performance. EBM and TabNet show low sensitivity under the fixed Gaussian-jitter robustness protocol, while Anchors produces high-precision rules with reduced coverage at stricter thresholds. Overall, the framework enables controlled comparison under explicit method–metric and focal-model assumptions, supporting more transparent XAI selection for tabular machine learning. Although demonstrated in healthcare, the framework is transferable to other high-stakes tabular domains. Source code: https://github.com/matifq/XAI_Tab_Health .

The proliferation of hate speech on digital platforms has become a significant issue, and automated content moderation systems built on machine learning are a proposed solution. However, they face challenges in multilingual and low-resource settings due to the need for extensive labelled data. This paper introduces an explainable AI framework designed to identify annotation discrepancies in low-resource languages, focusing on Hindi, the third most-spoken language worldwide, which lacks comprehensive research in hate speech detection. By examining the labelling quality of the Hate speech and Offensive Content Identification in English and Indo-Aryan Languages (HASOC) challenge, we use unsupervised learning methods to extract topical variations and annotation behavior and apply these features in an explainable AI-based classification model, TabNet. We release a relabelled Hindi hate speech benchmark dataset with label-flipping information and related metadata to facilitate research in this area. The source code has also been released for reproducibility purposes. Please be advised that this work contains examples of toxic content

Reaching marginal and other migrant communities to elicit their political views and opinions is a well-known challenge. Social media has enabled a certain amount of online activism and participation, especially in societies with abundant multicultural identities. However, it can be quite challenging to isolate the voice of the migrant in English-speaking countries, especially with an abundance of content in English on social media. In this paper, we pursue a case study of Ireland’s Twitter landscape, specifically migrant and native activists. We present a methodology that can accurately (> 80%) isolate the Irish migrant voice with as little as 25 English tweets without relying on user metadata and using simple, highly explainable, out-of-the-box machine learning methods. Using this, we distil (via sentiment analysis) polarities of views, segment (via BERT-based topic modelling) and summarise (via ChatGPT) differentiated views in a consumable manner for policymakers. Our approach enables policymakers to further their understanding of multicultural communities and use this to inform their decision-making processes.