Abstract Digital skills are crucial for career adaptability and long-term labour market outcomes, yet identifying interventions that deliver sustained benefits while ensuring equity remains challenging. To address this, we propose CausalSkillNet-RL, a fairness-aware causal reinforcement learning framework for personalised intervention recommendations. The framework combines for the first time (i) a novel fairness-aware reinforcement learning agent that optimises both immediate uplift and long-term skill growth while enforcing equity for protected groups; and (ii) a fairness-regularised DragonNet for principled estimation of heterogeneous treatment effects, to provide robust causal signals for policy learning. We evaluate CausalSkillNet-RL firstly on synthetic and real-world datasets (Stack OverFlow Survey, ClickStart), using heterogeneous effect metrics, ablation studies, and visualisation. Results show that it achieves higher average treatment effects, improved Uplift@K, and reduced fairness gaps compared to state-of-the-art baselines. These findings highlight the framework’s potential to support equitable, data-driven long-term interventions in digital skill development.

The Health Mention Classification (HMC) task plays a pivotal role in leveraging social media discourse for public health mention monitoring, particularly in identifying and tracking disease proliferation. Despite its potential, the task poses significant challenges, due to the nuanced nature of health-related discussion. Building upon recent insights that emotional context can enhance HMC performance, in this paper, we explore how affective information can be integrated into the HMC process. Our study pioneers two distinct methodological pipelines that are designed to embed emotional nuances into the HMC task: (1) a two-stage fine-tuning process, starting with an implicit affective knowledge injection to initialise the model, followed by intermediate HMC task fine-tuning; and (2) an explicit multi-feature fusion strategy to leverage affective knowledge for HMC prediction. We conducted comprehensive evaluations across five diverse HMC benchmark datasets, encompassing content from Twitter, Reddit, and a blend of other social media platforms. Our empirical findings reveal that our affective-enriched models achieve statistically significant improvements across HMC benchmarks. Notably, the explicit multi-feature fusion method yielded a minimum of 3% improvement in F1 score over established BERT baselines across all tested corpora. Intriguingly, our analysis also suggests that the exclusive consideration of negative emotional indicators does not detrimentally impact HMC efficacy compared to leveraging both positive and negative emotions. Furthermore, our affectiveness-aware models demonstrate promising utility as a viable alternative in scenarios where domain-specific HMC datasets are scarce or non-existent for fine-tuning purposes. The consistent performance uplift across datasets sourced from varied social media channels underscores the generalisability and resilience of our proposed framework, marking a significant step forward in the computational understanding of health-related conversations in the digital sphere.

During the COVID-19 pandemic, the Church closed its physical doors for the first time in about 800 years, which is, arguably, a cataclysmic event. Other religions have found themselves in a similar situation, and they were practically forced to move online, which is an unprecedented occasion. In this paper, we analyse this sudden change in religious activities twofold: we create and deliver a questionnaire, as well as analyse Twitter data, to understand people’s perceptions and activities related to religious activities online. Importantly, we also analyse the temporal variations in this process, by analysing a period of 3 months: July-September 2020. Additionally to the separate analysis of the two data sources, we also discuss the implications from triangulating the results.

Integrating transformers with graph representation learning has emerged as a research focal point. However, recent studies showed that positional encoding in Transformers does not capture enough structural information between nodes. Additionally, existing graph neural network (GNN) models face the oversquashing issue, impeding information retention from distant nodes. To address, we transform graphs into regular structures, such as tokens, to enhance positional understanding and leverage transformer strengths. Inspired by the visual transformer (ViT) model, we propose partitioning graphs into patches and apply GNN models obtain fixed size vectors. Notably, our approach adopts contrastive learning for in-depth graph structure and incorporate more topological information via Ricci curvature to alleviate over-squashing problem by attenuating the effects of negatively curved edges while preserving the original graph structure. Unlike existing graph rewiring methods that directly modify graph structure by adding or removing edges, this approach is potentially more suitable for applications such as molecular learning where structural preservation is important. Our innovative pipeline subsequently introduces the PerformerMixer, a transformer variant with linear complexity, ensuring efficient computation. Evaluations on real-world benchmarks demonstrate our framework’s superior performance, like Peptides-func and achieve 3-WL expressiveness.

People often use disease or symptom terms on social media and online forums in ways other than to describe their health. Thus the NLP health mention classification (HMC) task aims to identify posts where users are discussing health conditions literally, not figuratively. Existing computational research typically only studies health mentions within well-represented groups in developed nations. Developing countries with limited health surveillance abilities fail to benefit from such data to manage public health crises. To advance the HMC research and benefit more diverse populations, we present the Nairaland health mention dataset (NHMD), a new dataset collected from a dedicated web forum for Nigerians. NHMD consists of 7,763 manually labelled posts extracted based on four prevalent diseases (HIV/AIDS, Malaria, Stroke and Tuberculosis) in Nigeria. With NHMD, we conduct extensive experiments using current state-of-the-art models for HMC and identify that, compared to existing public datasets, NHMD contains out-of-distribution examples. Hence, it is well suited for domain adaptation studies. The introduction of the NHMD dataset imposes better diversity coverage of vulnerable populations and generalisation for HMC tasks in a global public health surveillance setting. Additionally, we present a novel multi-task learning approach for HMC tasks by combining literal word meaning prediction as an auxiliary task. Experimental results demonstrate that the proposed approach outperforms state-of-the-art methods statistically significantly (p < 0.01, Wilcoxon test) in terms of F1 score over the state-of-the-art and shows that our new dataset poses a strong challenge to the existing HMC methods.

Medical visual question answering (Med-VQA) is to answer medical questions based on clinical images provided. This field is still in its infancy due to the complexity of the trio formed of questions, multimodal features and expert knowledge. In this paper, we tackle, a ’myth’ in the Natural Language Processing area - that unimodal bias is always considered undesirable in learning models. Additionally, we study the effect of integrating a novel dynamic attention mechanism into such models, inspired by a recent graph deep learning study.Unlike traditional attention, dynamic attention scores are conditioned on different query words in a question and thus enhance the representation learning ability of texts. We propose that some questions are answered more accurately with a reinforcement of question embedding after fusing multimodal features. Extensive experiments have been implemented on the VQA-RAD datasets and demonstrate that our proposed model, reinforCe unimOdal dynamiC Attention (COCA), outperforms the state-of-the-art methods overall and performs competitively at open-ended question answering.

Recommendation systems aim to assist users to discover most preferred contents from an ever-growing corpus of items. Although recommenders have been greatly improved by deep learning, they still face several challenges: (1) Behaviors are much more com- plex than words in sentences, so traditional attentive and recurrent models have limitations capturing the temporal dynamics of user preferences. (2) The preferences of users are multiple and evolving, so it is difficult to integrate long-term memory and short-term intent.

Modelling user voting intention in social media is an important research area, with applications in analysing electorate behaviour, online political campaigning and advertising. Previous approaches mainly focus on predicting national general elections, which are regularly scheduled and where data of past results and opinion polls are available. However, there is no evidence of how such models would perform during a sudden vote under time-constrained circumstances. That poses a more challenging task compared to traditional elections, due to its spontaneous nature. In this paper, we focus on the 2015 Greek bailout referendum, aiming to nowcast on a daily basis the voting intention of 2,197 Twitter users. We propose a semi-supervised multiple convolution kernel learning approach, leveraging temporally sensitive text and network information. Our evaluation under a real-time simulation framework demonstrates the effectiveness and robustness of our approach against competitive baselines, achieving a significant 20% increase in F-score compared to solely text-based models.