Fully decentralised federated learning enables collaborative model training among edge devices without relying on a central coordinator, thereby avoiding single points of failure and supporting spontaneous collaboration in pervasive environments. However, the absence of coordination introduces challenges that go beyond data heterogeneity alone. In realistic decentralised settings, devices often start from different model initializations, possess limited and non-IID local data, and interact over unstructured communication graphs, making naive parameter averaging ineffective and potentially destructive. In this paper, we address decentralised learning under combined data and initial model heterogeneity by proposing DecDiff+VT, a coordination-free decentralised learning algorithm specifically designed for such environments. DecDiff+VT integrates two complementary mechanisms: DecDiff, a disruption-aware aggregation strategy that updates local models toward their neighborhood average with a magnitude inversely proportional to model disagreement, and a lightweight virtual teacher (VT) mechanism based on soft-label regularization to improve local generalization in the absence of strong or centralized teacher models. Extensive experiments on image classification and activity recognition benchmarks (MNIST, Fashion-MNIST, EMNIST, CIFAR-10, and UCI-HAR) show that DecDiff+VT consistently outperforms or matches state-of-the-art decentralised baselines, achieving faster convergence, improved generalization, and greater robustness to overfitting, without incurring additional communication or memory overhead compared to standard decentralised averaging.

Cannabis use is on the rise, driven by relaxing legal regulations and declining perceptions of harm. This trend, coupled with the increasing reliance on social media for health-related information, has sparked interest in cannabis use during pregnancy (CanPreg). This study examines online discourse about CanPreg on Twitter, analyzing 53,183 unique tweets from 32,744 users in the USA and Canada between 2012 and 2021. We investigate the spatio-temporal distribution of CanPreg discussions, key topical contexts within these conversations, and their correlations with socioeconomic and health indicators. The analysis reveals regional differences, with a relatively higher interest in CanPreg discussions in Canada compared to the USA. The online discourse is primarily focused on research, alongside criticism, personal experiences, queries, news sharing, and advertisements. Additionally, correlations between CanPreg tweet activity, poverty rates, and mental health metrics suggest a connection between online discussions and real-world behaviors. This study highlights the role of social media in health communication and provides insights to inform targeted intervention strategies.

Vaccinations are fundamental public health interventions. Yet, inequalities in vaccine uptake across socioeconomic groups can significantly undermine their impact. Moreover, heterogeneities in vaccination coverage across socioeconomic strata are typically neglected by epidemic models and considered, if at all, only at posteriori. This limitation reduces their ability to predict and assess the effectiveness of vaccination campaigns. Here, we study the impact of socioeconomic inequalities in vaccination uptake on disease burden, measured as attack rate. We consider a modeling framework based on generalized contact matrices that extend traditional age-stratified approaches to incorporate socioeconomic status (SES) variables. We simulate epidemic dynamics under two scenarios. In the first, vaccination campaigns are concurrent with epidemics. In the second, instead, vaccinations are completed before the onset of infection waves. By using both synthetic and empirical generalized contact matrices, we find that inequalities in vaccine uptake can lead to non-linear effects on disease outcomes and exacerbate disease burden in disadvantaged groups of the population. We demonstrate that simpler models ignoring SES heterogeneity produce incomplete or biased predictions of attack rates. Additionally, we show how inequalities in vaccine coverage interact with non-pharmaceutical interventions (NPIs), compounding differences across subgroups. Overall, our findings highlight the importance of integrating SES dimensions, alongside age, into epidemic models to inform more equitable and effective public health interventions and vaccination strategies.

Poverty maps are essential tools for governments and NGOs to track socioeconomic changes and adequately allocate infrastructure and services in places in need. Sensor and online crowd-sourced data combined with machine learning methods have provided a recent breakthrough in poverty map inference. However, these methods do not capture local wealth fluctuations, and are not optimized to produce accountable results that guarantee accurate predictions to all sub-populations. Here, we propose a pipeline of machine learning models to infer the mean and standard deviation of wealth across multiple geographically clustered populated places, and illustrate their performance in Sierra Leone and Uganda. These models leverage seven independent and freely available feature sources based on satellite images, and metadata collected via online crowd-sourcing and social media. Our models show that combined metadata features are the best predictors of wealth in rural areas, outperforming image-based models, which are the best for predicting the highest wealth quintiles. Our results recover the local mean and variation of wealth, and correctly capture the positive yet non-monotonous correlation between them. We further demonstrate the capabilities and limitations of model transfer across countries and the effects of data recency and other biases. Our methodology provides open tools to build towards more transparent and interpretable models to help governments and NGOs to make informed decisions based on data availability, urbanization level, and poverty thresholds.

Influence arriving via social ties may be relevant for a person to decide to buy a new product, share information, or to adopt a new behaviour. However, quantifying social influence is a difficult task, even in online social systems where the interactions and communication content can be closely followed. Here we study the information susceptibility and adoption thresholds of users on Twitter. We consider hashtag and retweet adoptions on different aggregation levels: items, users, and topic groups, and study these adoption mechanisms characterized by diversified or repeated influence stimuli. We find both metrics to be heterogeneously distributed, correlated, and dependent on the topics and aggregation level of social influence. We show that users adopt retweets easier than hashtags, and find that new influencing neighbors can effectively trigger adoptions. Our results may inform better models of adoption processes leading to a deeper empirical understanding of simple and complex contagion.

Our usage of language is not solely reliant on cognition but is arguably determined by myriad external factors leading to a global variability of linguistic patterns. This issue, which lies at the core of sociolinguistics and is backed by many small-scale studies on face-to-face communication, is addressed here by constructing a dataset combining the largest French Twitter corpus to date with detailed socioeconomic maps obtained from national census in France. We show how key linguistic variables measured in individual Twitter streams depend on factors like socioeconomic status, location, time, and the social network of individuals. We found that (i) people of higher socioeconomic status, active to a greater degree during the daytime, use a more standard language; (ii) the southern part of the country is more prone to use more standard language than the northern one, while locally the used variety or dialect is determined by the spatial distribution of socioeconomic status; and (iii) individuals connected in the social network are closer linguistically than disconnected ones, even after the effects of status homophily have been removed. Our results inform sociolinguistic theory and may inspire novel learning methods for the inference of socioeconomic status of people from the way they tweet.

We analyze a coupled dataset collecting the mobile phone communication and bank transactions history of a large number of individuals living in Mexico. After mapping the social structure and introducing indicators of socioeconomic status, demographic features, and purchasing habits of individuals we show that typical consumption patterns are strongly correlated with identified socioeconomic classes leading to patterns of stratification in the social structure. In addition we measure correlations between merchant categories and introduce a correlation network, which emerges with a meaningful community structure. We detect multivariate relations between merchant categories and show correlations in purchasing habits of individuals. Our work provides novel and detailed insight into the relations between social and consuming behaviour with potential applications in recommendation system design.