Abstract The growing electricity demand and use of smart appliances are placing pressures on power grids, making efficient energy management more important than ever. The existing energy management systems often prioritize system efficiency (balanced energy demand and supply) at the expense of consumer comfort. This paper addresses this gap by proposing a novel decentralized multi-agent coordination-based demand-side management system. The proposed system enables individual agents to coordinate for demand-side energy optimization while improving the consumer comfort and maintaining the system efficiency. A key innovation of this work is the introduction of a slot exchange mechanism, where agents first receive optimized appliance-level energy consumption schedules and then coordinate with each other to adjust these schedules through slot exchanges to improve their comfort even when agents show non-altruistic behaviour. It also scales well with large populations and promotes fairness by balancing satisfaction levels across consumers. For performance evaluation, a real-world dataset is used, and the results demonstrate that the proposed slot exchange mechanism increases consumer comfort and fairness without raising system inefficiency cost, making it a practical and scalable solution for future smart grids.

Digital technologies and AI promise to optimise complex systems through data-driven decisions, predictive modelling, and anticipatory action. However, this optimisation imperative creates a fundamental paradox: as systems excel at achieving measurable objectives, they may erode the collective intelligence and adaptive capacity of our societies. Recognising this tension, the field of Human-Centred AI (HCAI) has emerged to develop design principles such as explainability, fairness, and transparency to ensure that AI aligns with human values. However, research on HCAI often focuses on idealised interactions, neglecting the pressure, moral dilemmas, and social dynamics typical of today’s complex problems. This paper introduces and advocates for a paradigm shift towards Sensemaking AI: AI that supports collective meaning-making processes in evolving human-AI networks. This novel perspective recognises that algorithmic and AI systems actively participate in the social processes through which humans interpret information, coordinate responses, and adapt their values. Grounded in sensemaking and decision theory and informed by a scoping review of the HCAI literature, this paper identifies three connected research areas: (i) sensemaking-aware automation that preserves interpretive flexibility; (ii) collective agency for network-level control; and (iii) value-aware sensemaking that supports collective meaning-making. These principles form the basis for Sensemaking AI as a design and research agenda that prioritises collective meaning-making and democratic deliberation in networks.

The negative impact of floods on public health has been increasing, as climate change makes these events more frequent and intense. Floods are known to cause direct injury and favor the spread of many waterborne and vector-borne pathogens. Their effect on the circulation of respiratory pathogens, like influenza and SARS-CoV-2, is, however, still unclear. In this study, we quantify this effect through the analysis of large-scale behavioral data coupled to mathematical models of epidemic spread. We focus on the devastating floods occurred in Italy in 2023 and measure how they impacted human contact patterns within and between communities. We find a substantial increase in contacts occurring 3 weeks after the floods, both among residents of the affected areas and between them and those living in distant, unaffected areas of Italy. Then, through mathematical simulations, we determine that these disrupted contact patterns can carry a circulating pathogen to previously unaffected geographic areas, as well as increasing infection counts across the country. Our findings may help set up protocols to use large-scale human contact data to contain epidemic outbreaks before, during and in the aftermath of floods.

Large Language Models (LLMs) have significantly advanced prompt-based information retrieval, yet their potential to reproduce or amplify social biases remains insufficiently understood. In this study, we investigate this issue through the concrete task of reconstructing real-world co-authorship networks of computer science (CS) researchers using two widely used LLMs—GPT-3.5 Turbo and Mixtral 8x7B. This task offers a structured and quantifiable way to evaluate whether LLM-generated scholarly relationships reflect demographic disparities, as co-authorship is a key proxy for collaboration and recognition in academia. We compare the LLM-generated networks to baseline networks derived from DBLP and Google Scholar, employing both statistical and network science approaches to assess biases related to gender and ethnicity. Our findings show that both LLMs tend to produce more accurate co-authorship links for individuals with Asian or White names, particularly among researchers with lower visibility or limited academic impact. While we find no significant gender disparities in accuracy, the models systematically favor generating co-authorship links that overrepresent Asian and White individuals. Additionally, the structural properties of the LLM-generated networks differ from those of the baseline networks. These results highlight the importance of examining how LLMs represent social and scientific relationships, particularly in contexts where they are increasingly used for knowledge discovery and scholarly search.

This paper investigates identity bias (gender and race) in the South African news selection and representation of COVID-19 vaccination quotes. Social bias studies have qualitatively examined race and gender bias in South African news, given South Africa’s apartheid history; yet, studies that examine and quantify these biases at the speaker level using news quotes from a representative South African news corpus remain limited. To address this gap, we examined race and gender bias in news selection and framing of quotes. We used word embedding trained on 22,627 vaccination quotes from 76 South African news sources between 2020 and 2023. These large-scale processing embeddings are unbiased by design but can learn and uncover biases hidden in language. Our findings reveal gender and race bias in the news selection and framing of quotes – journalists privilege White voices as more authoritative and connected to global and technical vaccination discourse but confine black voices to primarily localised contexts. They also quote male speakers more frequently in the news than females. In an era where human biases are becoming increasingly implicit, we argue that embeddings offer a robust tool to unearth, monitor, and evaluate these biases at the micro or speaker level in the news.

Artificial Intelligence (AI) has been repeatedly associated with the potential for automating, or even augmenting, occupational tasks. However, the geographical impact of AI remains unclear. Building on previous work, we employed a deep learning natural language processing model to automatically identify AI patents impacting various occupational tasks. We analyzed a dataset of 17,879 task descriptions and quantified AI’s potential impact at metropolitan statistical areas (MSAs) within the U.S. by examining 24,758 AI patents filed with the United States Patent and Trademark Office (USPTO) between 2015 and 2022. Our findings reveal that MSAs that will be more likely to be impacted by AI are not just hubs of creative industries but will also be characterized by a lack of economic diversity. Indeed, the U.S. MSAs that will be more likely to be impacted are those heavily specialized, with little to no efforts at diversification. These dynamics suggest that AI could amplify existing divides, hitting hardest in areas where economic opportunities are already concentrated in a few sectors, leaving many behind in the race for innovation-driven growth.

Mobile phone data have played a key role in quantifying human mobility during the COVID-19 pandemic. Existing studies on mobility patterns have primarily focused on regional aggregates in high-income countries, obfuscating the accentuated impact of the pandemic on the most vulnerable populations. Leveraging geolocation data from mobile-phone users and population census for 6 middle-income countries across 3 continents between March and December 2020, we uncovered common disparities in the behavioral response to the pandemic across socioeconomic groups. Users living in low-wealth neighborhoods were less likely to respond by self-isolating, relocating to rural areas, or refraining from commuting to work. The gap in the behavioral responses between socioeconomic groups persisted during the entire observation period. Among users living in low-wealth neighborhoods, those who commute to work in high-wealth neighborhoods pre-pandemic were particularly at risk of experiencing economic stress, facing both the reduction in economic activity in the high-wealth neighborhood and being more likely to be affected by public transport closures due to their longer commute distances. While confinement policies were predominantly country-wide, these results suggest that, when data to identify vulnerable individuals are not readily available, GPS-based analytics could help design targeted place-based policies to aid the most vulnerable.