We have two new publications out!

1. Emergent social conventions and collective bias in LLM populations, by Ariel Flint Ashery, Luca Maria Aiello, and Andrea Baronchelli, published in Science Advances.
   We explore the collective behavior of LLMs starting from social conventions, the fundamental building blocks of coordinated societies. We used the naming game, a well-established framework that has been applied for decades to study conventions in humans. We found that LLM populations can spontaneously develop shared conventions through local interactions. These interactions can produce collective biases, invisible at the individual level, raising important considerations for AI alignment. Small committed minorities can trigger tipping points, steering the entire group toward new conventions—a dynamic well known in human societies
2. The experience of running: Recommending routes using sensory mapping in
   urban environments, by Katrin Hänsel, Luca Maria Aiello, Daniele Quercia, Rossano Schifanella, Krisztian Zsolt Varga, Linus W. Dietz, and Marios Constantinides, published in the International Journal of Human-Computer Studies.
   We set out to build running routes not around distance, but around how people feel: before, during, and after a run. We surveyed 387 runners and found that not everyone wants the same kind of run. Some seek quiet and greenery; others thrive on the buzz of people and traffic. Their preferences often match their personality. Runners who prefer scenic paths (quiet, green, and natural) tended to score higher in neuroticism. Those who preferred urban paths (lively and full of people) were more likely to be extroverted. Then, we built a routing engine. Using millions of geotagged Flickr photos and open London data, we scored streets for beauty, noise, safety, and surface quality. We tested the engine on hundreds of 5-km London loops. Most runners preferred the scenic routes.

We have two new publications out at CHIIR 2025 in Melbourne (Proceedings of the 2025 ACM SIGIR Conference on Human Information Interaction and Retrieval):

1. Exploring the Zero-Shot Known-Item Retrieval Capabilities of LLMs for Casual Leisure Information Needs, by T. Bogers, M. Gäde, M. Hall, M. Koolen, V. Petras, M. Skov, published in CHIIR 2025
   We compared four LLMs on their capability to answer a specific type of complex search task: known-item requests from casual leisure domains. We constructed a test collection by gathering known-item requests for books, games and movies from online forums along with verified answers by the original requester. We prompted four LLMs multiple times with the same prompt and analyzed the results with respect to accuracy and the degree to which answers were fabricated by the LLM. Our results show that LLMs are not particularly effective in fulfilling these complex casual leisure needs, but there are are big differences between LLMs and across domains.
2. From Queries to Candidates: Exploring Search and Source Interaction Behavior of Recruiters, by T. Bogers, M. Kaya, M. Gäde, published in CHIIR 2025
   
   Recruitment is a professional search domain that has been largely overlooked in IR research, even though better support of recruiters could have a big impact on job seekers, companies and society as a whole. In this paper, we analyze the search formulation and source selection behavior of the recruiters at one of Scandinavia’s largest job portals and recruitment agencies using search logs for close to 18,000 recruitment search tasks. We provide an analysis of the search sessions of recruiters in terms search tactics, query operators, query length, term re-use and filter usage, and break down their behavior both by task type and task complexity. We also relate their short-term tactics to different learning stages in the search process and investigate their influence on search success. We find that identifying and assessing relevant candidates for a job posting is a complex task: recruiters usually submit multiple queries during sessions that can last for hours and that increase in complexity. Recruiters all spend more time per query as their session progresses. We also observed query reformulation strategies that indicate distinct patterns of knowledge gaining during sessions. Relating these tactics to positive responses from candidates we aim at predicting successful strategies.

We have two new publications out, one on closing triads in social networks, and one on shaded walks (CoolWalks) in cities:

1. Exploring Time-Ordered Triadic Closure in Online Social Networks, by A. Galdeman, C. Tidiane Ba, M. Zignani, S. Gaito, published in ACM Transactions on the Web
   
   By analyzing networks with timestamped links from diverse platforms based on different architectures, including communication, Web3-based, and trade networks, we developed a comprehensive analytical pipeline to support the study of triadic closure patterns. This pipeline includes an algorithm for the census of time-ordered triads, a vector-based model for representing growing networks (growth triadic profile), the identification of triadic closure rules (TERs), and the evaluation of the speed of the formation of closed triads. Our findings reveal significant variations in the impact of triadic closure across different OSNs, marked by diverse growth triadic profiles and varying formation speeds of closed triads as well as diversity in the predictability of evolutionary patterns based on triads. This study not only enhances the comprehension of triadic closure in the temporal evolution of OSNs but also provides valuable insights to be taken into account for the design and administration of online social platforms.
2. CoolWalks for active mobility in urban street networks, by H. Wolf, A.R. Vierø & M. Szell, published in Scientific Reports
   
   Here we explore the potential for shaded walking, using building footprints and street networks from both synthetic and real cities. We introduce a route choice model with a sun avoidance parameter alpha and define the CoolWalkability metric to measure opportunities for walking in shade. We derive analytically that on a regular grid with constant building heights, CoolWalkability is independent of alpha, and that the grid provides no CoolWalkability benefit for shade-seeking individuals compared to the shortest path. However, variations in street geometry and building heights create such benefits. We further uncover that the potential for shaded routing differs between grid-like and irregular street networks, forms local clusters, and is sensitive to the mapped network geometry. Our research identifies the limitations and potential of shade for cool, active travel, and is a first step towards a rigorous understanding of shade provision for sustainable mobility in cities.

This week we are on fire and have yet another big paper out, long time in the making, led by Luca Maria Aiello with multi-NERDS involvement, just published in PNAS: Urban highways are barriers to social ties, by L.M. Aiello, A. Vybornova, S. Juhász, M. Szell, and E. Bokányi.

We have a big paper out today, long time in the making, led by Yanmeng Xin, our long-term PhD student visitor in 2021-2023, co-authored by Roberta Sinatra, just published in Nature Human Behavior: Academic mentees thrive in big groups, but survive in small groups, by Y. Xing, Y Ma, Y. Fan, R. Sinatra, and A. Zeng.

The main message of the paper is intriguing: If you “grow up” in a big research group, and if you survive, you will have high success. At the same time, in a big group it is also harder to survive, especially if your mentor is very productive. So what is then good mentorship, and what is a good group to be?

Interestingly, at NERDS we are a fairly big research group, but with several mentors who are by themselves not too busy, so we combine the best of both worlds 😁 In fact, this paper itself is another one in a long series of success stories where a visitor accomplished something great while staying at our ✨🦄 ~enchanted NERDS grounds~ 🧚‍♂️✨ in Copenhagen. (“NERDS is one of the best places I have ever stayed.”)

Mentoring is a key component of scientific achievements, contributing to overall measures of career success for mentees and mentors. Within the scientific community, possessing a large research group is often perceived as an indicator of exceptional mentorship and high-quality research. However, such large, competitive groups may also escalate dropout rates, particularly among early-career researchers. Overly high dropout rates of young researchers may lead to severe postdoc shortage and loss of top-tier academics in contemporary academia. In this context, we collect longitudinal genealogical data on mentor-mentee relations and their publication, and analyze the influence of a mentor’s group size on the future academic longevity and performance of their mentees. Our findings indicate that mentees trained in larger groups tend to exhibit superior academic performance compared to those from smaller groups, provided they remain in academia post-graduation. However, we also observe two surprising patterns: Academic survival rate is significantly lower for (1) mentees from larger groups, and for (2) mentees with more productive mentors. The trend is verified in institutions of different prestige. These findings highlight a negative correlation between a mentor’s success and the academic survival rate of their mentees, prompting a rethinking of effective mentorship and offering actionable insights for career advancement.

NERDS member Michele Coscia has updated his textbook for the Network Analysis and Advanced Network Science classes he teaches at ITU. This “Atlas for the Aspiring Network Scientist”, has now reached version 2.0, and 916 pages, and is available for anyone to read for free: https://arxiv.org/abs/2101.00863

Website: https://www.networkatlas.eu/

The new edition has a much improved coverage on graph neural networks, network data uncertainty, and background knowledge in statistics, machine learning, probability theory, and linear algebra.

Even version 2.0 has big margins for improvements. Please contact Michele with any comments.

Find a more detailed explanation of The Atlas for the Aspiring Network Scientist on Michele’s page: https://www.michelecoscia.com/?p=2393.

We have two new publications out, one on the Gamestop short squeeze by Reddit users, and one on bicycle network design with use case Copenhagen:

1. The dynamics of the Reddit collective action leading to the GameStop short squeeze, by A. Desiderio, L.M. Aiello, G. Cimini & L. Alessandretti , published in npj complexity
   
   In early 2021, the stock prices of GameStop, AMC, Nokia, and BlackBerry experienced dramatic increases, triggered by short-squeeze operations that have been largely attributed to Reddit’s retail investors. Here we shed light on the extent and timing of Reddit users’ influence on the GameStop short squeeze. Using statistical analysis tools with high temporal resolution, we find that increasing Reddit discussions anticipated high trading volumes. This effect emerged abruptly a few weeks before the event but waned once the community gained widespread visibility through Twitter. Meanwhile, the collective investment of the community, quantified through posts of individual positions, closely mirrored the market capitalization of the stock. This evidence suggests a coordinated action of users in developing a shared financial strategy through social media—targeting GameStop first and other stocks afterward. Overall, our results provide novel insights into the role of Reddit users in the dynamics of the GameStop short squeeze.
2. Cohesive urban bicycle infrastructure design through optimal transport routing in multilayer networks, by A. Lonardi, M. Szell and C. De Bacco, published in Journal of the Royal Society Interface
   
   Bicycle infrastructure networks must meet the needs of cyclists to position cycling as a viable transportation choice in cities. In particular, protected infrastructure should be planned cohesively for the whole city and spacious enough to accommodate all cyclists safely and prevent cyclist congestion—a common problem in cycling cities like Copenhagen. Here, we devise an adaptive method for optimal bicycle network design and for evaluating congestion criticalities on bicycle paths. The method goes beyond static network measures, using computationally efficient adaptation rules inspired by optimal transport on the dynamically updating multilayer network of roads and protected bicycle lanes. Street capacities and cyclist flows reciprocally control each other to optimally accommodate cyclists on streets with one control parameter that dictates the preference of bicycle infrastructure over roads. Applying our method to Copenhagen confirms that the city’s bicycle network is generally well-developed. However, we are able to identify the network’s bottlenecks, and we find, at a finer scale, disparities in network accessibility and criticalities between different neighbourhoods. Our model and results are generalizable beyond this particular case study to serve as a scalable and versatile tool for aiding urban planners in designing cycling-friendly cities.
   

We have been very productive over the winter! Five new NERDS publications were released in December and this January, on topics as diverse as archaeological networks, dynamic networks, spatial data science, climate change debates, and LLM-generated data:

1. “A Network of Mutualities of Being”: Socio-material Archaeological Networks and Biological Ties at Çatalhöyük, by C. Mazzucato, M. Coscia, A. Küçükakdağ Doğu, S. Haddow, M. Sıddık Kılıç, E. Yüncü & M. Somel, published in Journal of Archaeological Method and Theory
   
   In this paper, we propose a Network Science framework to integrate archaeogenomic data and material culture at an intra-site scale to study biological relatedness and social organization at the Neolithic site of Çatalhöyük. Methodologically, we propose the use of network variance to investigate the association between biological relatedness and material culture within networks of houses. This approach allows us to observe how material culture similarity between buildings is associated with biological relationships between individuals and how biogenetic ties concentrate at specific localities on site.
2. Graph Evolution Rules Meet Communities: Assessing Global and Local Patterns in the Evolution of Dynamic Networks, by A. Galdeman, M. Zignani & S. Gaito, published in Big Data Mining and Analytics
   
   In this paper, we comprehensively explore Graph Evolution Rules (GERs) in dynamic networks from diverse systems with a focus on the rules characterizing the formation and evolution of their modular structures, using EvoMine for GER extraction and the Leiden algorithm for community detection. We characterize network and module evolution through GER profiles, enabling cross-system comparisons. By combining GERs and network communities, we decompose network evolution into regions to uncover insights into global and mesoscopic network evolution patterns. From a mesoscopic standpoint, the evolution patterns characterizing communities emphasize a non-homogeneous nature, with each community, or groups of them, displaying specific evolution patterns, while other networks’ communities follow more uniform evolution patterns. Additionally, closely interconnected sets of communities tend to evolve similarly. Our findings offer valuable insights into the intricate mechanisms governing the growth and development of dynamic networks and their communities, shedding light on the interplay between modular structures and evolving network dynamics.
3. Teaching spatial data science, by A.R. Vierø & M. Szell, published in Geoforum Perspektiv
   
   Spatial data science is an emerging field building on geographic information science, geography, and data science. Here we first discuss the definition and history of the field, arguing that it indeed warrants a new label. Then, we present the design of our course Geospatial Data Science at IT University of Copenhagen and discuss the importance of teaching not just spatial data science tools but also spatial and critical thinking. We conclude with a perspective on the potential future for spatial data science, arguing that qualitative theory and methods will continue to play an important role despite new GeoAI-related advances.
4. Do You See What I See? Emotional Reaction to Visual Content in the Online Debate About Climate Change, by L. Rossi, A. Segerberg, L. Arminio & M. Magnani, in Environmental Communication.
   
   This paper explores the visual echo chamber effect in online climate change communication. We analyze communication by progressive actors and counteractors involved in the public debate about climate change on Facebook, to address the possibility that visual content can bridge ideologically diverse communities. Specifically, we investigate whether visual content depicting protest serves this purpose. The findings reveal a small amount of shared visual content. Interestingly, the emotional reactions to this content for the most part diverge significantly, suggesting that pre-existing attitudes, such as climate ideological position, influence interpretation. Contrary to our expectations, however, we do not observe visual content representing protest activity bridging the two groups. This work posits the possibility of a two-fold (de)polarization around visual content that both connects and divides, which contributes to a more nuanced understanding of the social dynamics that create and sustain the echo chamber effect observed in online climate change debates.
5. The Problems of LLM-generated Data in Social Science Research  by L. Rossi, K. Harrison & I Shklovski, in  Sociologica.
   The paper explores LLMs when used for generating synthetic data for social science and design research. Researchers have used LLM-generated data for data augmentation and prototyping, as well as for direct analysis where LLMs acted as proxies for real human subjects. LLM-based synthetic data build on fundamentally different epistemological assumptions than previous synthetically generated data and are justified by a different set of considerations. In this essay, we explore the various ways in which LLMs have been used to generate research data and consider the underlying epistemological (and accompanying methodological) assumptions. We challenge some of the assumptions made about LLM-generated data, and we highlight the main challenges that social sciences and humanities need to address if they want to adopt LLMs as synthetic data generators.
   

Two weeks ago the Nobel prize in physics was awarded to Hopfield and Hinton for their research on artificial neural networks. This caused quite some uproar, especially by many of our computer science and physics colleagues. As original-physicists-turned-data-scientists-dabbling-in-AI, who have done data-driven Science of Science research exactly on the crucial role of Hopfield and Hinton’s papers in physics, we penned a comment pointing to our clarifying research which was now published as a correspondence in Nature:

Was the Nobel prize for physics? Yes — not that it matters, by M. Szell, Y. Ma, and R. Sinatra

Here the entire correspondence:

The award of the 2024 Nobel Prize in Physics to John Hopfield and Geoffrey Hinton for their groundbreaking research on artificial neural networks (Nature 634, 523–524; 2024) has caused consternation in some quarters. Surely this is computer science, not physics?

Existing data can help to inform this debate. Almost a decade ago, two of us (M.S. and R.S.) co-authored an analysis of referencing and citation patterns that explicitly placed Hopfield’s seminal 1982 paper on neural networks among 3.2 million interdisciplinary papers in non-physics journals that were “indistinguishable from papers published in physics journals”. Six other physics Nobel-winning papers were also in this set (R. Sinatra et al. Nature Phys. 11, 791–796; 2015).

The physics Nobel prize has until recently rewarded conventional ‘core’ physics research, even though Hopfield’s and Hinton’s papers were ripe for recognition (M. Szell et al. Nature Phys. 14, 1075–1078; 2018). We hope that this year’s prize will expedite the breakdown of silos that obstruct thinking across disciplines. Clinging to the idea of research fields as fixed territories is at best small-minded, and at worst harmful, when it comes to solving global challenges such as climate change.

Our original version – before editorial changes – provides a slightly different angle and an instructive figure (that was cut for publication):