Continuing the network science installment, this time with Hyper Fleck Information Space.
Mate choice: How social trends influence mate diversity
Feb 2026, phys.org
If everyone performed "mate copying" behavior, then diversity would decline. This is what happens instead:Conformity: Here, the majority follows the trend. The model shows that this can paradoxically lead to the fixation of traits that have a lower biological quality. A rarer, actually fitter type then has little chance of asserting itself against the established social trend.Anti-conformity: If individuals deliberately copy the minority, diversity in the population remains stable.This new model makes it possible to identify the "critical copying probability." This threshold value marks the point at which social information overrides natural selection. If around 40% of the population follows the example of other individuals when choosing a mate, a biologically inferior type can suddenly dominate the group.The study emphasizes that evolution is not determined by genes alone. It is also shaped by the way information flows and is processed within a community.
via University of Würzburg: Srishti Patil et al, Phenotypic polymorphism via mate copying, Proceedings of the National Academy of Sciences (2026). DOI: 10.1073/pnas.2510849123
Image credit: Slime mold Arcyria denudata by Frederic Labaune - Nikon Small World Photomicrography Competition - 2025
Personal change thresholds may explain why popular policies fail to spread
Mar 2026, phys.org
Some people will try a new idea the moment they hear about it. Others wait until everyone else is doing it. In survey experiments, participants repeatedly chose between options such as energy policies or messaging apps while seeing different levels of social support for each one. Based on these choices, the team estimated each participant's personal threshold for change. "This approach lets us infer individual tipping points."Using extensive simulations on real social networks, they compared different strategies for "seeding" change. They found that strategies combining two types of information — social network structure and individual thresholds for change — consistently outperformed approaches based on only one of these factors.In scenarios where individuals with high thresholds were less responsive to targeting, the most effective strategy was to target those individuals connected to many others who were already close to adopting the change.In settings where targeting is costly, as in online influencer marketing, the best results came from more sophisticated algorithms that took both network structure and individual thresholds into account."By identifying who needs just a little nudge and how influence spreads through social networks, interventions can be designed to have a much larger impact."
via University of Zurich: Radu Tănase et al, Integrating behavioural experimental findings into dynamical models to inform social change interventions, Nature Human Behaviour (2026). DOI: 10.1038/s41562-026-02417-4
Bell-bottoms today, miniskirts tomorrow: Math reveals fashion's 20-year cycle
Mar 2026, phys.org
The 20-year-rule in fashion; it's true and it's here:
Analyzing roughly 37,000 images of women's clothing spanning from 1869 to 2025, taken from the historical sewing patterns of the Commercial Pattern Archive at the University of Rhode Island, and identifying datapoints (literal points on the pictures) of eyes, neckline, waistline, hemline, feet to measure the fashion trends.It's one of the most comprehensive quantitative datasets of fashion ever assembled.Also - "The system intrinsically wants to oscillate" ... and in this case, that oscillation is between the tension between wanting to stand out while still fitting in; once a style becomes too common, designers move away from it—but not so far that the clothes become unwearable.But not anymore, apparently - One of the clearest patterns involves hemline length; skirt lengths have repeatedly shortened and lengthened; but starting in the 1980s, the data show a wider range of skirt lengths appearing at the same time, suggesting that fashion trends are becoming more fragmented, and rather than one dominant trend, niches emerge, reflecting more diversity in fashion.
via Northwestern: Emma Zajdela, "Back in Fashion: Modeling the Cyclical Dynamics of Trends," of the session "Statistical Physics of Networks and Complex Society Systems" at the American Physical Society Global Physics Summit in Denver, March 17 2026
A new way to detect breakthroughs in science: Large-scale analysis reveals 'disruptive' innovations in research history
Mar 2026, phys.org
Hyper Fleck Infospace - Using a machine-learning technique known as neural embedding, the researchers built a map of approximately 55 million scientific papers and patents. Each paper is represented by two points—one reflecting the research it built upon, another reflecting the research it inspired. When a paper is truly disruptive, these two points are far apart, meaning it redirected future research away from what came before it. Unlike other disruption indexes, it is sensitive to broader contexts and can better identify "simultaneous discoveries."
(This below is from the paper proper)
"Bibliometric Data Artifacts"Here, we introduce an embedding-based measure that captures the extent to which a scientific work redirects the research trajectory.Our approach embeds each paper in a high-dimensional space reflecting its direct and indirect connections to prior and subsequent work. Just like neural language models that represent tokens and sentences as vectors, we imagine each paper as a vector that captures its intellectual “position.” We then train two distinct vectors for each paper in the same embedding space: one representing its past, or “antecedents,” context—the configuration of prior work it draws upon—and another representing its future, or “descendants,” context—the body of work it gives rise to. When a contribution substantially reshapes the trajectory connecting past to future, or initiates a new stream of research, these two contexts diverge; the distance between them therefore captures the extent to which subsequent work departs from the prior knowledge....As a reference point, we use the disruption index (“CD index”) (15, 16), a widely used indicator that uses the topology of local citation network. The disruption index captures how subsequent work diverges from earlier foundations, focusing on whether later papers cite the predecessors of a focal contribution through direct citation links....Using a dataset of more than 55 million scientific papers from the Web of Science (WoS) and the American Physical Society (APS), we show that our measure—“Embedding Disruptiveness Measure” (EDM)—provides a continuous, high-resolution view of how scientific contributions reconfigure the relationship between inherited knowledge and emerging directions....If the embedding model is trained such that the proximity between the vectors indicates higher connections between their papers, and if disruptive papers tend to eclipse the future knowledge from the past, making future knowledge less rely on the past, we expect that a paper’s past and future vectors diverge as the paper’s disruptiveness increases. Thus, by quantifying the distance between these two vectors—representing the past and future context of each paper—we can estimate their disruptiveness.Simultaneous disruption - To understand why some of the landmark papers have such low D scores, resulting in a bimodal distribution of D, we examine the top 10 papers with the largest difference between the disruption index score D and the EDM score (delta). We found that all 10 papers are related to the notable examples of simultaneous disruption.
via State University of New York Binghamton University and Center for Complex Networks and Systems Research, Luddy School of Informatics, Indiana University: Uncovering simultaneous breakthroughs with a robust measure of disruptiveness, Science Advances (2026). DOI: 10.1126/sciadv.adx3420
