Research articles

Yabe and colleagues use human mobility data to identify dependencies in visitation patterns between places in cities, with unexpected prevalence at long-distance connections. Leveraging this network enhances the accuracy of predicting business resilience during shocks.

Fabel et al. find that youth participation in the Fridays for Future climate movement relates to Green Party vote shares in Germany through the transmission of pro-environmental attitudes, social media presence and media coverage of environmental issues.

In the Bolivian Amazon, analyses of preserved human bone collagen show evidence of maize staple diets between ce 700 and 1400, as well as management of muscovy ducks, one of the few animals from the Americas to be domesticated.

Zhang and Rosenberg built a model that predicts surprise from brain network dynamics measured with fMRI revealing similarities across distinct contexts (task learning, basketball watching and cartoon viewing).

This genome-wide association study of occupational status in UK Biobank identified 106 genetic variants. Results highlight the role of family environment, childhood educational and occupational aspirations, and links to health.

Wild chimps in Bossou, Guinea, use stone tools to crack nuts. They show individual variation in how efficient they are, and some individuals are consistently more efficient than others.

This study assessed COVID-19 social science preprints’ replicability using structured groups. Both beginners and more-experienced participants used a elicitation protocol to make better-than-chance predictions about the reliability of research claims under high uncertainty.

Leng et al. find that framing options in a choice set as inclusive, rather than mutually exclusive, increases the efficiency of choices and reduces feelings of conflict.

Glickman and Sharot reveal a human–AI feedback loop, where AI amplifies subtle human biases, which are then further internalized by humans. This cycle, observed across various domains, leads to substantial increases in human bias over time.

Biobank data are extensively used. Schoeler and colleagues show that self-report inaccuracy and selective participation are sources of poor reproducibility for biobank-scale research.

Voloudakis and colleagues find statistically significant but weak associations between professional categories and genetic predisposition for neuropsychiatric traits.

In this Stage 2 Registered Report, Lin et al. show that trait impressions of faces are correlated with and causally influence mental state inferences from the same faces.

Kang et al. examine how patterns of knowledge diffusion can forecast the collapse of scientific ‘bubbles’, highlighting that sustained scientific advancement requires diverse audiences.

Gelabert et al. examine genomic and archaeological data from Europe’s earliest farming communities in Central Europe (5500–5000 bce). They find differentiated genetic networks but no evidence of unequal access to resources linked to sex or kin.

Large language models (LLMs) can synthesize vast amounts of information. Luo et al. show that LLMs—especially BrainGPT, an LLM the authors tuned on the neuroscience literature—outperform experts in predicting neuroscience results and could assist scientists in making future discoveries.

The introduction of mandatory calorie labelling in out-of-home food outlets in England was not associated with a reduction in calories purchased or consumed. There was a post-policy increase in self-reported noticing and use of calorie labelling.

In four studies, Kelly and Sharot reveal that web-browsing both reflects and affects mental health. Poorer mental health leads to more negative content consumption, which in turn worsens mood. Highlighting webpage emotional impacts reduced negative browsing and improved mood.

Using a battery of statistical tools, Alagöz et al. examine the genetic overlap between dyslexia and rhythm impairment and shed light on how the genome influences the neural bases of human language and musicality.

This study finds that decision markets can be a useful tool for selecting studies for replication. For a sample of 26 online experiments published in PNAS selected by a decision market, the authors find replication rates ranging between 54% and 62%.

Over 75% of links to news shared on Facebook during the election seasons of 2017–2020 were forwarded without the sharer first clicking on them. Extreme and user-aligned political content was shared more, aiding the spread of partisan misinformation.