arXiv subjects

Viola Priesemann

Publications and source records attributed to Viola Priesemann.

At least 19 recordsLinked to original sources

arXiv2026

Network Information Enhances Unreliable News Domain Detection

Content-based detection of unreliable news is increasingly difficult, as low-reliability sources mimic credible journalism and generative AI makes fabricated content harder to flag. We ask whether network structure can improve news reliability classification, taking a domain-level approach that shifts the focus from individual articles to source reliability. From URL-sharing patterns in Telegram chats, we build a statistically validated domain co-sharing network and find assortative mixing by reliability: low-reliability domains group together, as do reliable ones. Exploiting this structure, we compare Graph Neural Networks against network-unaware baselines using both content-aware features (multilingual text embeddings) and content-agnostic features (spreading dynamics). GNNs consistently outperform Multi-Layer Perceptrons on identical features, with GraphSAGE best in both settings (accuracy 0.63 with content, 0.53 without), a 13-14% relative gain over the network-unaware baseline. Network topology thus systematically improves domain reliability assessment, and remains effective even when content analysis is infeasible.

cs.SI↗

arXiv2026

Mechanics of Pandemics

COVID-19 and previous pandemics have shown how diseases can disrupt, threaten, and transform daily life. Since pathogens and societies are continuously evolving, every pandemic is different. However, certain fundamental principles of disease transmission appear to hold true across different outbreaks. These ``mechanisms'' are grounded in natural laws or the very structure of our biology and societies. This paper compiles ten fundamental mechanisms, curated by a multidisciplinary team with backgrounds spanning public health, medicine, epidemiology, political science, mathematics, physics, and psychology. These mechanisms, although perhaps underappreciated, substantially shape how pandemics unfold and are controlled. The better we succeed in understanding these mechanisms and establishing this knowledge in our societies, the better we will be able to prepare for future pandemics and respond appropriately when they occur.

q-bio.PE↗

arXiv2026

Learning Through Noise: Why Subliminal Learning Works and When It Fails

In the context of artificial neural networks, subliminal learning refers to the transfer of task-relevant knowledge or unintended biases from teacher to student models through distillation on task-unrelated input$\unicode{x2013}$output pairs. Prior explanations tie this effect to shared or closely matched teacher$\unicode{x2013}$student initialization. We show that a closely matched initialization is not necessary. Instead, subliminal learning is governed by compatible output heads. Using a controlled MNIST setting, we split outputs into an auxiliary head (for auxiliary, task-unrelated noise signals) and a class head (for classification) to demonstrate subliminal learning occurs$\unicode{x2014}$even when we randomly initialize hidden layers and remove layers, add new layers, or change the architecture (MLP-to-CNN). Compatible auxiliary heads enable transfer of a recoverable teacher signal, bringing the student's representations closer to the teacher's. When the class heads remain compatible as well, students trained only on task-unrelated noise can approach, and in favorable regimes match, teacher-level task performance. Our setting enables us to develop a theory that explains the mechanism of subliminal learning and to derive upper bounds on when subliminal learning fails. Together, our results turn subliminal learning from a surprising transfer effect into a theoretically grounded mechanism with predictable limits.

cs.LG↗

arXiv2026

TeraGram: A Structured Longitudinal Dataset of the Telegram Messenger

Here we present a massive longitudinal dataset of public Telegram content, comprising over 5.9 billion messages dating from 2015 to 2025, collected from 712 thousand channels and groups, enriched with metadata on forwards, reactions, and polls. The dataset spans multiple languages including Russian and Farsi, representing countries where Telegram shows mainstream adoption, as well as Western languages where Telegram is used in specific sub-communities. The dataset has several advantages. First, when restricted by language, it provides a versatile example of an algorithm-free platform, contrary to many other social media platforms that are strongly influenced by opaque content-curation algorithms. Second, it enables comparative studies across different languages, communities, and user bases under identical platform affordances. The dataset thus offers a foundation for studying engagement patterns, network evolution, and community formation in the absence of algorithmic curation.

physics.soc-ph↗

arXiv2026

Conformity Generates Collective Misalignment in AI Agents Societies

Artificial intelligence safety research focuses on aligning individual language models with human values, yet deployed AI systems increasingly operate as interacting populations where social influence may override individual alignment. Here we show that populations of individually aligned AI agents can be driven into stable misaligned states through conformity dynamics. Simulating opinion dynamics across nine large language models and one hundred opinion pairs, we find that each agent's behavior is governed by two competing forces: a tendency to follow the majority and an intrinsic bias toward specific positions. Using tools from statistical physics, we derive a quantitative theory that predicts when populations become trapped in long-lived misaligned configurations, and identifies predictable tipping points where small numbers of adversarial agents can irreversibly shift population-level alignment even after manipulation ceases. These results demonstrate that individual-level alignment provides no guarantee of collective safety, calling for evaluation frameworks that account for emergent behavior in AI populations.

physics.soc-ph↗

arXiv2026

Vocabulary shapes cross-lingual variation of word-order learnability in language models

Why do some languages like Czech permit free word order, while others like English do not? We address this question by pretraining transformer language models on a spectrum of synthetic word-order variants of natural languages. We observe that greater word-order irregularity consistently raises model surprisal, indicating reduced learnability. Sentence reversal, however, affects learnability only weakly. A coarse distinction of free- (e.g., Czech and Finnish) and fixed-word-order languages (e.g., English and French) does not explain cross-lingual variation. Instead, the structure of the word and subword vocabulary strongly predicts the model surprisal. Overall, vocabulary structure emerges as a key driver of computational word-order learnability across languages.

cs.CL↗

arXiv2026

Household size can explain 40% of the variance in cumulative COVID-19 incidence across Europe

Household size impacts the spread of respiratory infectious diseases: Larger households tend to boost transmission by acquiring external infections more frequently and subsequently transmitting them back into the community. Furthermore, mandatory interventions primarily modulate contagion between households rather than within them. We developed an approach to quantify the role of household size in epidemics by separating within-household from out-household transmission, and found that household size explains 41% of the variability in cumulative COVID-19 incidence across 34 European countries (95% confidence interval: [15%, 46%]). The contribution of households to the overall dynamics can be quantified by a boost factor that increases with the effective household size, implying that countries with larger households require more stringent interventions to achieve the same levels of containment. This suggests that households constitute a structural (dis-)advantage that must be considered when designing and evaluating mitigation strategies.

q-bio.PE↗

arXiv2026

Finite integration time can shift optimal sensitivity away from criticality

Sensitivity to small changes in the environment is crucial for many real-world tasks, enabling living and artificial systems to make correct behavioral decisions. It has been shown that such sensitivity is maximized when a system operates near the critical point of a phase transition. However, proximity to criticality introduces large fluctuations and diverging timescales. Hence, to leverage the maximal sensitivity, it would require impractically long integration periods. Here, we analytically and computationally demonstrate how the optimal tuning of a recurrent neural network is determined given a finite integration time. Rather than maximizing the theoretically available sensitivity, we find networks attain different sensitivities depending on the available time. Consequently, the optimal dynamic regime can shift away from criticality when integration times are finite, highlighting the necessity of incorporating finite-time considerations into studies of information processing.

cond-mat.dis-nn↗

arXiv2025

Optimizing infectious disease mitigation under dynamic conditions

Mitigation measures are essential for controlling the spread of infectious diseases during pandemics and epidemics, but they impose considerable societal, individual, and economic costs. We developed a general optimization framework to balance costs related to infection and to mitigation. Optimizing the trade-off between mitigation and infection cost, we identified three novel, surprising effects: First, assuming a constant reproduction number $R_0$, the optimal response to an infectious disease requires either strict mitigation or none at all, depending on disease severity, but never does one find an intermediate mitigation level to be optimal. Second, under seasonal variations, optimal mitigation is stricter during winter. Interestingly, a single wave of infections still arises in spring with 3 months delay to the seasonal peak of infectivity, replacing the autumn/winter waves known for classical influenza. Third, during steady vaccination campaigns, even optimal mitigation can result in transient infection waves. Finally, we quantify the cost of delayed mitigation onset and show that even short delays can substantially increase total costs -- if the disease is severe. Overall, our framework is easily applicable to general and complex settings and thereby presents a versatile tool to explore optimal mitigation strategies for endemic and pandemic infectious disease.

q-bio.PE↗

arXiv2025

Representation learning in cerebellum-like structures

Animals use past experiences to adapt future behavior. To enable this rapid learning, vertebrates and invertebrates have evolved analogous neural structures like the vertebrate cerebellum or insect mushroom body. A defining feature of these circuits is a large expansion layer, which re-codes sensory inputs to improve pattern separation, a prerequisite to learn non-overlapping associations between relevant sensorimotor inputs and adaptive changes in behavior. However, classical models of associative learning treat expansion layers as static, assuming that associations are learned through plasticity at the output synapses. Here, we review emerging evidence that also highlights the importance of plasticity within the expansion layer for associative learning. Because the underlying plasticity mechanisms and principles of this representation learning are only emerging, we systematically compare experimental data from two well-studied circuits for expansion coding -- the cerebellum granule layer and the mushroom body calyx. The data indicate remarkably similar interneuron circuits, dendritic morphology and plasticity mechanisms between both systems that hint at more general principles for representation learning. Moreover, the data show strong overlap with recent theoretical advances that consider interneuron circuits and dendritic computations for representation learning. However, they also hint at an interesting interaction of stimulus-induced, non-associative and reinforced, associative mechanisms of plasticity that is not well understood in current theories of representation learning. Therefore, studying expansion layer plasticity will be important to elucidate the mechanisms and full potential of representation learning for behavioral adaptation.

q-bio.NC↗

arXiv2025

Redundancy Maximization as a Principle of Associative Memory Learning in Hopfield Networks

Associative memory, traditionally modeled by Hopfield networks, enables the retrieval of previously stored patterns from partial or noisy cues. Yet, the local computational principles which are required to enable this function remain incompletely understood. To formally characterize the local information processing in such systems, we employ a recent extension of information theory -- Partial Information Decomposition (PID). PID decomposes the contribution of different inputs to an output into unique information from each input, redundant information across inputs, and synergistic information that emerges from combining different inputs. Applying this framework to individual neurons in classical Hopfield networks we find that below the memory capacity, the information in a neuron's activity is characterized by high redundancy between the external pattern input and the internal recurrent input, while synergy and unique information are close to zero until the memory capacity is surpassed and performance drops steeply. Inspired by this observation, we use redundancy maximization at each neuron as an information-theoretic learning goal. This dramatically increases the network's memory capacity to 1.59, a more than tenfold improvement over the 0.14 capacity of classical Hopfield networks, and also outperforming recent state-of-the-art implementations of Hopfield networks. Overall, this work establishes redundancy maximization as a new design principle for associative memories and opens pathways for new associative memory models based on information-theoretic goals.

cs.IT↗

arXiv2025

Inferring Mobility Reductions from COVID-19 Disease Spread along the Urban-Rural Gradient

The COVID-19 pandemic reshaped human mobility through policy interventions and voluntary behavioral changes. Mobility adaptions helped mitigate pandemic spread, however our knowledge which environmental, social, and demographic factors helped mobility reduction and pandemic mitigation is patchy. We introduce a Bayesian hierarchical model to quantify heterogeneity in mobility responses across time and space in Germany's 400 districts using anonymized mobile phone data. Decomposing mobility into a disease-responsive component and disease-independent factors (temperature, school vacations, public holidays) allows us to quantify the impact of each factor. We find significant differences in reaction to disease spread along the urban-rural gradient, with large cities reducing mobility most strongly. Employment sectors further help explain variance in reaction strength during the first wave, while political variables gain significance during the second wave. However, reduced mobility only partially translates to lower peak incidence, indicating the influence of other hidden factors. Our results identify key drivers of mobility reductions and demonstrate that mobility behavior can serve as an operational proxy for population response.

stat.AP↗

arXiv2025

Disentangling Interaction and Bias Effects in Opinion Dynamics of Large Language Models

Large Language Models are increasingly used to simulate human opinion dynamics, yet the effect of genuine interaction is often obscured by systematic biases. We develop a Bayesian framework to disentangle and quantify three such biases: (i) A topic bias toward the LLM's default stance; (ii) an agreement bias favoring agreement to the prompted statement irrespective of the question; and (iii) an anchoring bias toward the initiating agent's stance. We apply this framework to various LLMs that performed multi-step dialogues on 12 different questions from climate change and societal justice to music preferences. We find that opinion trajectories tend to quickly converge to a shared attractor, with the influence of both interaction and biases decaying over time, and with the impact of biases differing between LLMs. In addition, we show that fine-tuning an LLM on different sets of strongly opinionated statements (including misinformation) shifts the opinion attractor correspondingly. By exposing stark differences between LLMs and providing quantitative tools for comparing interaction and bias contributions to opinion shifts in LLM agent discussions, our approach highlights both promises and pitfalls of using LLMs as proxies for human behavior.

physics.soc-ph↗

arXiv2025

On the Statistical Physics of Wealth Distribution

Asset exchange models (AEMs) provide a physics-inspired framework for studying wealth formation. These models capture wealth distribution dynamics via pairwise money exchanges, yielding steady-state distributions from exponential to heavy-tailed power laws. However, empirical validation remains limited due to scarce real-world transaction data. Here, we bridge this gap by analyzing spectral properties of Markov transition matrices from both AEMs and Ethereum blockchain data, enabling quantitative comparison of model and empirical exchange dynamics. We assess thermodynamic equilibrium in exchange processes - specifically, detailed balance - and derive steady-state wealth distributions from transition matrices. We find that equilibrium systems' spectra contain only real eigenvalues and link Ethereum price changes to spectral shifts. We also investigate external factors (e.g., taxes), showing that advantages for richer individuals make wealth evolution path-dependent on initial distributions. Our work establishes a quantitative framework for validating AEMs with real data, advancing economic modeling and understanding of wealth formation.

physics.soc-ph↗

arXiv2025

Testing Paradox May Explain Increased Observed Prevalence of Bacterial STIs among MSM on HIV PrEP: A Modeling Study

HIV pre-exposure Prophylaxis (PrEP) has become essential for global HIV control, but its implementation coincides with rising bacterial STI rates among men who have sex with men (MSM). While risk-compensation behavioral changes like reduced condom use are frequently reported, we examine whether intensified asymptomatic screening in PrEP programs creates surveillance artifacts that could be misinterpreted. We developed a compartmental model to represent the simultaneous spread of HIV and chlamydia (as an example of a curable STI), integrating three mechanisms: 1) risk-mediated self-protective behavior, 2) condom use reduction post-PrEP initiation, and 3) PrEP-related asymptomatic STI screening. Increasing PrEP uptake may help to reduce chlamydia prevalence, only if the PrEP-related screening is frequent enough. Otherwise, the effect of PrEP can be disadvantageous, as the drop in self-protective actions caused by larger PrEP uptake cannot be compensated for. Additionally, the change in testing behavior may lead to situations where the trend in the number of positive tests is not a reliable sign of the actual dynamics. We found a plausible mechanism to reconcile conflicting observational evidence on the effect of PrEP on STI rates, showing that simultaneous changes in testing and spreading rates may generate conflicting signals, i.e., that observed trends increase while true prevalence decreases. Asymptomatic screening, together with personalized infection treatment to minimize putative pressure to generate antibiotic resistance, is one of the key determinants of the positive side effects of PrEP in reducing STI incidence.

q-bio.PE↗

arXiv2025

Neural mechanisms of predictive processing: a collaborative community experiment through the OpenScope program

This review synthesizes advances in predictive processing within the sensory cortex. Predictive processing theorizes that the brain continuously predicts sensory inputs, refining neuronal responses by highlighting prediction errors. We identify key computational primitives, such as stimulus adaptation, dendritic computation, excitatory/inhibitory balance and hierarchical processing, as central to this framework. Our review highlights convergences, such as top-down inputs and inhibitory interneurons shaping mismatch signals, and divergences, including species-specific hierarchies and modality-dependent layer roles. To address these conflicts, we propose experiments in mice and primates using in-vivo two-photon imaging and electrophysiological recordings to test whether temporal, motor, and omission mismatch stimuli engage shared or distinct mechanisms. The resulting dataset, collected and shared via the OpenScope program, will enable model validation and community analysis, fostering iterative refinement and refutability to decode the neural circuits of predictive processing.

q-bio.NC↗

arXiv2024

What should a neuron aim for? Designing local objective functions based on information theory

In modern deep neural networks, the learning dynamics of the individual neurons is often obscure, as the networks are trained via global optimization. Conversely, biological systems build on self-organized, local learning, achieving robustness and efficiency with limited global information. We here show how self-organization between individual artificial neurons can be achieved by designing abstract bio-inspired local learning goals. These goals are parameterized using a recent extension of information theory, Partial Information Decomposition (PID), which decomposes the information that a set of information sources holds about an outcome into unique, redundant and synergistic contributions. Our framework enables neurons to locally shape the integration of information from various input classes, i.e. feedforward, feedback, and lateral, by selecting which of the three inputs should contribute uniquely, redundantly or synergistically to the output. This selection is expressed as a weighted sum of PID terms, which, for a given problem, can be directly derived from intuitive reasoning or via numerical optimization, offering a window into understanding task-relevant local information processing. Achieving neuron-level interpretability while enabling strong performance using local learning, our work advances a principled information-theoretic foundation for local learning strategies.

cs.IT↗

arXiv2024

Signatures of hierarchical temporal processing in the mouse visual system

A core challenge for the brain is to process information across various timescales. This could be achieved by a hierarchical organization of temporal processing through intrinsic mechanisms (e.g., recurrent coupling or adaptation), but recent evidence from spike recordings of the rodent visual system seems to conflict with this hypothesis. Here, we used an optimized information-theoretic and classical autocorrelation analysis to show that information- and intrinsic timescales of spiking activity increase along the anatomical hierarchy of the mouse visual system, while information-theoretic predictability decreases. Moreover, the timescale hierarchy was invariant to the stimulus condition, whereas the decrease in predictability was strongest under natural movie stimulation. We could reproduce this effect in a basic recurrent network model with correlated sensory input. Our findings suggest that the rodent visual system indeed employs intrinsic mechanisms to achieve longer integration for higher cortical areas, while simultaneously reducing predictability for an efficient neural code.

q-bio.NC↗