About

Nicholas A. Christakis is the Director of the Human Nature Lab at Yale. The lab brings together scholars from various fields including medicine, engineering, sociology, economics, computer science, biology, statistics, applied math, and physics to explore diverse research areas. The lab provides an environment for researchers to engage with new approaches and methodologies in understanding human nature.

Research topics

• Computer Science
• Artificial Intelligence
• Sociology
• Psychology
• Geography
• Data science
• Demography
• Statistics
• World Wide Web
• Human–computer interaction
• Environmental health
• Biology
• Engineering
• Telecommunications
• Virology
• Engineering ethics
• Communication
• Medicine

Selected publications

• Countering the forgetting of novel health information with ‘social boosting’

  SSM - Population Health · 2026-01-24 · 1 citations

  articleOpen accessSenior author

  The prevalence of misleading information, especially with respect to health care practices, poses a threat. While studies have shown the effectiveness of various intervention techniques in mitigating the adverse effects of low-quality or false information, the effectiveness of such interventions can decay. Here, we investigate the role of the detailed social structure within which the intervened individuals live, which provides opportunities for the individuals to discuss and internalize new knowledge. We evaluated this with respect to information about maternal and child health care, delivered via a 22-month in-home intervention, among targeted individuals in 110 isolated Honduran villages. We hypothesize that individuals who receive specific knowledge can internalize and consolidate this information by engaging in social interactions where, for instance, they have an opportunity to discuss it with others. We found that well-connected individuals within a social network experience an enhanced effectiveness of knowledge interventions. These individuals may be more likely to internalize and retain the information and reinforce it in others, due to increased opportunities for social interaction where they teach or explain new knowledge, a mechanism we refer to as "social boosting". These findings underscore the role of social interactions in reinforcing health knowledge interventions. • Knowledge interventions often decay, limiting their long-term effectiveness. • Booster interventions help but are resource-intensive and challenging to scale. • Social networks offer scalable mechanism for reinforcement through peer interactions • Higher friendship ties intervened individuals exhibit significantly greater retention

  PublisherDOI

• Deep description of static and dynamic network ties in Honduran villages

  arXiv (Cornell University) · 2026-02-17

  preprintOpen accessSenior author

  We examine static and dynamic social network structure in 176 villages within the Copan Department of Honduras across two data waves (2016, 2019), using detailed data on multiplex networks for 20,232 individuals enrolled in a longitudinal survey. These networks capture friendship, health advice, financial help, and adversarial relationships, allowing us to show how cooperation and conflict jointly shape social structure. Using node-level network measures derived from near-census sociocentric village networks, we leverage mixed-effects zero-inflated negative binomial models to assess the influence of individual attributes, such as gender, marital status, education, religion, and indigenous status, and of village characteristics, on the dynamics of social networks over time. We complement these node-level models with dyadic assortativity (odds-ratio-based homophily) and community-level measures to describe how sorting by key attributes differs across network types and between waves. Our results demonstrate significant assortativity based on gender and religion, particularly within health and financial networks. Across networks, gender and religion exhibit the most consistent assortative mixing. Additionally, community-level assortativity metrics indicate that educational and financial factors increasingly influence social ties over time. Our findings provide insights into how personal attributes and community dynamics interact to shape network formation and socio-economic relationships in rural settings over time.

  PublisherDOI

• Rank Is Not Capacity: Spectral Occupancy for Latent Graph Models

  ArXiv.org · 2026-05-11

  articleOpen accessSenior author

  Graph representation learning has become a standard approach for analyzing networked data, with latent embeddings widely used for link prediction, community detection, and related tasks. Yet a basic design choice, the latent dimension, is still treated as a brittle hyperparameter, fixed before training and tuned by held-out performance. Learned factors are also identifiable only up to rotation and rescaling, so the nominal rank rarely coincides with the quantity that governs model behavior. We propose Spectral Prefix Extraction and Capacity-Targeted Representation Analysis (Spectra), which replaces rank as the unit of analysis with the spectrum of a learned positive semidefinite kernel, trace-normalized so that spectra are comparable across fits. The normalized eigenvalues form a distribution on the simplex, and their Shannon effective rank acts both as a summary of learned capacity and as a controllable training-time coordinate: a single scalar shapes this realized dimension during training, and bisection targets any desired value within the rank cap. To theoretically support that, we show local regularity and monotonicity of the realized-dimension profile. Across collaboration, social, biological, and infrastructure networks, Spectra traces performance--capacity frontiers that make the trade-off between predictive accuracy and realized dimension visible. It performs competitively with strong link-prediction baselines, yields aligned lower-capacity views of the same fitted model through spectral prefixes, and provides a principled handle on capacity in the overparameterized regime. Capacity thus becomes a property of the fitted model rather than a hyperparameter of the training.

  PublisherOA PDF

• Rank Is Not Capacity: Spectral Occupancy for Latent Graph Models

  arXiv (Cornell University) · 2026-05-11

  preprintOpen accessSenior author

  Graph representation learning has become a standard approach for analyzing networked data, with latent embeddings widely used for link prediction, community detection, and related tasks. Yet a basic design choice, the latent dimension, is still treated as a brittle hyperparameter, fixed before training and tuned by held-out performance. Learned factors are also identifiable only up to rotation and rescaling, so the nominal rank rarely coincides with the quantity that governs model behavior. We propose Spectral Prefix Extraction and Capacity-Targeted Representation Analysis (Spectra), which replaces rank as the unit of analysis with the spectrum of a learned positive semidefinite kernel, trace-normalized so that spectra are comparable across fits. The normalized eigenvalues form a distribution on the simplex, and their Shannon effective rank acts both as a summary of learned capacity and as a controllable training-time coordinate: a single scalar shapes this realized dimension during training, and bisection targets any desired value within the rank cap. To theoretically support that, we show local regularity and monotonicity of the realized-dimension profile. Across collaboration, social, biological, and infrastructure networks, Spectra traces performance--capacity frontiers that make the trade-off between predictive accuracy and realized dimension visible. It performs competitively with strong link-prediction baselines, yields aligned lower-capacity views of the same fitted model through spectral prefixes, and provides a principled handle on capacity in the overparameterized regime. Capacity thus becomes a property of the fitted model rather than a hyperparameter of the training.

  PublisherDOI

• Modeling roles and trade-offs in multiplex networks

  Nature Communications · 2026-03-07 · 2 citations

  articleOpen access

  Multiplex social networks capture multiple types of relations among the same people. Their structure reflects how exchanges arise from individual attributes related to independence, the status or resources of others related to dependence, and mutual influence related to interdependence. Understanding these systems is challenging because layers can play distinct yet complementary roles. We introduce the Multiplex Latent Trade-off Model, MLT, a framework for identifying roles in multiplex networks that incorporates independence, dependence, and interdependence. MLT represents roles as trade-offs, requiring each node to distribute source and target roles across layers while allocating community memberships within hierarchical structures. Applying MLT to 176 multiplex networks, including social, health, and economic layers from villages in western Honduras, we identify core principles of social exchange and reveal multi-scale communities. Link-prediction analyses show that modeling interdependence most improves predictions for social ties, whereas health and economic ties are shaped more strongly by individual status and behavior. People manage different relationships, including friendships and health-related and economic ties. The authors present a model that reveals how these layers of social life interact, showing that friendship ties rely strongly on interdependence, whereas health and economic ties are shaped more by status.

  PublisherOA PDFDOI

• Benchmark for Assessing Olfactory Perception of Large Language Models

  ArXiv.org · 2026-03-08

  articleOpen accessSenior author

  Here we introduce the Olfactory Perception (OP) benchmark, designed to assess the capability of large language models (LLMs) to reason about smell. The benchmark contains 1,010 questions across eight task categories spanning odor classification, odor primary descriptor identification, intensity and pleasantness judgments, multi-descriptor prediction, mixture similarity, olfactory receptor activation, and smell identification from real-world odor sources. Each question is presented in two prompt formats, compound names and isomeric SMILES, to evaluate the effect of molecular representations. Evaluating 21 model configurations across major model families, we find that compound-name prompts consistently outperform isomeric SMILES, with gains ranging from +2.4 to +18.9 percentage points (mean approx +7 points), suggesting current LLMs access olfactory knowledge primarily through lexical associations rather than structural molecular reasoning. The best-performing model reaches 64.4\% overall accuracy, which highlights both emerging capabilities and substantial remaining gaps in olfactory reasoning. We further evaluate a subset of the OP across 21 languages and find that aggregating predictions across languages improves olfactory prediction, with AUROC = 0.86 for the best performing language ensemble model. LLMs should be able to handle olfactory and not just visual or aural information.

  PublisherOA PDF

• Social networks and international migration in Honduran villages

  Proceedings of the National Academy of Sciences · 2026-01-02

  articleOpen access

  Social relationships are central to shaping international migration patterns, yet the link between widescale network structure and mobility decisions remains poorly understood. Here, we investigate two key mechanisms by which social networks influence migration behavior: transmission of information and resources, and comparison of social status. These mechanisms suggest distinct sets of alters that an ego may emulate with respect to their migration behaviors, resulting in divergent mobility trajectories within and across communities. Leveraging longitudinal data from 73 Honduran villages ([Formula: see text] individuals) over six years, we use a Linear Network Autocorrelation Modeling framework to disentangle the effects of kinship, friendship, and economic ties on international migration decisions. Our findings reveal that incorporating social network factors as predictors significantly improves model fit. While indicators for resource-sharing processes substantially contribute to model performance, the inclusion of structural comparison mechanisms does not provide additional explanatory power. These results underscore the critical role of information and resource transmission within social networks in facilitating migration behaviors.

  PublisherDOI

• On the optimal integration of intelligent agents into network systems to steer cooperation

  Proceedings of the National Academy of Sciences · 2026-03-16

  articleOpen accessSenior authorCorresponding

  Sociotechnical networks, in which humans and technologies act as interacting entities (also known as "hybrid systems"), increasingly face perturbations by automated agents. What this implies for immersive steering of collective behavior, and how this shapes the stability and resilience of cooperation, remain unclear. Here, we extend evolutionary graph theory by incorporating a distinct type of node representing embedded intelligent agents, namely, algorithmic nodes that autonomously implement prescribed behavioral responses during interactions. These agents are randomly placed within a social network and exert local influence in their neighborhood in social dilemma games. Individual behavior changes are driven by evolutionary dynamics. We derive closed-form analytical results characterizing evolutionary stability and long-run cooperation levels, and show that there exists an optimal, intermediate prevalence of intelligent agents that best promotes cooperation. Our work offers insights into the optimal alignment of human populations with respect to the social good using intelligent agents.

  PublisherDOI

• Benchmark for Assessing Olfactory Perception of Large Language Models

  arXiv (Cornell University) · 2026-03-08

  preprintOpen accessSenior author

  Here we introduce the Olfactory Perception (OP) benchmark, designed to assess the capability of large language models (LLMs) to reason about smell. The benchmark contains 1,010 questions across eight task categories spanning odor classification, odor primary descriptor identification, intensity and pleasantness judgments, multi-descriptor prediction, mixture similarity, olfactory receptor activation, and smell identification from real-world odor sources. Each question is presented in two prompt formats, compound names and isomeric SMILES, to evaluate the effect of molecular representations. Evaluating 21 model configurations across major model families, we find that compound-name prompts consistently outperform isomeric SMILES, with gains ranging from +2.4 to +18.9 percentage points (mean approx +7 points), suggesting current LLMs access olfactory knowledge primarily through lexical associations rather than structural molecular reasoning. The best-performing model reaches 64.4\% overall accuracy, which highlights both emerging capabilities and substantial remaining gaps in olfactory reasoning. We further evaluate a subset of the OP across 21 languages and find that aggregating predictions across languages improves olfactory prediction, with AUROC = 0.86 for the best performing language ensemble model. LLMs should be able to handle olfactory and not just visual or aural information.

  PublisherDOI

• How malicious AI swarms can threaten democracy

  Science · 2026-01-22 · 10 citations

  articleOpen access

  The fusion of agentic AI and LLMs marks a new frontier in information warfare.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R01AG017548

  NIH · $3.0M · 2007

• NIH Grant R01AG015326

  NIH · $910k · 2002

• NIH Grant P01AG031093

  NIH · $21.6M · 2014

• NIH Grant R01AG024448

  NIH · $2.0M · 2009

• NIH Grant P30AG034420

  NIH · $6.0M · 2020

Frequent coauthors

• Richard L. Schilsky

  American Society of Clinical Oncology

  137 shared

• James E. Herndon

  Duke University

  136 shared

• Jane C. Weeks

  King Edward Memorial Hospital

  136 shared

• I. Craig Henderson

  135 shared

• Elizabeth B. Lamont

  Medidata (United States)

  133 shared

• James H. Fowler

  University of California, San Diego

  100 shared

• Craig C. Earle

  81 shared

• Rogério Lilenbaum

  Yale University

  55 shared

Labs

• Human Nature LabPI

  We bring together scholars from throughout Yale in many fields, including medicine, engineering, sociology, economics, computer science, biology, statistics, applied math, and physics. The Human Nature Lab provides a fertile environment for our researchers to engage new approaches and methodologies.

Education

• Ph.D., Social Science

  Harvard University

  1995

• M.D.

  Harvard Medical School

  1990

• Other

  Harvard College

  1984