Contact Information

Contact Information

PhD Candidate, Demographer, Sociologist.

Sitemap

A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.

Pages

Posts

New Preprint: Model Diversity Over Model Size — Unanimous LLM Ensembles Correct Over-Classification in Survey Coding

1 minute read

Published:

I have a new preprint out on SocArXiv: Model Diversity Over Model Size: Unanimous LLM Ensembles Correct Over-Classification in Survey Coding. It’s a deep dive into one finding flagged in the CatLLM methods paper — that unanimous multi-model ensembling corrects over-classification — asking which ensemble ingredients actually drive the gain and where the gain shows up.

Presenting CatLLM at the University of Washington

3 minute read

Published:

On October 29th, 2025, I had the privilege of presenting at the University of Washington on how large language models can augment social science research. The presentation focused on CatLLM, an open-source Python package I developed to address a common challenge in demographic and social science research: analyzing open-ended survey responses and complex data at scale.

The Hispanic Health Paradox in Cognitive Aging

2 minute read

Published:

On July 1st, 2025, I had the opportunity to present joint research with Dr. William H. Dow, Professor of Health Policy and Management and Director of the Center on the Economics and Demography of Aging at UC Berkeley. Our presentation explored a fascinating contradiction in Hispanic aging: the Hispanic health paradox in cognitive outcomes.

I had a great time at PAA 2025

1 minute read

Published:

I presented shared work with Professor William Dow and Henry Dow, comparing sociodemographic characteristics of older Hispanic adults in U.S. immigrant populations with those in their countries of origin, using census microdata and the American Community Survey (publication pending).

I presented at the “Ask a Science Envoy” Event!

less than 1 minute read

Published:

I recently presented my research at the Science Envoy “Ask a Science Envoy” event at HopMonk Tavern in Novato—my first time presenting at a brewery and to a non-academic audience. The experience taught me the importance of navigating sensitive topics like partisanship and belief in science, especially when audience members began debating the subject.

Understanding the Three-Group SIR Step by Step

21 minute read

Published:

In our upcoming paper, “Measuring and Modeling the Impact of Partisan Differences in Health Behaviors on COVID-19 Dynamics,” we use a three-group Susceptible-Infected-Recovered model to highlight the importance of incorporating partisan differences into models of disease transmission. In this blog post, I want to fully explain what is happening in the background for readers who may be interested in utilizing it themselves. For those users, we also built an R shiny app (soon to be published as well). The link to the shiny app will be: here.

Field Data Collection in Puerto Rico

less than 1 minute read

Published:

I’m super grateful to the Puerto Rican team of The Caribbean American Dementia and Aging Study (CADAS). We got a lot done and things are moving quickly! Looking forward to visiting again soon.

Presenting our Research for the Center for Studies in Demography & Ecology at University of Washington, Seattle

1 minute read

Published:

The impact of non-pharmaceutical interventions and protective health behaviors, such as the use of face masks and physical distancing, on COVID-19 dynamics is well-documented, but sub-group heterogeneities in the adoption of these behaviors remains understudied. In this paper, we describe partisan differences in the adoption of protective health behaviors, and model how these differences can impact the dynamics of COVID-19.

How to Improve Match Quality on String Data Using Large Language Models

19 minute read

Published:

In the world of data analysis, ensuring the accuracy and consistency of datasets is crucial, especially when dealing with entities like school names that may be spelled differently across various sources. This discrepancy can pose significant challenges when trying to match records from different datasets. Traditional methods of data cleaning may fall short in addressing these inconsistencies effectively.

First Post in Honor of My Little Man, Charlie

less than 1 minute read

Published:

This first ever blog post is in honor of my recently passed little man, Charlie the cat, who for 12 years went together with me through thick and thin. He was the best cat anyone could’ve asked for, and no other will ever be able to replace him. I will miss you dearly, Charles, aka “Chester.” I love you always.

portfolio

publications

Connecting Fathers: Fathers’ Impact on Adult Children’s Social Networks

Published in The International Journal of Aging and Human Development, 2022

This study examines how having an emotionally close and active father in an adult child’s social network shapes network composition, using data from the 2015 UC Berkeley Social Networks Study (UCNets).

Recommended citation: Soria C, Lawton L. Connecting Fathers: Fathers Impact on Adult Childrens Social Networks. The International Journal of Aging and Human Development. 2023;96(1):19-32. doi:10.1177/00914150221106645 https://journals.sagepub.com/doi/full/10.1177/00914150221106645

Commentary: Examining Contextual Factors Contributing to Differentials in COVID-19 Mortality in U.S. vs. India

Published in Frontiers in Public Health, 2022

This commentary examines the disparities in COVID-19 mortality rates between the U.S. and India, exploring demographic dynamics and contextual factors contributing to the “Indian death paradox.”

Recommended citation: Zanwar PP, Wallace KL, Soria C, Perianayagam A. Commentary: Examining contextual factors contributing to differentials in COVID-19 mortality in U.S. vs. India. Front Public Health. 2022;10:995751. doi:10.3389/fpubh.2022.995751 https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2022.995751/full

Social Network Structure Rivals Smoking and Income as a Predictor of U.S. County Mortality

Published in SocArXiv, 2025

This study examines how US county-level social network structure relates to mortality disparities using measures from 21 billion Facebook friendships.

Recommended citation: Soria C, Feehan DM. Social Network Structure Rivals Smoking and Income as a Predictor of U.S. County Mortality. SocArXiv. 2025. https://osf.io/preprints/socarxiv/kvmx6_v3 https://osf.io/preprints/socarxiv/kvmx6_v3

Assessing the 10/66 Dementia Classification Algorithm for International Comparative Analyses with the U.S.

Published in American Journal of Epidemiology, 2025

Cross-national comparisons of dementia prevalence are essential for identifying unique determinants and cultural-specific risk factors, but methodological differences in dementia classification across countries hinder global comparisons. This study maps the 10/66 algorithm for dementia classification, widely used and validated in low- and middle-income countries (LMICs), to the U.S. Aging, Demographics, and Memory Study (ADAMS), the dementia sub-study of the Health and Retirement Study, and assesses its performance in ADAMS.

Recommended citation: Jorge J Llibre Guerra, Jordan Weiss, Jing Li, Chris Soria, Ana Rodriguez-Salgado, Juan de Jesús Llibre Rodriguez, Ivonne Z Jiménez Velázquez, Daisy Acosta, Mao-Mei Liu, William H Dow, Assessing the 10/66 Dementia Classification Algorithm for International Comparative Analyses with the U.S., American Journal of Epidemiology, 2024;, kwae470, https://doi.org/10.1093/aje/kwae470 https://pubmed.ncbi.nlm.nih.gov/39745806/

The Caribbean American Dementia and Aging Study: Protocol for a Population-Based Study of Older Adult Health and Dementia in Cuba, the Dominican Republic, and Puerto Rico

Published in BMC Geriatrics, 2025

CADAS is a multi-purpose household study of aging focused on the life course determinants and consequences of health and dementia in Puerto Rico, Dominican Republic, and Cuba.

Recommended citation: Liu MM, Llibre-Guerra J, Soria C, Li J, Zayas Llerena T, Rodriguez G, Acosta D, Jiménez Velázquez I, Llibre-Rodriguez JJ, Dow WH. The Caribbean American Dementia and Aging Study: protocol for a population-based study of older adult health and dementia in Cuba, the Dominican Republic, and Puerto Rico. BMC Geriatr. 2025;25(1). doi:10.1186/s12877-025-06131-0 https://bmcgeriatr.biomedcentral.com/articles/10.1186/s12877-025-06131-0

Partisan differences in health behaviors can impact respiratory disease dynamics

Published in medRxiv, 2026

This study examines how partisan differences in contact rates, mask usage, and vaccination patterns shape respiratory disease transmission dynamics.

Recommended citation: Soria C, Dorelien A, Feehan D, Mahmud A. Partisan differences in health behaviors can impact respiratory disease dynamics. medRxiv. 2026. doi:10.64898/2026.01.14.26344076 https://www.medrxiv.org/content/10.64898/2026.01.14.26344076v1

Scaling Open-Ended Survey Coding: An LLM Pipeline Where Definitions Do the Heavy Lifting

Published in SocArXiv, 2026

CatLLM is an open-source Python and R package offering a three-stage pipeline for coding open-ended survey responses with large language models, with defaults calibrated by a systematic empirical study evaluating 21 LLMs across three capability tiers, six providers, and four survey questions.

Recommended citation: Soria C. Scaling Open-Ended Survey Coding: An LLM Pipeline Where Definitions Do the Heavy Lifting. SocArXiv. 2026. https://osf.io/preprints/socarxiv/gjvcf_v1 https://osf.io/preprints/socarxiv/gjvcf_v1

CatLLM: A Python package for Generating, Assigning, and Scoring Open-Ended Survey Data and Images

Published in Journal of Open Source Software, 2026

Peer-reviewed software paper introducing CatLLM, an open-source Python and R toolkit for reproducible LLM-powered text classification, with defaults calibrated against expert human coders across multiple survey datasets.

Recommended citation: Soria C. CatLLM: A Python package for Generating, Assigning, and Scoring Open-Ended Survey Data and Images. Journal of Open Source Software. 2026. doi:10.21105/joss.09678 https://doi.org/10.21105/joss.09678

High Agreement, Different Stories: How LLM Classifiers Reshape Demographic Patterns in Survey Data

Published in SocArXiv, 2026

A multilabel evaluation of eight LLMs against human coders on 3,200 open-ended responses from the UC Berkeley Social Networks Study, showing how high per-category agreement can mask thematic over-classification and demographic divergence.

Recommended citation: Soria C. High Agreement, Different Stories: How LLM Classifiers Reshape Demographic Patterns in Survey Data. SocArXiv. 2026. https://osf.io/preprints/socarxiv/85kyd_v1 https://osf.io/preprints/socarxiv/85kyd_v1

Model Diversity Over Model Size: Unanimous LLM Ensembles Correct Over-Classification in Survey Coding

Published in SocArXiv, 2026

Across 16 LLMs and four open-ended survey questions, unanimous voting across diverse models corrects over-classification on subjectively ambiguous categories, with cross-provider diversity—not temperature or within-family size—as the active ingredient. As few as three diverse lower-tier models can reliably exceed GPT-5.

Recommended citation: Soria C. Model Diversity Over Model Size: Unanimous LLM Ensembles Correct Over-Classification in Survey Coding. SocArXiv. 2026. https://osf.io/preprints/socarxiv/er6mz_v1 https://osf.io/preprints/socarxiv/er6mz_v1

Prevalence of Dementia and Modifiable Risk Factors among Older Adults in Cuba: A Population-Based Study

Published in Neuroepidemiology, 2026

A population-based study estimating the prevalence of dementia and associated modifiable risk factors among older adults in Cuba. Accepted for publication in Neuroepidemiology.

Recommended citation: Llibre-Rodriguez JJ, Zayas T, Llibre-Guerra JJ, Diaz N, Martinez Z, Santos A, Liu MM, Soria C, Acosta D, Jiménez-Velázquez IZ, Dow WH. Prevalence of Dementia and Modifiable Risk Factors among Older Adults in Cuba: A Population-Based Study. Neuroepidemiology. 2026. In press.

Why Would They Help Me?: Reasons People Expect (or Do Not Expect) Social Support

Published in Social Psychology Quarterly, 2026

A study of the reasons people give for expecting—or not expecting—social support from family and friends, drawing on the UC Berkeley Social Networks Study (UCNets).

Recommended citation: Fischer CS, Soria C. Why Would They Help Me?: Reasons People Expect (or Do Not Expect) Social Support. Social Psychology Quarterly. 2026. In press.

talks

Father Connection & Support in Adulthood (UC Berkeley Social Networks Study)

Published:

Despite rapidly expanding interest in fathers, scholars know little about the impact of fathers on adult health and well-being. How does the positive presence – or lack thereof - of fathers in adult children’s lives affect their social networks? Drawing on attachment theory and social capital theory to examine novel UC Berkeley Social Networks Study (UC Nets) data, I seek to extend understanding of how father attachment and socialization can influence adult social well-being. I find that individuals who name a father in their social network have significantly more social ties. Those with a “close” father have larger social networks than those who name a father who is not “close.” This has consequences for individuals’ broader networks: respondents with a “close” father report more males – but not females – in social activities networks. In contrast, having a “close” mother was associated with more females to confide in, but not males.

Evaluating LLM Accuracy for Survey Data Classification

Published:

This talk examined how accurately leading large language models classify open-ended survey responses compared to human coders. Findings suggest that LLMs are best used to augment human coders in a loop rather than as a sole source of classification.

Partisan Context and Its Associations on Individual Behavior During the COVID-19 Pandemic

Published:

Joint presentation with Dr. Audrey Dorélien examining partisan differences in the adoption of protective health behaviors during COVID-19 and modeling how these differences impact disease dynamics. Using detailed survey data on partisanship, contact rates, mask usage, and vaccination rates, we demonstrated that partisan differences in health behaviors exceed racial and gender differences. We incorporated these observations into a Susceptible-Infected-Recovered (SIR) model framework that explicitly incorporates partisanship to identify the most significant mechanisms driving disease spread, emphasizing the importance of considering partisan identification in public health policy and pandemic preparedness.

Partisan-Influenced Health Outcomes

Published:

Radio interview as a Wonderfest Science Envoy on KPOO’s “Let Me Touch Your Mind” show, hosted by Marilynn Fowler. Discussed research on partisan-influenced health outcomes and how political partisanship shapes public health behaviors and disease spread.

Partisan Differences in the Spread of Disease

Published:

Presented as a Wonderfest “Science Envoy” at “Ask a Science Envoy: Anthropocene Alarm; Partisan Contagion,” a public science event at HopMonk Tavern co-hosted with Marin Science Seminar, alongside UC Berkeley biologist Kristy Mualim (Genetic Biodiversity Loss in the Anthropocene). The talk explored how political partisanship significantly influences how different groups respond to public health guidance, affecting their adherence to protective measures against infectious diseases. Understanding partisan-based disparities in acceptance of scientific evidence helps us to predict the population-wide spread of diseases and to improve public health outcomes.

Benefits and Challenges of AI in Demography

Published:

Participated in a panel discussion on the benefits and challenges of AI in demography at the 2025 PAA Annual Meeting. Discussed use-cases and benefits/limitations of large language models (LLMs) for survey research alongside other panelists.

Hispanic Migrant Cognitive Aging

Published:

Joint presentation with Dr. William H. Dow examining the Hispanic health paradox in cognitive aging. The talk explored the contradiction in Hispanic aging where Mexican migrants, despite being more socioeconomically disadvantaged, demonstrate better cognitive aging outcomes than other Hispanic migrant groups. This research contributes to understanding health disparities and the complex relationship between socioeconomic status and health outcomes in aging populations.

CatLLM: Augmenting Social Science Research with Large Language Models

Published:

Presented on computational tools that augment social science research using large language models. Introduced CatLLM—an open-source Python package that enables researchers to apply language and vision models to survey coding, image analysis, and data categorization without machine learning expertise. The talk covered practical applications in demographic research, including automated coding of open-ended survey responses and analysis of cognitive health data.

Automating Survey Analysis with Large Language Models: A Hands-On Workshop Using CatLLM

Published:

Hands-on workshop on computational tools that augment social science research using large language models, presented as part of Session 103, “Innovative Data Methods and AI Applications,” at the 2026 Population Association of America (PAA) Applied Demography Conference. Introduced CatLLM—an open-source Python package that enables researchers to apply language and vision models to survey coding, image analysis, and data categorization without machine learning expertise. The workshop covered practical applications in demographic research, including automated coding of open-ended survey responses and analysis of cognitive health data.

The Role of Social Structure in County-Level Mortality Disparities

Published:

Personal networks influence health and mortality at the individual level, but less is known about how population-scale social network structure relates to mortality. This study examines how US county-level social network structure relates to mortality disparities. Using measures from 21 billion Facebook friendships, we investigate how two structural features of population social networks — cohesiveness and diversity — are associated with age-standardized and age-specific mortality rates. Bivariate results show that measures of social network structure rival smoking rates, median income, and educational attainment in their association with mortality rates. Social network structure remains predictive of mortality even after controlling for traditional measures like socioeconomic status and rural/urban classification. Network diversity is associated with lower mortality in both bivariate and multivariate analyses. Network clustering is associated with higher mortality bivariately, but this association reverses after controlling for county-level demographic and socioeconomic factors, revealing a protective effect masked by confounding. Age-stratified analyses further complicate this picture, showing that clustering predicts lower mortality among adults aged 15-64 but higher mortality among those 70 and older. These findings highlight social network structure as an important dimension of place-based health disparities, one not fully captured by conventional measures of socioeconomic composition or spatial segregation.

An Empirical Investigation into the Utility of Large Language Models in Open-Ended Survey Data Categorization

Published:

How effectively can Large Language Models (LLMs) approximate social scientist judgment in categorizing open-ended survey responses? This study compares eight contemporary LLMs—GPT-5, Claude Sonnet 4.5, Gemini 2.5 Flash, Grok 4 Fast, Qwen 3, DeepSeek v3.1, Llama 4, and Mistral Medium—to human annotators on 3,208 responses from the UC Berkeley Social Networks Study, spanning four question types and yielding 19,248 multi-label coding decisions. Models do not reach human-like inter-rater reliability in comparisons with individual coders (Krippendorff’s alpha 0.58–0.59 vs. 0.77 for humans), yet they achieve high accuracy rates of 82–97% relative to a human consensus standard, depending on task complexity. Accuracy declines for longer, more ambiguous responses and for rare thematic categories, indicating that model performance is sensitive to both response length and category prevalence. Demographic differences in performance are present—for example, responses from female respondents are classified less accurately—but much of this gap is associated with differences in response style, such as length and complexity, rather than clearly attributable to direct demographic targeting.

Scaling Open-Ended Survey Coding: Definitions, Ensembles, and the Limits of Prompt Engineering

Published:

As large language model (LLM)–based text classification becomes routine in the social sciences, researchers confront dozens of competing models, inconsistent advice on prompting, and little standardized tooling with evidence-based defaults. CatLLM, an open-source Python and R package, addresses this gap with a three-stage pipeline — exploration, extraction, classification — for coding open-ended survey responses. The package supports multi-model ensembles, batch processing, and fully local deployment via open-weight models, allowing researchers working with sensitive data to avoid transmitting responses to external servers.

Social Isolation, Loneliness, and Post-COVID Cognitive Decline in Older Adults

Published:

Using four waves of the Health and Retirement Study (2016-2022), we document a sharp post-COVID cognitive decline in a nationally representative panel of older adults. Mean performance on the 27-point cognitive composite was flat across three pre-pandemic waves, then fell by approximately 0.7-0.9 points in 2022 — three to four times larger than any prior wave-to-wave fluctuation. This decline replicated across the full balanced panel and two non-overlapping Leave Behind cohorts, ruling out sample attrition and rotation-group assignment as explanations. Descriptively, both loneliness and structural social isolation were associated with lower baseline cognition and steeper decline. Loneliness showed an acute spike at COVID onset followed by partial recovery; structural isolation rose monotonically through 2022 with no sign of attenuation. Preliminary fixed-effects models provide initial causal evidence that within-person increases in structural isolation predict within-person cognitive decline (b = -0.30, p < .001). The UCLA loneliness scale adds an independent negative association (b = -0.47, p < .001). These estimates survive elimination of all stable individual traits via person fixed effects, strengthening causal inference. Whether isolation’s cognitive penalty was specifically amplified by COVID — beyond individual illness and mortality — awaits a fuller model with area-level contextual controls.

Survey Data Collection in Social Science

Published:

Contributed to the Sage Research Methods Video collection Doing Survey Research, produced by Sage Publishing as a teaching and training resource for students and researchers. In the interview I discussed dementia classification using survey data, the use of large language models for classifying open-ended survey responses, and approaches to quality control in survey data workflows.

A Multinational Framework for Online Cognitive Impairment Diagnosis in the Caribbean American Dementia and Aging Study

Published:

Presented as part of a Gateway/HCAP consensus session at AAIC on the Caribbean American Dementia and Aging Study (CADAS). Twenty-seven clinicians from five countries (Dominican Republic, Mexico, Puerto Rico, the United States, and India) independently classified 300 older adults using CADAS survey data, applying a structured 10-question NIA-AA diagnostic algorithm through a web-based consensus platform. Each case was reviewed by at least three clinicians per country team; cases with within-country disagreement underwent moderated consensus discussion, and international consensus was defined by majority rule across all five teams.

Advances in Modeling Human Behavior and Infectious Disease Spread

Published:

Recent infectious disease outbreaks highlighted the critical role of human behavior and social processes in shaping infectious disease dynamics. Individual decisions—such as whether to wear a mask, get vaccinated, or work remotely—are not made in isolation, but are strongly influenced by the current state of the disease and individuals’ perceptions of risk. Such behavioral responses are also not fixed and change over the course of a pandemic. As these behavior changes accumulate across individuals, they scale up to shape population-level outcomes, including the timing, magnitude, and geographic spread of outbreaks. In order to curtail disease spread with minimal impact to a population, governmental interventions to control epidemics must therefore account for individual decision-making and, particularly, noncompliance or even defiance of recommended or even mandated policies. Modeling techniques developed to account for such dynamic human responses are therefore essential for understanding and managing infectious disease spread.

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.

Teaching experience 2

Workshop, University 1, Department, 2015

This is a description of a teaching experience. You can use markdown like any other post.