arXiv subjects

Serina Chang

Publications and source records attributed to Serina Chang.

17 recordsLinked to original sources

arXiv2026

Anamnesis: An Open-Source Platform for Large-Scale Backstory-Conditioned Survey Simulation

We present Anamnesis, an interactive system for demographically controllable survey simulation using large language models. Open-source and designed for non-technical users/researchers, Anamnesis enables the prototyping and stress-testing of survey instruments on virtual populations rather than real human subjects. The platform operationalizes the recently introduced Anthology and Alterity frameworks, which use structured narrative backstories to condition model responses, within a unified web interface. It supports open-ended generation, probabilistic demographic resampling, and multimodal (image and audio) surveys. We evaluate the system through two case studies: (1) replicating segments of Pew Research Center's American Trends Panel (ATP) on political typology and biomedical issues and (2) emulating human preference in the New Yorker Caption Contest. In both cases, Anamnesis produces opinion distributions that more closely match real-world survey data than standard persona-prompting baselines, offering a transparent, reproducible, and open-source alternative to proprietary simulation services.

cs.CL↗

arXiv2026

Political Neutrality as Balanced Approval: A Large-Scale Human Evaluation of AI Responses

As AI systems increasingly shape political views, defining and evaluating AI political neutrality is an urgent problem. Here, we propose a new definition of AI political neutrality and design a large-scale user study to test it, releasing a new dataset PARETO with 7,434 participants and 208,152 evaluations of AI responses. Our definition follows a simple principle grounded in political theory: when asked about a controversial issue, an AI model should generate responses that maximize approval across groups with opposing viewpoints, while balancing approval between groups. This definition allows empirical testing of whether an AI response is "neutral" and generalizes to any political context without pre-supposing a single left-right axis of division. We construct a benchmark of controversial U.S. issues, with prompts sourced from politically charged questions on Reddit and responses from frontier AI models, and recruit human participants to rate AI responses. Across all 20 issues, we find that it is possible for AI responses to achieve high rates of approval on both sides, even as those sides disagree strongly with each other on the substance of the issues. We also find that default responses lean liberal for GPT, Gemini, Claude, and Llama, but not Grok, and that user prompts with political charges are harder to respond to than neutral prompts. This work introduces a rigorous definition and benchmark of AI political neutrality, and a dataset to measure progress toward it.

cs.CY↗

arXiv2026

Quantifying the Utility of User Simulators for Building Collaborative LLM Assistants

User simulators are increasingly leveraged to build interactive AI assistants, yet how to measure the quality of these simulators remains an open question. In this work, we show how simulator quality can be quantified in terms of its downstream utility: how an LLM assistant trained with this user simulator performs in the wild when interacting with real humans. In a controlled experiment where only the user simulator varies, we train LLM assistants via reinforcement learning against a spectrum of simulators, from an LLM prompted to role-play a user to one fine-tuned on human utterances from WildChat. As evaluation, we measure pairwise win rates in a user study with 283 participants and on WildBench, a benchmark derived from real human--AI conversations. Training against the role-playing LLM yields an assistant statistically indistinguishable from the initial assistant in our user study (51% win rate), whereas training against the fine-tuned simulator yields significant gains (58% over the initial and 57% over the one trained against role-playing). Closer inspection reveals three further patterns: methods for making role-playing LLMs more realistic (e.g., persona conditioning) improve trained assistants but do not close the gap to the fine-tuned simulator; scaling the simulator's model size benefits the fine-tuned simulator but yields no gain for role-playing ones; and assistants trained against role-playing simulators fail to generalize when paired with other simulators at test time, while the one trained against fine-tuned simulator does. Together, these results argue for grounding user simulators in real human behavior and measuring their quality by their downstream effect on real users.

cs.CL↗

arXiv2026

Learning Demographic-Conditioned Mobility Trajectories with Aggregate Supervision

Human mobility trajectories are widely studied in public health and social science, where different demographic groups exhibit significantly different mobility patterns. However, existing trajectory generation models rarely capture this heterogeneity because most trajectory datasets lack demographic labels. To address this gap in data, we propose ATLAS, a weakly supervised approach for demographic-conditioned trajectory generation using only (i) individual trajectories without demographic labels, (ii) region-level aggregated mobility features, and (iii) region-level demographic compositions from census data. ATLAS trains a trajectory generator and fine-tunes it so that simulated mobility matches observed regional aggregates while conditioning on demographics. Experiments on real trajectory data with demographic labels show that ATLAS substantially improves demographic realism over baselines (JSD $\downarrow$ 12%--69%) and closes much of the gap to strongly supervised training. We further develop theoretical analyses for when and why ATLAS works, identifying key factors including demographic diversity across regions and the informativeness of the aggregate feature, paired with experiments demonstrating the practical implications of our theory. We release our code at https://github.com/schang-lab/ATLAS.

cs.LG↗

arXiv2025

Graph-Based Alternatives to LLMs for Human Simulation

Large language models (LLMs) have become a popular approach for simulating human behaviors, yet it remains unclear if LLMs are necessary for all simulation tasks. We study a broad family of close-ended simulation tasks, with applications from survey prediction to test-taking, and show that a graph neural network can match or surpass strong LLM-based methods. We introduce Graph-basEd Models for Human Simulation (GEMS) which formulates close-ended simulation as link prediction on a heterogeneous graph of individuals and choices. Across three datasets and three evaluation settings, GEMS matches or outperforms the strongest LLM-based methods while using three orders of magnitude fewer parameters. These results suggest that graph-based modeling can complement LLMs as an efficient and transparent approach to simulating human behaviors. Code is available at https://github.com/schang-lab/gems.

cs.CL↗

arXiv2025

Valid Survey Simulations with Limited Human Data: The Roles of Prompting, Fine-Tuning, and Rectification

Surveys provide valuable insights into public opinion and behavior, but their execution is costly and slow. Large language models (LLMs) have been proposed as a scalable, low-cost substitute for human respondents, but their outputs are often biased and yield invalid estimates. We study the interplay between synthesis methods that use LLMs to generate survey responses and rectification methods that debias population estimates, and explore how human responses are best allocated between them. Using two panel surveys with questions on nutrition, politics, and economics, we find that synthesis alone introduces substantial bias (24-86%), whereas combining it with rectification reduces bias below 5% and increases effective sample size by up to 14%. Overall, we challenge the common practice of using all human responses for fine-tuning, showing that under a fixed budget, allocating most to rectification results in far more effective estimation.

cs.CL↗

arXiv2025

ChatBench: From Static Benchmarks to Human-AI Evaluation

With the rapid adoption of LLM-based chatbots, there is a pressing need to evaluate what humans and LLMs can achieve together. However, standard benchmarks, such as MMLU, measure LLM capabilities in isolation (i.e., "AI-alone"). Here, we design and conduct a user study to convert MMLU questions into user-AI conversations, by seeding the user with the question and having them carry out a conversation with the LLM to answer their question. We release ChatBench, a new dataset with AI-alone, user-alone, and user-AI data for 396 questions and two LLMs, including 144K answers and 7,336 user-AI conversations. We find that AI-alone accuracy fails to predict user-AI accuracy, with significant differences across multiple subjects (math, physics, and moral reasoning), and we analyze the user-AI conversations to provide insight into how they diverge from AI-alone benchmarks. Finally, we show that fine-tuning a user simulator on a subset of ChatBench improves its ability to estimate user-AI accuracies, increasing correlation on held-out questions by more than 20 points, creating possibilities for scaling interactive evaluation.

cs.CL↗

arXiv2025

Language Model Fine-Tuning on Scaled Survey Data for Predicting Distributions of Public Opinions

Large language models (LLMs) present novel opportunities in public opinion research by predicting survey responses in advance during the early stages of survey design. Prior methods steer LLMs via descriptions of subpopulations as LLMs' input prompt, yet such prompt engineering approaches have struggled to faithfully predict the distribution of survey responses from human subjects. In this work, we propose directly fine-tuning LLMs to predict response distributions by leveraging unique structural characteristics of survey data. To enable fine-tuning, we curate SubPOP, a significantly scaled dataset of 3,362 questions and 70K subpopulation-response pairs from well-established public opinion surveys. We show that fine-tuning on SubPOP greatly improves the match between LLM predictions and human responses across various subpopulations, reducing the LLM-human gap by up to 46% compared to baselines, and achieves strong generalization to unseen surveys and subpopulations. Our findings highlight the potential of survey-based fine-tuning to improve opinion prediction for diverse, real-world subpopulations and therefore enable more efficient survey designs. Our code is available at https://github.com/JosephJeesungSuh/subpop.

cs.CL↗

arXiv2024

LLMs generate structurally realistic social networks but overestimate political homophily

Generating social networks is essential for many applications, such as epidemic modeling and social simulations. The emergence of generative AI, especially large language models (LLMs), offers new possibilities for social network generation: LLMs can generate networks without additional training or need to define network parameters, and users can flexibly define individuals in the network using natural language. However, this potential raises two critical questions: 1) are the social networks generated by LLMs realistic, and 2) what are risks of bias, given the importance of demographics in forming social ties? To answer these questions, we develop three prompting methods for network generation and compare the generated networks to a suite of real social networks. We find that more realistic networks are generated with "local" methods, where the LLM constructs relations for one persona at a time, compared to "global" methods that construct the entire network at once. We also find that the generated networks match real networks on many characteristics, including density, clustering, connectivity, and degree distribution. However, we find that LLMs emphasize political homophily over all other types of homophily and significantly overestimate political homophily compared to real social networks.

cs.CY↗

arXiv2024

Inferring Dynamic Networks from Marginals with Iterative Proportional Fitting

A common network inference problem, arising from real-world data constraints, is how to infer a dynamic network from its time-aggregated adjacency matrix and time-varying marginals (i.e., row and column sums). Prior approaches to this problem have repurposed the classic iterative proportional fitting (IPF) procedure, also known as Sinkhorn's algorithm, with promising empirical results. However, the statistical foundation for using IPF has not been well understood: under what settings does IPF provide principled estimation of a dynamic network from its marginals, and how well does it estimate the network? In this work, we establish such a setting, by identifying a generative network model whose maximum likelihood estimates are recovered by IPF. Our model both reveals implicit assumptions on the use of IPF in such settings and enables new analyses, such as structure-dependent error bounds on IPF's parameter estimates. When IPF fails to converge on sparse network data, we introduce a principled algorithm that guarantees IPF converges under minimal changes to the network structure. Finally, we conduct experiments with synthetic and real-world data, which demonstrate the practical value of our theoretical and algorithmic contributions.

stat.ML↗

arXiv2024

Learning production functions for supply chains with graph neural networks

The global economy relies on the flow of goods over supply chain networks, with nodes as firms and edges as transactions between firms. While we may observe these external transactions, they are governed by unseen production functions, which determine how firms internally transform the input products they receive into output products that they sell. In this setting, it can be extremely valuable to infer these production functions, to improve supply chain visibility and to forecast future transactions more accurately. However, existing graph neural networks (GNNs) cannot capture these hidden relationships between nodes' inputs and outputs. Here, we introduce a new class of models for this setting by combining temporal GNNs with a novel inventory module, which learns production functions via attention weights and a special loss function. We evaluate our models extensively on real supply chains data and data generated from our new open-source simulator, SupplySim. Our models successfully infer production functions, outperforming the strongest baseline by 6%-50% (across datasets), and forecast future transactions, outperforming the strongest baseline by 11%-62%

cs.LG↗

arXiv2023

Machine Learning for Health symposium 2023 -- Findings track

A collection of the accepted Findings papers that were presented at the 3rd Machine Learning for Health symposium (ML4H 2023), which was held on December 10, 2023, in New Orleans, Louisiana, USA. ML4H 2023 invited high-quality submissions on relevant problems in a variety of health-related disciplines including healthcare, biomedicine, and public health. Two submission tracks were offered: the archival Proceedings track, and the non-archival Findings track. Proceedings were targeted at mature work with strong technical sophistication and a high impact to health. The Findings track looked for new ideas that could spark insightful discussion, serve as valuable resources for the community, or could enable new collaborations. Submissions to the Proceedings track, if not accepted, were automatically considered for the Findings track. All the manuscripts submitted to ML4H Symposium underwent a double-blind peer-review process.

cs.LG↗

arXiv2023

Accurate Measures of Vaccination and Concerns of Vaccine Holdouts from Web Search Logs

To design effective vaccine policies, policymakers need detailed data about who has been vaccinated, who is holding out, and why. However, existing data in the US are insufficient: reported vaccination rates are often delayed or missing, and surveys of vaccine hesitancy are limited by high-level questions and self-report biases. Here, we show how large-scale search engine logs and machine learning can be leveraged to fill these gaps and provide novel insights about vaccine intentions and behaviors. First, we develop a vaccine intent classifier that can accurately detect when a user is seeking the COVID-19 vaccine on search. Our classifier demonstrates strong agreement with CDC vaccination rates, with correlations above 0.86, and estimates vaccine intent rates to the level of ZIP codes in real time, allowing us to pinpoint more granular trends in vaccine seeking across regions, demographics, and time. To investigate vaccine hesitancy, we use our classifier to identify two groups, vaccine early adopters and vaccine holdouts. We find that holdouts, compared to early adopters matched on covariates, are 69% more likely to click on untrusted news sites. Furthermore, we organize 25,000 vaccine-related URLs into a hierarchical ontology of vaccine concerns, and we find that holdouts are far more concerned about vaccine requirements, vaccine development and approval, and vaccine myths, and even within holdouts, concerns vary significantly across demographic groups. Finally, we explore the temporal dynamics of vaccine concerns and vaccine seeking, and find that key indicators emerge when individuals convert from holding out to preparing to accept the vaccine.

cs.CY↗

arXiv2022

Estimating Geographic Spillover Effects of COVID-19 Policies From Large-Scale Mobility Networks

Many policies in the US are determined locally, e.g., at the county-level. Local policy regimes provide flexibility between regions, but may become less effective in the presence of geographic spillovers, where populations circumvent local restrictions by traveling to less restricted regions nearby. Due to the endogenous nature of policymaking, there have been few opportunities to reliably estimate causal spillover effects or evaluate their impact on local policies. In this work, we identify a novel setting and develop a suitable methodology that allow us to make unconfounded estimates of spillover effects of local policies. Focusing on California's Blueprint for a Safer Economy, we leverage how county-level mobility restrictions were deterministically set by public COVID-19 severity statistics, enabling a regression discontinuity design framework to estimate spillovers between counties. We estimate these effects using a mobility network with billions of timestamped edges and find significant spillover movement, with larger effects in retail, eating places, and gyms. Contrasting local and global policy regimes, our spillover estimates suggest that county-level restrictions are only 54% as effective as statewide restrictions at reducing mobility. However, an intermediate strategy of macro-county restrictions -- where we optimize county partitions by solving a minimum k-cut problem on a graph weighted by our spillover estimates -- can recover over 90% of statewide mobility reductions, while maintaining substantial flexibility between counties.

cs.CY↗

arXiv2021

To Recommend or Not? A Model-Based Comparison of Item-Matching Processes

Recommender systems are central to modern online platforms, but a popular concern is that they may be pulling society in dangerous directions (e.g., towards filter bubbles). However, a challenge with measuring the effects of recommender systems is how to compare user outcomes under these systems to outcomes under a credible counterfactual world without such systems. We take a model-based approach to this challenge, introducing a dichotomy of process models that we can compare: (1) a "recommender" model describing a generic item-matching process under a personalized recommender system and (2) an "organic" model describing a baseline counterfactual where users search for items without the mediation of any system. Our key finding is that the recommender and organic models result in dramatically different outcomes at both the individual and societal level, as supported by theorems and simulation experiments with real data. The two process models also induce different trade-offs during inference, where standard performance-improving techniques such as regularization/shrinkage have divergent effects. Shrinkage improves the mean squared error of matches in both settings, as expected, but at the cost of less diverse (less radical) items chosen in the recommender model but more diverse (more radical) items chosen in the organic model. These findings provide a formal language for how recommender systems may be fundamentally altering how we search for and interact with content, in a world increasingly mediated by such systems.

cs.CY↗

arXiv2019

Automatically Inferring Gender Associations from Language

In this paper, we pose the question: do people talk about women and men in different ways? We introduce two datasets and a novel integration of approaches for automatically inferring gender associations from language, discovering coherent word clusters, and labeling the clusters for the semantic concepts they represent. The datasets allow us to compare how people write about women and men in two different settings - one set draws from celebrity news and the other from student reviews of computer science professors. We demonstrate that there are large-scale differences in the ways that people talk about women and men and that these differences vary across domains. Human evaluations show that our methods significantly outperform strong baselines.

cs.CL↗

arXiv2018

Detecting Gang-Involved Escalation on Social Media Using Context

Gang-involved youth in cities such as Chicago have increasingly turned to social media to post about their experiences and intents online. In some situations, when they experience the loss of a loved one, their online expression of emotion may evolve into aggression towards rival gangs and ultimately into real-world violence. In this paper, we present a novel system for detecting Aggression and Loss in social media. Our system features the use of domain-specific resources automatically derived from a large unlabeled corpus, and contextual representations of the emotional and semantic content of the user's recent tweets as well as their interactions with other users. Incorporating context in our Convolutional Neural Network (CNN) leads to a significant improvement.

cs.CL↗