Tingyu Dai • Sociology Researcher

About

Tracing how AI and digital metrics reorganize inequality.

My research examines how socio-technical systems, particularly platforms and AI tools, reshape information environments and opportunity structures, and thereby reproduce or transform social inequality. Broadening information exposure has long been thought to narrow inequality. But in the digital era, exposure can no longer be treated as a binary variable: information mediated through platforms and AI arrives instantaneously and in abundance, yet it is also stochastic and fragmentary. What looks like wider access often turns out to be a new gradient of advantage.

In my current projects, I take organizations as the mediating mechanism through which these dynamics travel. I study how people respond to digital metrics that are constantly updated and increasingly central to evaluation and competition, and how these responses accumulate into broader organizational changes and reshape the structure of the entire field. My master’s thesis approaches these questions through the lens of immigrant labor markets, while another ongoing project investigates how citation-metric visibility may influence patterns of academic collaboration.

Alongside this work, I am actively seeking research opportunities in quantitative and computational social science. Practice with large-scale data, I believe, will sharpen my ability to operationalize structural inequality, and allow me to carry these concerns into subfields like labor markets, occupations, and education.

I studied philosophy at Peking University, spent a semester on exchange in the EAA Program at the University of Tokyo, and completed my M.A. in Sociology at Columbia University, advised by Prof. David Stark. I now serve as a Research Assistant in the COMAP Lab at Northwestern with Prof. Yingdan Lu, and am preparing Ph.D. applications in sociology for the Fall 2027 cycle.

Research Topics

Three angles on a single question: who is sorted by digital infrastructure, where they land, and how we make those processes legible.

Inequality

I approach inequality as a structural and categorical phenomenon rather than a distribution of individual outcomes. The question that animates my work is whether digital systems reproduce older hierarchies of race, migration, and class, or generate new ones whose contours we still need to map.

An early project, working with Opportunity Insights data, found racial segregation to be a substantially stronger drag on upward mobility than income segregation, a finding that drew me toward how durable categorical structures continue to organize opportunity even as the infrastructures that distribute it shift.

Labor Markets and Occupation

Labor markets are where the digital reordering of information becomes materially consequential. Platform metrics and AI tools do not merely change what workers do; they reorganize who can enter an occupation, how mobility happens inside it, and how whole industries restructure around new evaluative regimes.

My thesis follows immigrant labor into the U.S. vertical-drama industry, where platforms drove rapid occupational entry, only for subsequent industrial upgrading to threaten to render that same labor disposable. Building on this, I hope to extend my work into AI exposure, occupational mobility, and the intergenerational reproduction of advantage.

Computational Social Science

Studying these processes at scale requires methods equal to their complexity. At the COMAP Lab, I contribute to LLM-assisted measurement, multimodal data pipelines, and causal-identification strategies for observational research on social media.

An exploratory project of my own treats the 2011 publicization of Google Scholar profiles as a causal shock, asking how citation visibility reshapes patterns of academic collaboration.

Projects

Below are working papers and data projects at different stages of development. I welcome comments and conversations about possible collaborations or extensions.

01
Working paper · Master’s thesis

Sponsored by Metrics: How Platforms Organize Immigrants into an Industry
- qualitative
- computational triangulation
- labor markets and occupation
Theorizes how platform metrics, by becoming simultaneously legible to immigration governance and commercial evaluation, can assemble an immigrant labor force at unprecedented speed and dispose of it at the same pace.

Within three years, the U.S. vertical drama industry, a mobile-first cultural sector built by Chinese entrepreneurs and staffed primarily by Chinese international film students and recent graduates, recruited thousands of workers, differentiated a full production value chain, achieved internal consensus around metric-oriented production norms, and then began displacing the very workers it had incorporated. This paper argues that the mechanism behind this trajectory is what I call dual-legibility: the capacity of platform metrics to be simultaneously recognized as credible evidence by two distinct institutional systems, in this case U.S. immigration governance (specifically the O-1 extraordinary-ability visa) and commercial market evaluation. Drawing on ten in-depth interviews, fieldwork in Los Angeles and online co-ethnic communities, and an original cross-platform dataset covering ReelShort’s first fifty U.S. productions, the analysis traces organizational genesis across four stages (aggregation, configuration, alignment, self-reinforcement), and shows how the same property that lowered entry barriers for immigrant workers structurally encoded their disposability.

Working sample (PDF) →

Status Drafting. A second phase of fieldwork from May to August 2026 will extend the interview pool and support mechanism-oriented qualitative coding, with particular attention to the displacement dynamics identified in the self-reinforcement stage.
02
In progress

Counting in Public: Citation Visibility and the Reorganization of Scholarly Collaboration
- causal inference
- social network analysis
- science of science
Treats the 2011 publicization of Google Scholar citation profiles as a causal shock to test how visible metrics reorganize the collaboration networks of science.

Since 2011, Google Scholar Citations has rendered individual scholars’ citation profiles publicly viewable and continuously updated, transforming what was previously a relatively opaque evaluative environment into one in which scholarly performance is legible in real time to colleagues, hiring committees, and competitors. This project treats the rollout of public citation profiles as a causal shock and asks how scholars adjusted their collaboration behavior in response to a newly transparent regime of metric competition. Using bibliometric data on computer science publications, I plan a difference-in-differences design to examine several margins of adjustment: changes in co-authorship team size, the propensity to recruit senior or already highly-cited co-authors, the persistence of prior collaboration ties, and the rate of cross-disciplinary collaboration. The aim is to identify how visible quantification reorganizes the social networks through which scientific knowledge is produced.

Status Feasibility stage. I am currently using computer-science publication data to test parallel pre-trends and refine the identification strategy.
03
Data project

Indexing Disadvantage: A Tract-Level Measure of Neighborhood Deprivation and Its Health Associations
- index construction
- factor analysis
- machine learning
- neighborhood inequality
Builds a tract-level index of neighborhood disadvantage from ACS data and tests its associations with obesity and mental health prevalence across 55,534 U.S. census tracts.

This project constructs a tract-level Neighborhood Disadvantage Index (NDI) using a hybrid theory-and-data-driven strategy: established frameworks (the ADI, SDI, reADI, and Sampson and Raudenbush’s concentrated-disadvantage construct) define the relevant theoretical dimensions, and maximum-likelihood factor analysis then extracts the latent structure within that domain. The resulting index integrates 13 ACS indicators across six dimensions (economic deprivation, employment, education, family structure, housing, and material resources) and incorporates two indicators absent from prior frameworks: rent burden and household broadband access. Across 55,534 U.S. census tracts, the NDI explains 43.7% of the variance in obesity prevalence and 55.8% of the variance in poor-mental-health prevalence reported in CDC PLACES, with results robust to alternative transformation choices and to indicator-level LASSO benchmarks. The exercise illustrates both the analytic gains from disciplined measurement of structural disadvantage and the limits of cross-sectional, OLS-based inference when spatial autocorrelation is left unaddressed.

Report → Code (.R) →

Status Completed data assessment.

Teaching & Service

Peer Reviewer

Reviewed six submissions (five papers and one symposium proposal) across the Organizations and Management Theory (OMT) and Communication, Digital Technology, and Organization (CTO) divisions for the 2026 Annual Meeting of the Academy of Management.

Academy of Management · 2026

Volunteer ESL Aide

Served as a weekly classroom aide at Flushing International High School, supporting English-language instruction for immigrant students and mentoring seniors on their college application essays.

Flushing International High School · Fall 2024

Contact

Email td2853@columbia.edu
Office Frances Searle Building
2240 Campus Drive
Evanston, IL 60208

Get in touch

Open to conversations about research, collaboration, and graduate study — particularly on inequality, labor markets, computational social science, and how AI and digital metrics reshape opportunity structures. Email is the surest way to reach me.