Christopher Barrie
Assistant Professor of Sociology, New York University
Core Faculty, Center for Social Media, AI, and Politics
About
I am a computational sociologist at New York University. My substantive work is on protest, conflict, communication, and political attitudes.
Methodologically, I work on natural language processing, large language models, and repurposing digital trace data for social science research.
Published in American Journal of Sociology · PNAS · Nature Human Behaviour · Political Analysis · British Journal of Political Science · Perspectives on Politics · Sociological Science
Publications
- 2026
- 2026
- 2026
- 2026
- 2026
- 2026
- 2025
- 2025
- 2025
- 2024
- 2024
- 2024
- 2024
- 2023
- 2023
- 2022
- 2021
- 2021
- 2021
- 2021
- 2021
- 2020
- 2020
- 2019
- 2018
- 2017
No publications match this filter.
Working papers
·
- 2026
- 2026
- 2026
- 2026
- 2026
- 2026
- 2026
- 2026
- 2025
- 2024
- 2024
No working papers in this area yet.
Software
You kept scrolling
This is Thomas Schelling’s model of segregation. Each square is content when enough of its neighbours share its colour; outlined squares are not, and move. Set how much alike company each square wants, press play, and watch mild preferences add up to a sorted board. Pause at any point to move squares yourself.
- Round
- 000
- Content
- 0%
- Alike
- 0%
- Random mix
- 0%
How the model works
The rule. A square’s neighbourhood is the 8 cells around it, or the 24 within two steps in Wide view. Its alike share is s = alike neighbours ÷ occupied neighbours. It is content when s is at least the threshold t you set; in Prefers mixing it must also be at most 75%. A square with no neighbours is content.
A round. Moving takes time, so in each round a discontented square moves with probability 30%. A square that moves looks at 6 empty cells chosen at random and takes the nearest one where it would be content; if none of them suits, it stays where it is. Separately, 1% of all squares move to a random empty cell each round for reasons of their own, as people do, so the board never quite freezes and some squares are always unsettled. The run stops once alike neighbours change by less than one point over 20 rounds, or after 150 rounds.
The measures. Alike neighbours is the average of s over all squares that have neighbours. If mixed at random is the value expected if the same squares were shuffled: Σg ng(ng − 1) / N(N − 1). That is 50% for two equal groups, about 17% for six, and about 62% for a 75:25 split. Content is the share of squares that are content right now. Changing the threshold replays the same starting board, so runs can be compared.
What to look for. Schelling’s point is the gap between the dashed line and the orange one. With a threshold of 35%, no square minds being in the minority, yet the board typically levels off well above the random mix. Because a few squares keep moving for other reasons, content never quite reaches 100%, and alike patches keep slowly growing even once nearly everyone is content. Set the threshold to 75% and the board stays frustrated: many squares can never be satisfied. With a 75:25 split, the majority’s alike share barely moves from its 75% starting point, while the minority’s rises far above its 25% share of the population; some minority squares may never find a place that suits them. In Prefers mixing, squares that dislike being a large majority still end up well above the random mix and often never settle, in line with Zhang (2004) and Pancs and Vriend (2007).
Schelling, T. C. (1971). Dynamic models of segregation. Journal of Mathematical Sociology, 1(2), 143–186. · Schelling, T. C. (1978). Micromotives and Macrobehavior. Norton. · Zhang, J. (2004). Residential segregation in an all-integrationist world. Journal of Economic Behavior & Organization, 54(4), 533–550. · Pancs, R., & Vriend, N. J. (2007). Schelling’s spatial proximity model of segregation revisited. Journal of Public Economics, 91(1–2), 1–24.