Latent patterns of urban mixing in mobility analysis across five global cities

B. P. Y. Loo; C. Ratti; E. Moro; F. Duarte; Z. Fan

arxiv: 2604.12202 · v1 · submitted 2026-04-14 · 💻 cs.AI · cs.SI

Latent patterns of urban mixing in mobility analysis across five global cities

Z. Fan , B. P. Y. Loo , F. Duarte , C. Ratti , E. Moro This is my paper

Pith reviewed 2026-05-10 15:56 UTC · model grok-4.3

classification 💻 cs.AI cs.SI

keywords urban mobilitysocial mixingactivity spaceplace exposuretravel surveysgraph neural networksautoencoderssocioeconomic mixing

0 comments

The pith

The structure of where people travel shapes experienced social mixing more than home location, demographics, or transit access across five cities.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper uses travel surveys from over 200,000 people in Boston, Chicago, Hong Kong, London, and Sao Paulo to compare how different factors affect social mixing in daily places. It shows that estimating status from neighborhoods alone produces mixing levels 16 percent lower than self-reported data, that older adults mix more than those nearing retirement while teens and caregivers mix less, and that living near major transit reduces the role of income in mixing. A model that combines home locations, travel destinations, and demographics finds that the pattern of destinations accounts for most differences in place exposure. This suggests movement choices create distinct mixing experiences even when overall levels look similar across income groups.

Core claim

Using the same individual-level data across five cities, the authors show that the configuration of an individual's activity space accounts for the largest share of variation in exposure to places of different socioeconomic character, while residential neighborhood, demographics, and transit proximity contribute less. Ablation tests confirm that income groups can reach comparable overall mixing levels yet maintain structurally separate activity spaces that produce different sequences of encounters.

What carries the argument

A supervised autoencoder that embeds home-space attributes, activity-space structure from a graph neural network of spatio-temporal place networks, and demographic inputs to predict individual exposure vectors, with ablation isolating the contribution of activity-space structure.

If this is right

Income-stratified activity spaces produce qualitatively different sequences of social encounters even when total mixing levels are similar.
Proximity to major transit stations weakens the direct link between individual socioeconomic status and mixing outcomes.
Adults over age 66 show higher mixing than those aged 55-65, consistent with increased travel after retirement.
Teenagers and women with caregiving duties display measurably lower mixing levels than other groups.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Policies that expand affordable travel options could increase mixing more directly than neighborhood redesign alone.
Persistent separation of activity spaces by income may maintain distinct social experiences even in cities that appear well-mixed on aggregate metrics.
The five-city comparison implies that similar activity-space dominance would appear in additional cities once comparable survey data become available.

Load-bearing premise

Self-reported survey responses on income, age, gender, and travel destinations accurately capture real behaviors and that the autoencoder can isolate activity-space effects without unmeasured confounding variables.

What would settle it

A follow-up study that replaces self-reported destinations with verified GPS traces for the same individuals and finds that the activity-space structure no longer predicts exposure vectors better than home neighborhood or demographics alone.

read the original abstract

This study leverages large-scale travel surveys for over 200,000 residents across Boston, Chicago, Hong Kong, London, and Sao Paulo. With rich individual-level data, we make systematic comparisons and reveal patterns in social mixing, which cannot be identified by analyzing high-resolution mobility data alone. Using the same set of data, inferring socioeconomic status from residential neighborhoods yield social mixing levels 16% lower than using self-reported survey data. Besides, individuals over the age of 66 experience greater social mixing than those in late working life (aged 55 to 65), lending data-driven support to the "second youth" hypothesis. Teenagers and women with caregiving responsibilities exhibit lower social mixing levels. Across the five cities, proximity to major transit stations reduces the influence of individual socioeconomic status on social mixing. Finally, we construct detailed spatio-temporal place networks for each city using a graph neural network. Inputs of home-space, activity-space and demographic attributes are embedded and fed into a supervised autoencoder to predict individual exposure vectors. Results show that the structure of individual activity space, i.e., where people travel to, explains most of the variations in place exposure, suggesting that mobility shapes experienced social mixing more than sociodemographic characteristics, home environment, and transit proximity. The ablation tests further discover that, while different income groups may experience similar levels of social mixing, their activity spaces remain stratified by income, resulting in structurally different social mixing experiences.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Activity space explains most variation in experienced mixing across five cities, but the GNN ablation likely overstates its role because exposure targets and activity features come from the same trip data.

read the letter

The main point is that this work pulls large travel surveys from Boston, Chicago, Hong Kong, London, and Sao Paulo to show activity space structure predicts individual exposure better than home location or demographics, with a 16% gap when SES is inferred from neighborhoods instead of self-reports. It also finds older adults over 66 mix more than those 55-65, teens and caregiving women mix less, and transit proximity weakens the SES-mixing link. The GNN ablation is presented as evidence that activity space dominates, plus that income groups have stratified activity patterns even at similar mixing levels. The scale and cross-city comparisons are the solid parts here, giving concrete numbers that single-city studies often lack and tying directly to planning questions about infrastructure versus residential design. The age and gender patterns add useful detail too. The soft spots are real though. The abstract gives no information on model validation, error bars, sample weights, or preprocessing, so the specific percentages and 'explains most' claim stay hard to assess. The stress-test concern lands: exposure vectors are computed from the identical mobility traces used for activity-space inputs, so dropping the activity embedding removes the direct computational path to the target and guarantees a bigger drop than removing demographics or home inputs. Nothing in the description shows they used orthogonalization, instruments, or fixed-activity counterfactuals to separate that dependence from genuine contribution. This is for urban mobility and segregation researchers who want multi-city empirical patterns rather than new theory. A reader working on policy for social integration through transit would get the most out of the transit and activity-space findings. It deserves peer review because the data volume and comparative angle are substantive enough to warrant referee time, even with the modeling questions.

Referee Report

1 major / 1 minor

Summary. This paper analyzes social mixing patterns across Boston, Chicago, Hong Kong, London, and Sao Paulo using travel surveys from over 200,000 residents. It compares mixing levels from self-reported versus residential-inferred socioeconomic status (finding 16% lower levels with the latter), identifies age and gender differences (higher mixing for those over 66, lower for teens and caregiving women), shows transit proximity reduces SES influence on mixing, and employs graph neural networks for spatio-temporal place networks plus a supervised autoencoder with home, activity-space, and demographic embeddings to predict exposure vectors. The central result is that activity-space structure explains most variation in place exposure, with ablation tests indicating income-stratified activity spaces despite similar mixing levels.

Significance. If the central claims hold after addressing data-dependence issues, the work provides a valuable multi-city, individual-level empirical contribution to understanding how mobility shapes experienced social mixing beyond sociodemographics or home location. The scale of the dataset and the use of supervised autoencoders for exposure prediction offer a replicable framework for urban mobility studies, with potential policy relevance for transit planning and social integration.

major comments (1)

[Supervised autoencoder and ablation tests] In the section on the supervised autoencoder and ablation tests: the claim that activity-space inputs explain most variation in predicted individual exposure vectors is undermined by the construction of the data. Both the activity-space features (locations visited and their structure) and the place exposure vectors are derived from the identical travel-survey trip records. Removing the activity-space embedding therefore removes the direct computational pathway to the target variable, guaranteeing a larger performance drop irrespective of whether sociodemographics are the ultimate driver. The model lacks built-in mechanisms (e.g., orthogonalization, instrumental variables, or fixed-activity counterfactuals) to separate direct dependence from independent contribution, which is load-bearing for the headline result that mobility shapes experienced social mixing more than other factors.

minor comments (1)

[Abstract] The abstract reports quantitative claims (e.g., '16% lower') and model conclusions without reference to statistical significance, error bars, or validation metrics; adding these would improve verifiability.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive and detailed comments, which help clarify the interpretation of our supervised autoencoder results. We address the major comment on the ablation tests and data dependence below, providing an honest assessment of the analysis while outlining revisions to strengthen the manuscript.

read point-by-point responses

Referee: In the section on the supervised autoencoder and ablation tests: the claim that activity-space inputs explain most variation in predicted individual exposure vectors is undermined by the construction of the data. Both the activity-space features (locations visited and their structure) and the place exposure vectors are derived from the identical travel-survey trip records. Removing the activity-space embedding therefore removes the direct computational pathway to the target variable, guaranteeing a larger performance drop irrespective of whether sociodemographics are the ultimate driver. The model lacks built-in mechanisms (e.g., orthogonalization, instrumental variables, or fixed-activity counterfactuals) to separate direct dependence from independent contribution, which is load-bearing for the headline result that mobility shapes experienced social mixing more than other factors.

Authors: We agree that the activity-space features and exposure vectors share a common origin in the trip records, creating an inherent dependence that the ablation necessarily exploits. This dependence, however, aligns with the substantive claim: experienced social mixing is realized through the specific places visited, so the structure of activity space is the direct mechanism. The ablation quantifies relative predictive utility: home location and demographics alone yield substantially weaker reconstruction of exposure vectors, while including activity-space embeddings recovers most of the variance. This pattern indicates that mobility patterns contain information about mixing that is not fully reducible to sociodemographic traits or residential context. We acknowledge that the architecture does not incorporate orthogonalization, instrumental variables, or counterfactual fixed-activity designs, which would be required for stricter causal separation; such extensions lie beyond the scope of the current travel-survey data. In the revised manuscript we will (i) explicitly state the shared data provenance, (ii) reframe the ablation results as evidence of predictive dominance rather than isolated causal effect, and (iii) add a limitations paragraph discussing the absence of disentangling mechanisms and the consequent interpretive bounds. revision: partial

Circularity Check

0 steps flagged

No significant circularity in the derivation or analysis chain.

full rationale

The paper conducts empirical data analysis on travel survey inputs across five cities, using embeddings fed into a supervised autoencoder to predict exposure vectors and ablation tests to assess input importance. No mathematical derivation, equation, or self-referential definition is presented in which a claimed result (such as activity space explaining variations in exposure) reduces by construction to its own fitted inputs or definitions. The central claim arises from model performance differences on real survey data rather than a tautological reduction or renamed fit. No load-bearing self-citations, uniqueness theorems, or smuggled ansatzes are invoked. This is a standard self-contained empirical ML study on observational data, consistent with the default expectation of low circularity scores for such work.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The central claims rest on the validity and representativeness of self-reported survey data across culturally diverse cities and on the ability of the GNN-autoencoder pipeline to attribute variance to activity space without residual confounding.

free parameters (1)

GNN and autoencoder hyperparameters
Parameters of the graph neural network and supervised autoencoder are tuned to the survey data to predict individual exposure vectors.

axioms (2)

domain assumption Self-reported survey responses provide a more accurate measure of socioeconomic status than neighborhood-based inference
The 16% difference is presented as evidence favoring self-reported data.
domain assumption Travel survey data from the five cities are comparable for systematic analysis despite differences in culture and infrastructure
The study performs cross-city comparisons using the same data processing pipeline.

pith-pipeline@v0.9.0 · 5569 in / 1560 out tokens · 74449 ms · 2026-05-10T15:56:51.988677+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages

[1]

Ratti, C., Frenchman, D., Pulselli, R. M. & Williams, S. Mobile Landscapes: Using L ocation Data from Cell Phones for Urban Analysis. Environ. Plan. B Plan. Des. 33, 727–748 (2006)

work page 2006
[2]

& Pentland, A

Eagle, N. & Pentland, A. S. Eigenbehaviors: identifying structure in routine. Behav. Ecol. Sociobiol. 63, 1057–1066 (2009)

work page 2009
[3]

& Pentland, A

Moro, E., Calacci, D., Dong, X. & Pentland, A. Mobility patterns are associated with experienced income segregation in large US cities. Nat. Commun. 12, 4633 (2021)

work page 2021
[4]

Wang, A. et al. Personal Mobility Choices and Disparities in Carbon Emissions. Environ. Sci. Technol. 57, 8548–8558 (2023)

work page 2023
[5]

& Kong, H

Zhang, Y., Fu, W., Chao, H., Mi, Z. & Kong, H. A comparative analysis of the potential of carbon emission reductions from shared micro-mobility. Sustain. Energy Technol. Assess. 72, 104088 (2024). Fan, Z., Loo, B.P.Y., Duarte, F., Ratti, C., & Moro, E. (2026). Latent patterns of urban mixing in mobility analysis across five global cities. Nature Cities, a...

work page 2024
[6]

M., Kloog, I., Britter, R., Ratti, C

Nyhan, M. M., Kloog, I., Britter, R., Ratti, C. & Koutrakis, P. Quantifying population exposure to air pollution using individual mobility patterns inferred from mobile phone data. J. Expo. Sci. Environ. Epidemiol. 29, 238–247 (2019)

work page 2019
[7]

Loo, B. P. Y. & Axhausen, K. Getting out of energy-intensive and "dirty" transport. Innovation 3, 100339 (2022)

work page 2022
[8]

& Lyons, T

Ewing, R., Tian, G. & Lyons, T. Does compact development increase or reduce traffic congestion? Cities 72, 94–101 (2018)

work page 2018
[9]

& Pouladi, R

Sardari, R., Hamidi, S. & Pouladi, R. Effects of Traffic Congestio n on Vehicle Miles Traveled. Transp. Res. Rec. 2672, 92–102 (2018)

work page 2018
[10]

& Chai, Y

Ma, J., Rao, J., Kwan, M.-P. & Chai, Y. Examining the effects of mobility-based air and noise pollution on activity satisfaction. Transp. Res. Part D Transp. Environ. 89, 102633 (2020)

work page 2020
[11]

& Franklin, M

Yin, X., Fallah -Shorshani, M., McConnell, R., Fruin, S. & Franklin, M. Predicting Fine Spatial Scale Traffic Noise Using Mobile Measurements and Machine Learning. Environ. Sci. Technol. 54, 12860– 12869 (2020)

work page 2020
[12]

Lian, T., Loo, B. P. Y. & Fan, Z. Adva nces in estimating pedestrian measures through artificial intelligence: From data sources, computer vision, video analytics to the prediction of crash frequency. Comput. Environ. Urban Syst. 107, 102057 (2024)

work page 2024
[13]

Loo, B. P. Y., Tsoi, K. H., Wong, P. P. & Lai, P. C. Identification of superspreading environment under COVID-19 through human mobility data. Sci. Rep. 11, 4699 (2021)

work page 2021
[14]

The flight of the creative class: The new global competition for talent

Florida, R. The flight of the creative class: The new global competition for talent. Lib. Educ. 92, 22– 29 (2006)

work page 2006
[15]

Major, B., Mendes, W. B. & Dovidio, J. F. Intergroup relations and health disparities: a social psychological perspective. Health Psychol. 32, 514–524 (2013)

work page 2013
[16]

Chetty, R. et al. Social capital I: measurement and associations with economic mobility. Nature 608, 108–121 (2022)

work page 2022
[17]

Cabrera, J. F. & Najarian, J. C. How the Built Environment Shapes Spatial Bridging Ties and Social Capital. Environ. Behav. 47, 239–267 (2015)

work page 2015
[18]

SDG Indicators

United Nations. SDG Indicators. https://unstats.un.org/sdgs/indicators/indicators-list/ (2023)

work page 2023
[19]

Nilforoshan, H. et al. Human mobility networks reveal increased segregation in large cities. Nature 624, 586–592 (2023). Fan, Z., Loo, B.P.Y., Duarte, F., Ratti, C., & Moro, E. (2026). Latent patterns of urban mixing in mobility analysis across five global cities. Nature Cities, accepted. Pre-print uploaded on 10/04/2026. Supplementary Materials are not u...

work page 2023
[20]

Loo, B. P. Y., Fan, Z. & Moro, E. Residential and experienced social segregation: the roles of different transport mo des, metro extensions, and longitudinal changes in Hong Kong. Humanit. Soc. Sci. Commun. 11, 1439 (2024)

work page 2024
[21]

& Schmidt, T

Athey, S., Ferguson, B., Gentzkow, M. & Schmidt, T. Estimating experienced racial segregation in US cities using large-scale GPS data. Proc. Natl. Acad. Sci. U. S. A. 118, e2026160118 (2021)

work page 2021
[22]

& Wang, Q

Han, Y., Liao, P., Li, W. & Wang, Q. R. Visitation patterns reveal service access disparities for ageing populations in the USA. Nat. Hum. Behav. https://doi.org/10.1038/s41562-025-02285-4 (2025)

work page doi:10.1038/s41562-025-02285-4 2025
[23]

& Glaeser, E

Cook, C., Currier, L. & Glaeser, E. Urban mobility and the experienced isolation of students. Nat. Cities 1, 73–82 (2024)

work page 2024
[24]

Jung, H., Lee, J. S. & Kim, S. Beyond Population Density: A New Framework for Compact Cities through Amenity Survival Analysis. Cities 168, 105535 (2026)

work page 2026
[25]

& Burgess, R

Jenks, M. & Burgess, R. Compact Cities: Sustainable Urban Forms for Developing Countries (Routledge, 2000)

work page 2000
[26]

Chen, C., Bian, L. & Ma, J. From Traces to Trajectories: How Well Can We Guess Activity Locations from Mobile Phone Traces? Transp. Res. Part C Emerg. Technol. 46, 326–337 (2014)

work page 2014
[27]

Innovations in Travel Survey Methods

National Research Council Transportation Research Board. Innovations in Travel Survey Methods. Transportation Research Record no. 1412 (National Academy Press, 1993)

work page 1993
[28]

& Madre, J.-L

Armoogum, J. & Madre, J.-L. Weighting or imputations? The example of nonresponses for daily trips in French NPTS. J. Transp. Stat. 1, 53–63 (1998)

work page 1998
[29]

& Stopher, P

Jones, P. & Stopher, P. R. (eds) Transport Survey Quality and Innovation (Pergamon, 2003)

work page 2003
[30]

Moore, J. C. & Durrant, G. Non-response bias risks in sample sub-groups in Understanding Society: the UK Household Longitudinal Study. Understanding Society Working Paper Series 2025 -03 https://www.understandingsociety.ac.uk/wp-content/uploads/working-papers/2025-03.pdf (2025)

work page 2025
[31]

Discrete Global Grid Systems as scalable geospatial frameworks for characterizing coastal environments

Bousquin, J. Discrete Global Grid Systems as scalable geospatial frameworks for characterizing coastal environments. Environ. Model. Softw. 146, 105210 (2021)

work page 2021
[32]

Salazar-Miranda, A. et al. Exploring the social life of urban spaces through AI. Proc. Natl. Acad. Sci. U. S. A. 122, e2424662122 (2025)

work page 2025
[33]

Loo, B. P. Y. & Fan, Z. Social interaction in public space: Spatial edges, moveable furniture, and visual landmarks. Environ. Plan. B Urban Anal. City Sci. 50, 2510–2526 (2023). Fan, Z., Loo, B.P.Y., Duarte, F., Ratti, C., & Moro, E. (2026). Latent patterns of urban mixing in mobility analysis across five global cities. Nature Cities, accepted. Pre-print ...

work page 2023
[34]

M., Batty, M

Roth, C., Kang, S. M., Batty, M. & Barthélemy, M. Structure of Urban Movements: Polycentric Activity and Entangled Hierarchical Flows. PLoS ONE 6, e15923 (2011)

work page 2011
[35]

Lindeman, R. H. Introduction to Bivariate and Multivariate Analysis (Scott Foresman, 1980)

work page 1980
[36]

Aston, L. et al . Addressing Transit Mode Location Bias in Built Environment -Transit Mode Use Research. J. Transp. Geogr. 87, 102819 (2020)

work page 2020
[37]

Weng, G., Kim, M., Ahn, Y. -Y. & Moro, E. Beyond Distance: Mobility Neural Embeddings Reveal Visible and Invisible Barriers in U rban Space . Preprint at https://doi.org/10.48550/arXiv.2506.24061 (2025)

work page doi:10.48550/arxiv.2506.24061 2025
[38]

& Guimerà , R

Cabanas -Tirapu, O., Danús, L., Moro, E., Sales -Pardo, M. & Guimerà , R. Human mobility is well described by closed-form gravity-like models learned automatically from data. Nat. Commun. 16, 1336 (2025)

work page 2025
[39]

& González, M

Jiang, S., Ferreira, J. & González, M. C. Clustering daily patterns of human activities in the city. Data Min. Knowl. Discov. 25, 478–510 (2012)

work page 2012
[40]

H., Parra, D

de Sá, T. H., Parra, D. C. & Monteiro, C. A. Impact of travel mode shift and trip distance on active and non-active transportation in the São Paulo Metropolitan Area in Brazil. Prev. Med. Rep. 2, 183–188 (2015)

work page 2015
[41]

A., Wilby, D

Fairnie, G. A., Wilby, D. J. & Saunders, L. E. Active travel in London: the role of travel survey data in describing population physical activity. J. Transp. Health 3, 161–172 (2016)

work page 2016
[42]

Forrest, T. L. & Pearson, D. F. Comparison of Trip Determination Methods i n Household Travel Surveys Enhanced by a Global Positioning System. Transp. Res. Rec. 1917, 63–71 (2005)

work page 1917
[43]

OSMnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks

Boeing, G. OSMnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks. Comput. Environ. Urban Syst. 65, 126–139 (2017)

work page 2017
[44]

R., Pentland, A

Yabe, T., García Bulle Bueno, B., Frank, M. R., Pentland, A. & Moro, E. Behaviour-based dependency networks between places shape urban economic resilience. Nat. Hum. Behav. 9, 496–506 (2025)

work page 2025
[45]

& Vetter, C

Luxen, D. & Vetter, C. Real -time routing with OpenStreetMap data. in Proceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems 513 –516 (ACM, 2011)

work page 2011
[46]

Goodchild, M. F. First law of geography. in International Encyclopedia of Human Geography (eds Kitchin, R. & Thrift, N.) 179–182 (Elsevier, 2009)

work page 2009
[47]

On the First Law of Geography: A Reply

Tobler, W. On the First Law of Geography: A Reply. Ann. Assoc. Am. Geogr. 94, 304–310 (2004). Fan, Z., Loo, B.P.Y., Duarte, F., Ratti, C., & Moro, E. (2026). Latent patterns of urban mixing in mobility analysis across five global cities. Nature Cities, accepted. Pre-print uploaded on 10/04/2026. Supplementary Materials are not uploaded. 26

work page 2004
[48]

C., Hagenbuchner, M

Scarselli, F., Gori, M., Tsoi, A. C., Hagenbuchner, M. & Monfardini, G. The graph neural network model. IEEE Trans. Neural Netw. 20, 61–80 (2008)

work page 2008
[49]

Zhai, J., Zhang, S., Chen, J. & He, Q. Autoencoder and its various variants. in 2018 IEEE International Conference on Systems, Man, and Cybernetics (SMC) 415–419 (IEEE, 2018). Correspondence Correspondence should be addressed to Professor Becky P.Y. Loo at bpyloo@hku.hk. The aggregated data for repeating the analysis are available via this repository (htt...

work page 2018

[1] [1]

Ratti, C., Frenchman, D., Pulselli, R. M. & Williams, S. Mobile Landscapes: Using L ocation Data from Cell Phones for Urban Analysis. Environ. Plan. B Plan. Des. 33, 727–748 (2006)

work page 2006

[2] [2]

& Pentland, A

Eagle, N. & Pentland, A. S. Eigenbehaviors: identifying structure in routine. Behav. Ecol. Sociobiol. 63, 1057–1066 (2009)

work page 2009

[3] [3]

& Pentland, A

Moro, E., Calacci, D., Dong, X. & Pentland, A. Mobility patterns are associated with experienced income segregation in large US cities. Nat. Commun. 12, 4633 (2021)

work page 2021

[4] [4]

Wang, A. et al. Personal Mobility Choices and Disparities in Carbon Emissions. Environ. Sci. Technol. 57, 8548–8558 (2023)

work page 2023

[5] [5]

& Kong, H

Zhang, Y., Fu, W., Chao, H., Mi, Z. & Kong, H. A comparative analysis of the potential of carbon emission reductions from shared micro-mobility. Sustain. Energy Technol. Assess. 72, 104088 (2024). Fan, Z., Loo, B.P.Y., Duarte, F., Ratti, C., & Moro, E. (2026). Latent patterns of urban mixing in mobility analysis across five global cities. Nature Cities, a...

work page 2024

[6] [6]

M., Kloog, I., Britter, R., Ratti, C

Nyhan, M. M., Kloog, I., Britter, R., Ratti, C. & Koutrakis, P. Quantifying population exposure to air pollution using individual mobility patterns inferred from mobile phone data. J. Expo. Sci. Environ. Epidemiol. 29, 238–247 (2019)

work page 2019

[7] [7]

Loo, B. P. Y. & Axhausen, K. Getting out of energy-intensive and "dirty" transport. Innovation 3, 100339 (2022)

work page 2022

[8] [8]

& Lyons, T

Ewing, R., Tian, G. & Lyons, T. Does compact development increase or reduce traffic congestion? Cities 72, 94–101 (2018)

work page 2018

[9] [9]

& Pouladi, R

Sardari, R., Hamidi, S. & Pouladi, R. Effects of Traffic Congestio n on Vehicle Miles Traveled. Transp. Res. Rec. 2672, 92–102 (2018)

work page 2018

[10] [10]

& Chai, Y

Ma, J., Rao, J., Kwan, M.-P. & Chai, Y. Examining the effects of mobility-based air and noise pollution on activity satisfaction. Transp. Res. Part D Transp. Environ. 89, 102633 (2020)

work page 2020

[11] [11]

& Franklin, M

Yin, X., Fallah -Shorshani, M., McConnell, R., Fruin, S. & Franklin, M. Predicting Fine Spatial Scale Traffic Noise Using Mobile Measurements and Machine Learning. Environ. Sci. Technol. 54, 12860– 12869 (2020)

work page 2020

[12] [12]

Lian, T., Loo, B. P. Y. & Fan, Z. Adva nces in estimating pedestrian measures through artificial intelligence: From data sources, computer vision, video analytics to the prediction of crash frequency. Comput. Environ. Urban Syst. 107, 102057 (2024)

work page 2024

[13] [13]

Loo, B. P. Y., Tsoi, K. H., Wong, P. P. & Lai, P. C. Identification of superspreading environment under COVID-19 through human mobility data. Sci. Rep. 11, 4699 (2021)

work page 2021

[14] [14]

The flight of the creative class: The new global competition for talent

Florida, R. The flight of the creative class: The new global competition for talent. Lib. Educ. 92, 22– 29 (2006)

work page 2006

[15] [15]

Major, B., Mendes, W. B. & Dovidio, J. F. Intergroup relations and health disparities: a social psychological perspective. Health Psychol. 32, 514–524 (2013)

work page 2013

[16] [16]

Chetty, R. et al. Social capital I: measurement and associations with economic mobility. Nature 608, 108–121 (2022)

work page 2022

[17] [17]

Cabrera, J. F. & Najarian, J. C. How the Built Environment Shapes Spatial Bridging Ties and Social Capital. Environ. Behav. 47, 239–267 (2015)

work page 2015

[18] [18]

SDG Indicators

United Nations. SDG Indicators. https://unstats.un.org/sdgs/indicators/indicators-list/ (2023)

work page 2023

[19] [19]

Nilforoshan, H. et al. Human mobility networks reveal increased segregation in large cities. Nature 624, 586–592 (2023). Fan, Z., Loo, B.P.Y., Duarte, F., Ratti, C., & Moro, E. (2026). Latent patterns of urban mixing in mobility analysis across five global cities. Nature Cities, accepted. Pre-print uploaded on 10/04/2026. Supplementary Materials are not u...

work page 2023

[20] [20]

Loo, B. P. Y., Fan, Z. & Moro, E. Residential and experienced social segregation: the roles of different transport mo des, metro extensions, and longitudinal changes in Hong Kong. Humanit. Soc. Sci. Commun. 11, 1439 (2024)

work page 2024

[21] [21]

& Schmidt, T

Athey, S., Ferguson, B., Gentzkow, M. & Schmidt, T. Estimating experienced racial segregation in US cities using large-scale GPS data. Proc. Natl. Acad. Sci. U. S. A. 118, e2026160118 (2021)

work page 2021

[22] [22]

& Wang, Q

Han, Y., Liao, P., Li, W. & Wang, Q. R. Visitation patterns reveal service access disparities for ageing populations in the USA. Nat. Hum. Behav. https://doi.org/10.1038/s41562-025-02285-4 (2025)

work page doi:10.1038/s41562-025-02285-4 2025

[23] [23]

& Glaeser, E

Cook, C., Currier, L. & Glaeser, E. Urban mobility and the experienced isolation of students. Nat. Cities 1, 73–82 (2024)

work page 2024

[24] [24]

Jung, H., Lee, J. S. & Kim, S. Beyond Population Density: A New Framework for Compact Cities through Amenity Survival Analysis. Cities 168, 105535 (2026)

work page 2026

[25] [25]

& Burgess, R

Jenks, M. & Burgess, R. Compact Cities: Sustainable Urban Forms for Developing Countries (Routledge, 2000)

work page 2000

[26] [26]

Chen, C., Bian, L. & Ma, J. From Traces to Trajectories: How Well Can We Guess Activity Locations from Mobile Phone Traces? Transp. Res. Part C Emerg. Technol. 46, 326–337 (2014)

work page 2014

[27] [27]

Innovations in Travel Survey Methods

National Research Council Transportation Research Board. Innovations in Travel Survey Methods. Transportation Research Record no. 1412 (National Academy Press, 1993)

work page 1993

[28] [28]

& Madre, J.-L

Armoogum, J. & Madre, J.-L. Weighting or imputations? The example of nonresponses for daily trips in French NPTS. J. Transp. Stat. 1, 53–63 (1998)

work page 1998

[29] [29]

& Stopher, P

Jones, P. & Stopher, P. R. (eds) Transport Survey Quality and Innovation (Pergamon, 2003)

work page 2003

[30] [30]

Moore, J. C. & Durrant, G. Non-response bias risks in sample sub-groups in Understanding Society: the UK Household Longitudinal Study. Understanding Society Working Paper Series 2025 -03 https://www.understandingsociety.ac.uk/wp-content/uploads/working-papers/2025-03.pdf (2025)

work page 2025

[31] [31]

Discrete Global Grid Systems as scalable geospatial frameworks for characterizing coastal environments

Bousquin, J. Discrete Global Grid Systems as scalable geospatial frameworks for characterizing coastal environments. Environ. Model. Softw. 146, 105210 (2021)

work page 2021

[32] [32]

Salazar-Miranda, A. et al. Exploring the social life of urban spaces through AI. Proc. Natl. Acad. Sci. U. S. A. 122, e2424662122 (2025)

work page 2025

[33] [33]

Loo, B. P. Y. & Fan, Z. Social interaction in public space: Spatial edges, moveable furniture, and visual landmarks. Environ. Plan. B Urban Anal. City Sci. 50, 2510–2526 (2023). Fan, Z., Loo, B.P.Y., Duarte, F., Ratti, C., & Moro, E. (2026). Latent patterns of urban mixing in mobility analysis across five global cities. Nature Cities, accepted. Pre-print ...

work page 2023

[34] [34]

M., Batty, M

Roth, C., Kang, S. M., Batty, M. & Barthélemy, M. Structure of Urban Movements: Polycentric Activity and Entangled Hierarchical Flows. PLoS ONE 6, e15923 (2011)

work page 2011

[35] [35]

Lindeman, R. H. Introduction to Bivariate and Multivariate Analysis (Scott Foresman, 1980)

work page 1980

[36] [36]

Aston, L. et al . Addressing Transit Mode Location Bias in Built Environment -Transit Mode Use Research. J. Transp. Geogr. 87, 102819 (2020)

work page 2020

[37] [37]

Weng, G., Kim, M., Ahn, Y. -Y. & Moro, E. Beyond Distance: Mobility Neural Embeddings Reveal Visible and Invisible Barriers in U rban Space . Preprint at https://doi.org/10.48550/arXiv.2506.24061 (2025)

work page doi:10.48550/arxiv.2506.24061 2025

[38] [38]

& Guimerà , R

Cabanas -Tirapu, O., Danús, L., Moro, E., Sales -Pardo, M. & Guimerà , R. Human mobility is well described by closed-form gravity-like models learned automatically from data. Nat. Commun. 16, 1336 (2025)

work page 2025

[39] [39]

& González, M

Jiang, S., Ferreira, J. & González, M. C. Clustering daily patterns of human activities in the city. Data Min. Knowl. Discov. 25, 478–510 (2012)

work page 2012

[40] [40]

H., Parra, D

de Sá, T. H., Parra, D. C. & Monteiro, C. A. Impact of travel mode shift and trip distance on active and non-active transportation in the São Paulo Metropolitan Area in Brazil. Prev. Med. Rep. 2, 183–188 (2015)

work page 2015

[41] [41]

A., Wilby, D

Fairnie, G. A., Wilby, D. J. & Saunders, L. E. Active travel in London: the role of travel survey data in describing population physical activity. J. Transp. Health 3, 161–172 (2016)

work page 2016

[42] [42]

Forrest, T. L. & Pearson, D. F. Comparison of Trip Determination Methods i n Household Travel Surveys Enhanced by a Global Positioning System. Transp. Res. Rec. 1917, 63–71 (2005)

work page 1917

[43] [43]

OSMnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks

Boeing, G. OSMnx: New methods for acquiring, constructing, analyzing, and visualizing complex street networks. Comput. Environ. Urban Syst. 65, 126–139 (2017)

work page 2017

[44] [44]

R., Pentland, A

Yabe, T., García Bulle Bueno, B., Frank, M. R., Pentland, A. & Moro, E. Behaviour-based dependency networks between places shape urban economic resilience. Nat. Hum. Behav. 9, 496–506 (2025)

work page 2025

[45] [45]

& Vetter, C

Luxen, D. & Vetter, C. Real -time routing with OpenStreetMap data. in Proceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems 513 –516 (ACM, 2011)

work page 2011

[46] [46]

Goodchild, M. F. First law of geography. in International Encyclopedia of Human Geography (eds Kitchin, R. & Thrift, N.) 179–182 (Elsevier, 2009)

work page 2009

[47] [47]

On the First Law of Geography: A Reply

Tobler, W. On the First Law of Geography: A Reply. Ann. Assoc. Am. Geogr. 94, 304–310 (2004). Fan, Z., Loo, B.P.Y., Duarte, F., Ratti, C., & Moro, E. (2026). Latent patterns of urban mixing in mobility analysis across five global cities. Nature Cities, accepted. Pre-print uploaded on 10/04/2026. Supplementary Materials are not uploaded. 26

work page 2004

[48] [48]

C., Hagenbuchner, M

Scarselli, F., Gori, M., Tsoi, A. C., Hagenbuchner, M. & Monfardini, G. The graph neural network model. IEEE Trans. Neural Netw. 20, 61–80 (2008)

work page 2008

[49] [49]

Zhai, J., Zhang, S., Chen, J. & He, Q. Autoencoder and its various variants. in 2018 IEEE International Conference on Systems, Man, and Cybernetics (SMC) 415–419 (IEEE, 2018). Correspondence Correspondence should be addressed to Professor Becky P.Y. Loo at bpyloo@hku.hk. The aggregated data for repeating the analysis are available via this repository (htt...

work page 2018