COVID & the city

A model of social inequality in the pandemic

TU Delft, Urban Studies

Lunch & Learn 6th October 2020

Clémentine Cottineau

CoVprehension collective

Drawings by Odile Plattard, 2020

March 2020

Get informed

Panic

Act

Explain

move to NL!

Connect

Understand

Test theory

Agent Based Modelling

Simulation modelling framework

Agents:

  • attributes
  • aims & goals
  • memory
  • heterogeneous 

Dynamics:

  • rules of action
  • evolution (≠equilibrium)
  • bottom-up

Environment:

  • spatial
  • local interactions
  • can evolve too

Why Model?

 

Predict: validating theories in natural sciences.       > Ferguson's model: needs time, data, hypotheses

  1. Explain (very distinct from predict)
  2. Guide data collection
  3. Illuminate core dynamics
  4. Suggest dynamical analogies
  5. Discover new questions
  6. Promote a scientific habit of mind
  7. Bound (bracket) outcomes to plausible ranges
  8. Illuminate core uncertainties.
  9. Offer crisis options in near-real time
  10. Demonstrate tradeoffs / suggest efficiencies
  11. Challenge the robustness of prevailing theory through perturbations
  12. Expose prevailing wisdom as incompatible with available data
  13. Train practitioners
  14. Discipline the policy dialogue
  15. Educate the general public
  16. Reveal the apparently simple (complex) to be complex (simple)

Epstein, J. M. (2008).  Journal of Artificial Societies and Social Simulation, 11(4), 12.

or at least compare them (Swedish policy versus total lockdown)

engage in dissemination with feedback by asking for their questions

Effect of density, of lockdown, visualise hospital beds

by highlighting unknowns

simplification of models

Agent Based Modelling

Simple integrated platform: Netlogo

Social differentiation in COVID spread during the first wave

  • First affected: travelling professionals, celebrities, holiday makers from ski resorts, etc.
  • Then: "the virus is blind, everyone can catch it" / "The great equalizer"
  • Finally: higher prevalence of the disease among low income groups, deprived neighborhoods, minorities.

 

> Can we integrate such three facts into one model?

Work status: remote or not

Some can work from home whereas others are exposed on their workplaces (shops, hospitals, restaurants, nuclear sites, etc.)

Work status: remote or not

We identify essential & non essential jobs

Essential jobs are filled by workers preferentially within a radius r, non-essential jobs are less constrained

Residential status:

individual or collective

For some, being home means being exposed to their household only. 

For others, it's sharing doors, lifts, corridors, courtyards & bins with others.

Transmission of the virus

When people are put in contact, their probability of being infected is the same.

Lockdown measures

After a certain number of fatalities,

a lockdown is automatically imposed.

People from non-essential sectors work from home

Evaluation of simulations

  • Dynamic of the epidemics (susceptible, infected, recovered, dead)
  • Distribution by social class

Privi-

leged

working

class

work 

on

site

collec-

tive

living

work 

re-

mote

indi-

vidual

living

"middle" class

Online simulator

Offline simulator

Motor of simulation

Model results

Model results

t + 35

t + 70

The epidemic follows classic SIR dynamics

Privileged workers are put in contact first (longer commutes)

Then they are overtaken by lower and middle class infections

Model results

At the end of this simulation, a larger % of working class agents have been affected by the virus.

 

They also have a higher mortality rate (does not take into account co-morbidities & poorer access to health care).

 

Model results

Through generative simulation, we can trace through which pathways agents get infected (difficult in real life)

 

> The virus is blind but the exposure to the virus has been socially biased.

Additional experiment

Limits

  • Definition of "social class" is highly contextual to the epidemic question 
  • Urban, medical and demographic factors still absent from the model
  • Exploration of the model remains to be done (calibration, sensitivity, robustness analysis, adequacy with empirical data, etc.)

Perspectives

  • Keep multilevel & integrated approach to simulate human interactions dynamically in cities
  • Add dimensions of inequality (on top of work status & residential setting), such as education level, income, etc.
  • Calibrate on empirical individual data
  • Use to test policy scenarii

Conclusion

  • Stimulating collective modelling experience in face of a pandemic
  • Inquiring, testing & sharing: our tools
  • Urban models of inequality

Going further

The website: https://covprehension.org/en/

 

The repository: https://github.com/covprehension

 

A collective publication:

  • CoVprehension collective, 2020, "Understanding the current COVID-19 epidemic: one question, one model", Review of Artificial Societies and Social Simulation, COVID-19 thread, http://rofasss.org/2020/04/30/covprehension/

 

More questions: https://framaforms.org/covprehension-challenge-us-ask-your-questions-1585581105 

Made with Slides.com