Published on
Summer 2020-Fall 2021

COVID-19 Contact Tracing Agent-Based Model

Stochastic models created in python and and Agent-Based Models created in NetLogo to simulate the effects of reactive and preventative contact tracing on the spread of viruses

Role

Primary Investigator, Software Engineer

Skills Used

  • NetLogo Agent-Based Modeling
  • Python Stochastic Modeling

Collaborators

none

Background

In my Engineering and Epidemics class, we were told to create anything we wanted for our final project. I decided to learn the NetLogo software language and model the effects of different contact tracing methods on the spread of a virus, like COVID-19. The following semester, I joined an interdisciplinary team of undergraduates, graduates, and professors from Duke's Sanford School of Public Policy to assess the privacy implications of COVID-19 contact tracing. Centralized and decentralized contact tracing were widely discussed and still remain a public safety concern. I presented my findings from these agent-based models and my team's findings regarding privacy at Intel's Future for Privacy Forum in October, 2020.

Links

Final Video Summary

Schematics

Reactive Quarantine

Preventative Isolation

Stochastic Models

Reactive Quarantine

diagram of reative model
reactive quarantine stochastic model

Preventative Isolation

diagram of preventative model
preventative isolation stochastic model

Agent-Based Models

No Contact Tracing

Reactive Contact Tracing

Preventative Contact Tracing

Results

Raw Results After 10 Trials

average max infections and average max total deaths for different types of CT

Summary of Results

  • Agent-based models have a greater capacity to model contact-tracing interventions

  • Increasing percent of population using contact-tracing apps, both reactive and preventative, decreases max # infections and total deaths

summary of results including 1 degree versus 2 degree isolation difference

  • More trials necessary and can expand models to inform public policy

Privacy Implications

The Future for Privacy Forum was an international conference sponsored by Intel in 2020 that brough together the world's leading experts in privacy and tech. I worked with a team from the Sanford School of Public Policy at Duke and presented our collective findings.

Check out my team's proposal for the Future for Privacy Forum.

FPF forum

(Bonus: A Picture of Me Presenting At the Forum!)

Reflection

This project was very applicable when I did it because COVID-19 had only recently been declared a pandemic and there was serious panic about infection and death rates. This project show that when empirical evidence is limited, it is possible to acquire theoretical data through many different types of models, including predictive, stochastic, and agent-based models. NetLogo was such an awesome language to learn, and I look forward to using it in the future for applications other than modeling viral spread.