Northeastern London Wins £327,500 Grant to Model Pandemic Behavior

Northeastern University London’s Network Science Institute secured a £327,500 Leverhulme Trust grant to build epidemic models that integrate human psychology and social identity, aiming to fix the “blind spots” exposed by the UK COVID-19 Inquiry.

Northeastern University London and Sussex University won a $327,500 Leverhulme Trust grant to simulate how millions of agents respond to public health messaging.

The project, announced this year and run with Sussex University, will simulate how millions of distinct “agents” respond to public health messaging based on their unique combination of political, professional, and social traits. Researchers argue current models fail because they treat populations as uniformly rational actors, a flaw Baroness Heather Hallett’s inquiry highlighted when it concluded British decision-makers were “effectively blind” to the virus’s early spread.

Why do current epidemic models miss the mark?

Traditional compartmental models — such as Susceptible-Infected-Recovered (SIR) frameworks — rely on average contact rates and homogeneous mixing assumptions. They cannot capture the behavioral feedback loops where risk perception alters contact patterns, which in turn alters transmission dynamics.

During COVID-19, this limitation meant models struggled to predict adherence to non-pharmaceutical interventions (NPIs) like masking or isolation, particularly when adherence fractured along political or cultural lines. The UK COVID-19 Inquiry’s first report, published in 2024, explicitly criticized the lack of behavioral science integration in the Scientific Advisory Group for Emergencies (SAGE) modeling pipeline.

How does the “social identity” approach differ?

Instead of a single “average person,” the NetSI model assigns each software agent a “social identity DNA” — a vector of characteristics including political conviction, profession, religion, education, and social habits. Project leader István Kiss, professor at the Network Science Institute, explained the mechanism using a concrete example: a Republican nurse in the US who follows strict infection control at work but attends anti-vaccine protests in her personal time.

“When she goes to the hospital, she obeys by all the rules, she wears masks, she even maybe gets her vaccine,” Kiss told Northeastern Global News. “But in the afternoon, she’s going to go out and protest with other Republicans against vaccination, she might not wear a mask, she might behave completely differently – because she belongs to that political conviction.”

“There is a kind of DNA to every individual in terms of what social identities you have and how they kind of manifest themselves.”

— István Kiss, Professor, Network Science Institute, Northeastern University London, told Northeastern Global News

Northeastern Global News

This granularity allows the simulation to capture “context switching” — where the same individual exhibits high compliance in one setting and low compliance in another — a dynamic invisible to aggregate models.

In Plain English: The Clinical Takeaway

  • Current pandemic models assume people act rationally and uniformly; this project builds “virtual people” with real-world contradictions (e.g., a nurse who masks at work but protests vaccines off-duty).
  • The goal is not to predict the next virus, but to give governments a stress-test tool for public health messaging before a crisis hits.
  • Funding comes from the Leverhulme Trust (£327,500), a UK charity, ensuring academic independence from pharmaceutical or political interests.

What data drives the simulation?

Kiss cited research from his Boston colleagues at Northeastern’s Network Science Institute showing a “consistent gap” between how Republicans and Democrats acted during COVID-19 despite receiving identical federal guidance.

The simulation will also incorporate cross-national differences: Kiss said that during the last pandemic, cooperation across the political spectrum was more effective in Britain than in the US, though the causal drivers — trust in institutions, media environment, welfare structures — remain unresolved. Additional variables include population density, household composition, workplace mixing patterns, and leisure-driven contact rates.

Can a computer model actually change policy?

Andreia Sofia Teixeira, associate professor at NetSI and co-investigator, emphasized the tool’s advisory role. “We are not in control of any emerging pathogens,” she told Northeastern Global News. “But we will be able to provide the tool that can hopefully better inform what could be the best preventive measures.”

The output aims to help agencies like the UK Health Security Agency (UKHSA) or the US Centers for Disease Control and Prevention (CDC) test messaging strategies in silico — for example, comparing the epidemic trajectory under a mandate framed as “personal protection” versus “community duty” across different demographic segments.

— Andreia Sofia Teixeira, Associate Professor, Network Science Institute, Northeastern University London, told Northeastern Global News

How does this fit into global pandemic preparedness?

The World Health Organization’s Pandemic Influenza Risk Management guidance and the subsequent Pandemic Agreement negotiations both stress the need for “whole-of-society” approaches and behavioral insights. However, operational tools remain scarce. The NetSI project addresses a specific gap identified by the UK COVID-19 Inquiry Module 1 report: the failure to model “adherence fatigue” and identity-based resistance.

Model Feature Traditional SIR / Compartmental NetSI Agent-Based (Proposed)
Population Representation Homogeneous averages Heterogeneous agents with “Social Identity DNA”
Behavioral Response Fixed parameters (e.g., static compliance rate) Dynamic, context-dependent (work vs. leisure vs. protest)
Identity Variables None Politics, profession, religion, education, social habits
Policy Testing Capability Limited to NPI intensity/timing Message framing, targeting, identity-specific interventions
Calibration Data Case counts, mobility aggregates Partisan behavior gaps, cross-national surveys, contact surveys

Contraindications & When to Consult a Doctor

Public health authorities should treat model outputs as scenario-planning aids, not forecasts. Decision-makers must validate simulated adherence patterns against real-time behavioral surveillance (e.g., the ONS COVID-19 Infection Survey methodology) before operationalizing policies. Over-reliance on unverified agent rules risks “garbage in, gospel out” errors.

The Leverhulme Trust, established in 1925 under the will of the 1st Viscount Leverhulme, dispenses over £100 million a year, making it one of the largest providers of research funding in the United Kingdom. It operates across all academic disciplines with a focus on originality, significance, and the removal of barriers between traditional disciplines.

The project begins in January, funded by a $327,500 Leverhulme Trust grant awarded to Northeastern University London and Sussex University.

References

  • UK COVID-19 Inquiry.
  • World Health Organization.
  • Kiss IZ, et al.
  • Leverhulme Trust.
Photo of author

Health Desk

Health Desk at Archyde.

Baltimore Health Officials Investigate Legionnaires’ Cases at Residence Inn Near Johns Hopkins