Oxford Martin School

Professor Alessandro Vespignani of Indiana University Bloomington (USA) spoke at the 21st Century School on 25 February about mobility, pandemics and the challenges of predicting the behaviour of techno-social systems.

"We are entering the age of the social collider," he said. "Everything we do leaves traces online." Vespignani was speaking not as a privacy researcher, nor as a concerned private citizen, but as a researcher who is interested in networks and social behaviour on a grand scale.

In seeking to understand the statistical bases of society, Vespignani uses data networks to model the spread of diseases like H1N1 ; patterns of human mobility are key to understanding those of viruses as well.  His group uses multiple sources of data - census data on commuter movements, demographics of population density, and airline schedules for long distance travel - to model and project travel patterns looking forward.

While our notions of predictions are often based on ones we know, such as weather forecasts, techno-social systems like the ones Vespignani studies are harder to predict. The knowledge that a storm is coming may cause us to stay inside that day, but the weather arrives regardless. Not so with infectious diseases, where a change in people's behaviour can alter the outcome - for both better and worse.

The biggest change in people's behaviour since the last major pandemic in 1918 is the much greater mobility we have today. In addition to the millions of local commutes made each day, thousands of planes take off and land across the country or around the world, potentially expanding the global reach of pathogens. However, though containment may seem like a logical strategy during an outbreak, Vespignani presented mathematical models to show why it doesn't work. In order for it to be effective, travel would need to be restricted by more than 90%, which is simply not feasible in the modern world.

He presented several other models for anticipating the statistical likelihood of an infectious outbreak, from large-scale computational models to a simplistic S-I-R model that tracks those who are susceptible, infected, and have recovered from an infection, but warned that the results are statistical likelihoods rather than fact. This problem is exacerbated by chaos theory - also known as the butterfly effect - where small differences in initial conditions can lead to widely divergent outcomes. It is also based on the assumption that human mobility patterns don't change radically in aggregate, which may in fact happen if there is panic about an outbreak of disease.

"What advice would you give to policymakers about the spread of a pandemic?" asked someone from the audience. Vespignani explained his role as offering information, and demonstrating to what degree it can be relied on, but sees the role of interpretation as one of policy.  The social component is the most difficult one to predict, he says.

Join the discussion

• Post a comment – click ‘add comment’ below to share your thoughts
• Subscribe to the RSS podcast feed for an audio recording of the seminars 

This blog was written by Susan Curran, Web and Publications Officer, at the Institute for Science, Innovation and Society.