Big data is not always enough to predict the course of a particular situation, according to a team from the Ludwig-Maximilians-University in Munich, Germany (1) When big data analysis reaches its limits, information obtained from individual observations can be extremely valuable.
Social researchers may have unprecedented access to enormous amounts of big data, but there are still many aspects of social events that are virtually impossible to predict. Situations like national elections and the coronavirus pandemic outbreak are just a few examples where current big data analysis may not provide a complete picture. Changes develop rapidly, sometimes with unexpected turns and twists. This makes it impossible to gather up-to-date passive data to allow researchers to follow the trajectory of such events.
This is where a team of German researchers tested a method called “human social sensing” to uncover information that goes beyond what can be found from digital trace data. Frauke Kreuter, Professor of Statistics and Data Science in the Social Sciences and the Humanities at LMU, is now using this method with the global “Covid Trends & Impact Survey” to predict the course of the pandemic.
“We shouldn’t focus too narrowly on the analysis of digital trace data. It’s a mistake to ignore the fact that people are equipped with sensory capacities. Those can be particularly useful in areas that are difficult to capture with data from behavioural traces”, said Kreuter.
Kreuter and her team showed that interviews with individual people still have a vital role to play, even when big data often attracts more attention. “One should never lose sight of the fact that surveys can serve as a very valuable source of supplementary information”, continued the researcher.
The idea is that individuals can be used as “social sensors”. In other words, each person can provide information not only about themselves but also about others in their local environment. This is an aspect that is often ignored in surveys, but Kreuter defends that, during the current pandemic, it is vital to take advantage of everything each individual knows about what’s happening around them.
The team used this approach in the “Covid Trends & Impact Survey”, which is trying to predict the development of the coronavirus pandemic on a global level. The survey started in April 2020, and more than 55 million people around the world have already participated. Relying on the “social sensor” method, one of the questions included in the survey is whether they personally knew anybody with COVID-19 symptoms.
The team believes this particular question is vital, and answers so far seem to be a solid predictor for the development of the pandemic. The only requisite to allow people to participate in this survey is that they must be a representative sample of the population of interest, allowing them to be reliable social sensors. The ability to “sense” our environment – and thereby understand the thoughts and actions of other people – is what allows us to fit into our social group and learn more than any big data dataset.
(1) Galesic, M., Bruine de Bruin, W., Dalege, J. et al.Human social sensing is an untapped resource for computational social science. Nature (2021). https://doi.org/10.1038/s41586-021-03649-2