Drugs that are developed for individual patients, car insurance that is tailored to driving styles, or burglars who are caught before they commit a crime: Big data and better algorithms ensure that predictions become more accurate and more individualised. Bielefeld sociology professor Dr Elena Esposito is studying the effects this has on society. The European Research Council is funding her research project “Predict” with an ERC Advanced Grant.
Police that can predict future crimes: it sounds like the plot from the science fiction thriller ‘Minority Report’ in which Tom Cruise hunts murderers before they can commit their crimes. Such a scenario is not that far distant from reality—already today, for example, the police are using computers to prevent burglaries. ‘Predictive policing’ aims to use enormous datasets and complex algorithms to predict offences as accurately as possible.
Similar developments are to be found in insurance and medicine. Here as well, more data and better algorithms are making more accurate and more individualized predictions. Car insurance, for example, is using data collected on driving style when calculating tariffs; and cancer medicine is promising personalized treatment methods for patients with specific genetic characteristics.
Predictions for single individuals instead of general prognoses
‘Uncertainties, even though such predictions claim to manage them, continue to play an important role in our society,’ says Professor Dr Elena Esposito from the Faculty of Sociology. In her research project ‘Predict’, she is studying the social effects that come from predictions based on algorithms dealing with large amounts of data. Her project is being funded by the European Research Council (ERC).
‘In the past, our society developed mechanisms for dealing with our shared uncertainty about the future. In medicine, for example, clinical studies are used to determine which treatment will have, on average, the greatest chance of delivering a cure,’ says Esposito. Such predictions are based on frequencies of occurrence within a representative group. These are then used to derive probabilities for a great number of people; in other words, uncertain prognoses that nonetheless possess a certain degree of validity for everybody. In contrast, the goal of a personalized insurance sector, of precision medicine, and of predictive policing is to make predictions for single individuals—regarding whether they will cause an accident, fall sick, or commit a crime.
© Bielefeld University/ M. Adamski
Three aspects: individualization, generalization, and bias
‘Algorithmic predictions are becoming increasingly important in both research and practice. Nonetheless, we still do not know enough about the social dimension of this trend,’ says Esposito. She is working on three different aspects of algorithmic predictions: individualization, generalization, and bias.
She is using the case of the personalized insurance sector to study individualization: will personalized insurances disadvantage certain groups of individuals—for example, if those exposed to greater risks suddenly have to pay more? Who can still have any interest in selling or buying an insurance policy if they already know what their future damage will be?
‘Generalization is particularly important in precision medicine, because the development of algorithms is based on training data. Can these algorithms be applied to cases in which other data may play a role—data that had previously not been considered?’ asks Esposito. Finally, she is using the example of predictive policing to study the aspect of bias: if existing data are biased because, for example, members of certain ethnic groups are arrested more frequently, will the algorithm then strengthen this bias?
In all three areas of research, Eposito is also focusing on the interplay between the traditional probabilistic methods used to predict the future and the new algorithmic forms of forecast—because both methods are still being applied today.
Applying Luhmann’s systems theory
‘Proceeding from the single areas of research, I am also studying the effects on the overall system of society, and I am doing this referring to Luhmann’s systems theory,’ says Esposito. She is one of the leading representatives of sociological systems theory. Bielefeld sociology professor Niklas Luhmann supervised her doctorate in 1990. Since 2016, Esposito has been a professor of sociology at Bielefeld University, and with her interdisciplinary networking, she has also been working in parallel at the Università di Modena e Reggio Emilia in Italy.
For her research project ‘Predict’, Esposito is receiving an ERC Advanced Grant from the European Research Council (ERC). This grant is awarded to outstanding researchers who are established, leading principal investigators in their field of research. It is worth 2.1 million euro over a period of five years. Officially, the project will be starting at the beginning of 2020. However, the first studies are already up and running — particularly those on precision medicine and personalized insurance.
This article is a publication from ‘BI.research‘, the research magazine of Bielefeld University. The current issue of the magazine was published in November 2019.
