17 September 2019

Researchers at the Institute for Analytical Sociology, IAS, have been awarded SEK 13 million to use machine learning to provide in-depth knowledge about inequalities in society. “The Swedish population registers are a gold mine”, says Maria Brandén, leading the project.

Maria Brandén, researcher at IAS, in Campus Norrköping. Photo credit: Mikael Sönne

A cross-disciplinary research group

Researchers from seven departments and divisions in several European countries are members of a group that has been awarded around SEK three million a year for four years from the Swedish Research Council. To put it simply, machine learning uses algorithms that have been developed to reveal patterns and correlations in data. The method is used in, for example, disease diagnosis, facial recognition and realtime translation.

In the social sciences, machine learning has been relatively little used, at least until now.

“One important reason for this is that huge amounts of data are required, and in reality it is only Sweden that has sufficiently extensive population registers. It is also common in the quantitative social sciences to start with a specific hypothesis and then test it, while machine learning involves a more inductive approach, allowing the data to find its own voice. Many people find this a new way of thinking”, says Maria Brandén.

Definition of neighbourhood

The research programme will combine machine learning with large amounts of data from registers to develop and refine social indicators important for our understanding of inequality. Such an indicator is “neighbourhood”, an important concept that is often used in a rather imprecise manner. The researchers are planning to develop more exact definitions that consider the population on the basis of not only geographical region but also factors in the physical landscape that influence interactions between people.

Maria Brandén.

Machine learning will be also used to improve our understanding of various social processes, in particular social mobility. The method will make it possible to see how the same factor – such as education, economy or social networks – can influence different individuals in different ways. The study of such heterogeneous effects is a major advantage of machine learning.

“We will be able to study smaller groups and will not be compelled to focus on averages, as social scientists usually do. The knowledge we gain will be more insightful and deeper, quite simply”, says Maria Brandén.

In order to understand the effects of various political initiatives, the researchers will attempt to create what they refer to as a “policy machine”. This will enable them to investigate which measures influence which individuals, and in which way.

Register-based research

The research group includes people with a background in sociology, demographics, statistics and computer science.

“It will be truly cross-disciplinary: I’m really looking forward to working with this group”, says Maria Brandén.

The research grant has been granted following a call for applications from the Swedish Research Council for register-based research, for which a total of SEK 102 million has been granted for the period 2020-2023.

• More information about register-based research is available at the website for Swedish Research Council.

Read more about IAS

Latest news from LiU

Woman outdoors.

Linda Carlsson is the new Director of Communications at LiU

Linköping University has appointed Linda Carlsson as new Director of Communications. She comes most recently from the Swedish Council for Higher Education.

Four persons in chairs on a stage.

AI can boost financial decision making

Do you have limited financial knowledge, or prefer not to think about financial issues? Then there is great potential that AI can help, according to LiU researcher Kinga Barrafrem.

Nathalie Hallin and Hajdi Moche in conversation.

Religious people are not more generous – with one exception

Believers are no more generous than atheists – at least as long as they don’t know what the recipient believes in. This is the conclusion of a study carried out at LiU