When big data fails: Researcher from ITU unveils gaps in mobility data  

In today’s world, whether you are commuting to work or fleeing a crisis, travel cards and smartphones leave behind digital traces, that provide valuable insights into human movement. These mobility data, collected through interactions with cell towers and GPS satellites, serve as a powerful resource for decision-makers, urban planners, and health authorities. But what if these digital maps of human activity only reflect the experiences of some groups while overlooking others?
 
This is the central question posed by Vedran Sekara, Associate Professor at the IT University of Copenhagen. Over the past year, with support from the Digital Research Centre Denmark (DIREC), he has explored this issue in the research project “Understanding Biases and Diversity of Big Data used for Mobility Analysis.”

The hidden inequality in data

Today, mobility data is employed in a wide range of applications, from urban planning and epidemic tracking to disaster management. However, Sekara has identified alarming biases within the algorithms that process these data – biases he seeks to correct.
 
According to Sekara, most mobility data is skewed. In some datasets, he found that half of the mobility data is generated by the wealthiest 20% of the population, while only 5% comes from the poorest 20%.
 
“There are distinct patterns in smartphone ownership – typically, it is the wealthy, men, and the highly educated who possess them. As a result, women, the elderly, and children are significantly underrepresented in mobility data collected via phones. Despite this bias, these data are used to shape societal decisions,” Vedran Sekara explains.
 
The biases in mobility data are not just an academic concern – they have real world consequences for resource allocation.
 
“There is a significant risk that our efforts are disproportionately directed towards helping those who generate the most data, who are often the wealthiest and most powerful in society, while the needs of the poor are overlooked. During earthquakes or epidemics, we have seen that poor regions are less represented in the data, which could lead to fewer resources being allocated to them. Since we have better data for the wealthier areas, more aid is sent there”, says Vedran Sekara.

Searching for solutions

In his research project, Vedran Sekara and his team are working to develop new algorithms to correct these biases. However, the challenge is proving to be highly complex.
 
“We found that to correct these biases, we need to create specific models for each local area. Every area has its own unique biases, which means we can’t apply the same algorithms universally,” says Vedran Sekara.
 
With DIREC’s support of 500,000 DKK, Vedran and his team have published several articles on the subject. The project has also opened doors for additional funding, potentially enabling further research on biases in mobility data.
 
“DIREC’s support has been invaluable. They have provided the necessary funding to explore the problem on a smaller scale, laying the groundwork for larger projects in the future. DIREC has truly been a catalyst for this research,” concludes Vedran Sekara.
 
All data used in Vedran Sekara’s research is pseudonymized, in compliance with both European and American data regulations.
 
Learn more about the project here.