MobiSenseUs: Inferring Population Well-Being from Mobile Data

Measures of population well-being usually come from surveys and official statistics. Because they are costly to collect and not always updated frequently, we built MobiSenseUs to test whether aggregated patterns of mobile-phone use could complement them.

We used two pseudonymized mobile datasets from more than one million smartphone users in the United Kingdom. Sixteen features described mobility, application use, and communication; the call and message records contained metadata, not conversational or textual content. We aggregated these features for each local authority and compared them with the UK's Index of Multiple Deprivation (IMD) and survey measures of subjective well-being.

Our models estimated the IMD, which includes income, employment, health, and education, more accurately than self-reported life satisfaction. In a four-quartile classification, accuracy was 66% for average IMD and 46% for life satisfaction. When the task was limited to the highest and lowest quartiles, accuracy rose to 99% and 84%, respectively. Mobility patterns were most informative for deprivation, whereas communication patterns were more useful for life satisfaction.

These are area-level estimates from our models, not measurements of an individual's well-being. The mobile data were collected in 2017–2018, while the available IMD values were from 2015, and the relationships found in the UK may not hold in other places or periods. The patterns may complement surveys and public statistics, but they cannot replace information collected directly from people.

A bar chart compares how accurately mobility, app-use, and communication signals estimated deprivation and life satisfaction.

Combining all three types of mobile behavior worked best for estimating deprivation, while communication signals performed best for life satisfaction.

Publication

Conference paper
PDF DOI
Cite
Formatted citation

Martin Hillebrand, Imran Khan, Filipa Peleja, Nuria Oliver (2020). MobiSenseUs: Inferring Aggregate Objective and Subjective Well-Being from Mobile Data. Proceedings of the European Conference on Artificial Intelligence (ECAI 2020). https://doi.org/10.3233/faia200297

BibTeX
@inproceedings{hillebrand2020mobisenseus,
  author = {Martin Hillebrand and Imran Khan and Filipa Peleja and Nuria Oliver},
  title = {{MobiSenseUs}: Inferring Aggregate Objective and Subjective Well-Being from Mobile Data},
  booktitle = {Proceedings of the European Conference on Artificial Intelligence (ECAI 2020)},
  address = {Santiago de Compostela, Spain},
  year = 2020,
  doi = {10.3233/faia200297},
  file = {/projects/mobisenseus/mobisenseus.pdf},
  cites = 8,
  citesdate = {2026-03-23}
}

MobiSenseUs: Inferring Aggregate Objective and Subjective Well-Being from Mobile Data

Martin Hillebrand, Imran Khan, Filipa Peleja, Nuria Oliver
Proceedings of the European Conference on Artificial Intelligence (ECAI 2020) · 2020

More information