Measuring sociability through traces rather than questions

Communication records provide continuous series, on very large samples, without drawing on people's memory. They have made it possible to trace curves by age and to measure the stability of contact hierarchies. They correct a long-standing flaw of questionnaires: people remember imperfectly whom they have seen or called.

Proximity devices, badges or sensors worn during experiments, measure face-to-face interaction with a precision inaccessible to questionnaires. The European SocioPatterns project has thus described contacts in schools, hospitals and conferences.

A recorded contact is not a relationship, and its absence is not the absence of a relationship. A person may be very close to another without any exchange recorded over the observed period.

The observed channel is also partial: an analysis covering one operator or platform ignores everything passing elsewhere, and that share varies systematically by age and background.

The difficulty is also one of interpretation. Frequent exchange may reflect closeness, practical coordination, as between colleagues or between the parents of the same child, or mere obligation. Traces record the form of the exchange, never the meaning it has for those who conduct it.

The data come from an operator's customers or a service's subscribers, which do not constitute a representative sample. The usual weightings do not apply, for want of knowing the reference population.

That limit is structural and distinguishes this work from probability surveys, whose weakness is self-report but whose strength is controlled representativeness. The volume of data does not make up for that flaw: a very large, ill-defined sample remains ill-defined.

These data belong to firms. Making them available runs through agreements that determine what may be analysed and published, which raises a question of independence distinct from that of validity.

Several regulated access arrangements have been created in Europe, with uneven results, the main difficulty being to reconcile personal data protection with scientific usefulness. The European Digital Services Act, adopted in 2022, provides for vetted researchers to access the data of very large platforms, though it is too early to gauge its reach.

The most convincing work combines traces and questionnaires on the same people, which allows agreement between the two to be estimated and the gaps interpreted. Nathan Eagle, Alex Pentland and David Lazer thus compared in 2009 the phone records of students and researchers with the friendships they reported.

Those protocols are costly and cover small samples, which places them in a complementary rather than competing position relative to large self-report surveys.