Share precise location: a natural experiment on Twitter users’ privacy concerns

The explosion of location-based services allowed for a new range of technological application influencing our every day life in several contexts including health, sport, transportation, entertainment, work, and personal life. However, it also represents a new source of concerns regarding personal or sensitive information shared by the user and/or stored on the platform, which cannot be simply overcome by data anonymisation [1].
Here we show how an open and customizable user interface may support the achievement of the level of privacy desired by the user. In April 2015, the Twitter interface has been modified, allowing to toggle on or off the sharing of user’s precise location (latitude and longitude) for each tweet written. This change produced a massive drop in the number of geo-referenced tweets, the extent of the which varies significantly between countries.
Comparing these empirical observations with a survey performed over 15 different regions of the world [2], we find that the observed drop is smaller in countries where users more aware of the privacy issues and where, consequently, the quantity of geo-referenced tweets were already low before the interface change. Remarkably, we also find that, as a consequence of the new interface, the fraction of geo-located tweets now really aligns to different countries' willingness to trade privacy for convenience for the “social persona”.
Finally, using multiple regression analysis, we predict the average answers in the survey using the characteristics of the drop and the country GDP [3] (Fig.1). Thanks to this reconstruction, we can extend worldwide the results of the survey, using this natural experiment on Twitter users for inferring the users’ awareness and real desires concerning online privacy.