Development and evaluation of Dona, a privacy-preserving donation platform for messaging data from WhatsApp, Facebook, and Instagram.

Abstract

Social interactions are a fundamental aspect of human life, yet, their objective and naturalistic measurement remains challenging for scientific research. This challenge can be addressed using digital communication data. To this end, we have developed Dona, an open-source platform for donating messaging data from WhatsApp, Facebook, and Instagram. Given the highly sensitive nature of messaging data, we ensure participant privacy through rigorous data minimization. Dona removes all sensitive information on the user side prior to donation, retaining only de-identified meta-data such as message length and timestamps. This paper presents an overview of the platform, a deployment guide, and example use cases. In addition, we evaluate the informativeness of minimized messaging data for studying social interactions with two approaches. First, we conducted a user study in which 85 participants donated their data, received visualizations of their messaging behavior and evaluated the informativeness of this visual feedback. Second, we performed a quantitative analysis using over 1500 donated chats to confirm whether minimized messaging data captures known aspects of human interactions, such as interaction balance, heterogeneity, and burstiness. The results demonstrate that minimized, de-identified messaging data reflects informative interaction features as assessed by both self-reports and objective metrics. In conclusion, Dona is a donation platform well suited for sensitive contexts in which researchers aim to balance participant privacy with the acquisition of objective and informative data on social interactions.