Hate Speech & Misinformation on WhatsApp: Insights from Data Donation Programs

• In Event: What’s Up with WhatsApp? Social Media and Politics in the Global South

Abstract

Over the past few years, alarming press reports assigning blame to WhatsApp usage for a variety of events have proliferated. In countries like Brazil and India, analyses have repeatedly suggested that group-based interactions on WhatsApp distort beliefs among the electorate (see, for instance Perrigo 2019; Bengani 2019; Benevenuto et al 2018), and beyond, that they impact various outcomes, including (but not limited to) individuals’ propensity to engage in hostile, radical or violent behaviors (Chopra 2019, Magenta et al 2018; Ozawa et al 2023).

While academic research has recently started to examine these dramatic narratives, much research remains to be done to evaluate their validity, and to disentangle the mechanisms through which WhatsApp may or may not be associated to these outcomes. Specifically, in order to ascertain the platform’s role in the dissemination of problematic content, as well as the consequences of this dissemination, researchers still need to better understand (1) the type and the style of hateful content and/or disinformation that circulates on the platform, (2) the overall volumes of such contents, (3) their degree of virality, (4) the networks through which such contents are most likely to circulate, (5) the profiles of those forwarding and receiving it and (6) the political, social and contextual factors in which these contents emerge.

Yet, given the many practical, legal and ethical challenges that such research implies, social scientists interested in describing the media diets of WhatsApp users in the Global South – where the app is central to daily communication, networking as well as access to information – have so far not been able to access WhatsApp data on a large scale. As result, the type of high-quality evidence about Facebook and Twitter users’ “information diets” that has existed for some time (Barbera et al 2015, Guess et al 2019) so far does not exist about WhatsApp, despite researchers’ long-standing awareness of the fact that the platform is used to disseminate this type of content in much of the Global South (Tucker et al 2018).

In this paper, we rely on the world’s largest-to-date (WhatsApp) data donation program in India and Brazil to assemble a unique dataset of private WhatsApp groups in both countries and address each of the 6 aforementioned research questions.

Relying on a new data extraction tool called WhatApp Explorer (Chauchard and Garimella 2024) and on an extensive, GDPR-compliant anonymization and privacy protocol developed as part of the ERC POLARCHATS project, we field (the data collection is ongoing at the time of this submission) two large, face-to-face data collection campaigns in a large number of locations in each of the two countries. During these campaigns, diverse sample of users in both countries are able to donate past and future data about their groups/threads, with the assistance of local implementing partners, in exchange for small incentives. Altogether around 3000 users in each country each donate up to 4 months of content about their groups (donating an average of around 5 groups each, as per recent pre-tests), allowing us to analyze the prevalence of specific types of content over vast amounts of data, their virality, to explore the type of groups or communities on which problematic content tend to proliferate, as well as document the profiles of users that tend to forward and/or receive such content.

Authors

• Simon Chauchard, University Carlos 3 of Madrid

• Kiran Kiran Garimella

• Fernando Barros de Mello, Carlos III-Juan March Institute of Social Sciences