The next speaker in this session at the 5th Data Donation Symposium at the Weizenbaum-Institut in Berlin is Eva-Maria Vogel, whose study is a field experiment in the context of South Africa’s upcoming local elections in November 2026 that explores the impact of TikTok influencers on the election. There are already concerns about disinformation during the elections, and this project works with the AfricaCheck fact-checking organisation, which will develop fact-checks of current disinformation to be distributed by the project.
This tests, then, whether a single fact-checker account, a large-scale influencer, or a fleet of smaller-scale influencers are more effective at distributing such fact-checks on TikTok. Data donations in this context provide a proof of compliance by members of the various participant groups, dose of manipulated content received, organic contamination of the non-exposed control group, independent checks on participant self-reports, patterns of overall TikTok use, donor selection effects, and broader rumour exposure.
TikTok data donations will provide many of these data points, but also various unwanted information which will need to be redacted from the donated data. The particular approach taken addresses this by exporting only the relevant data from TikTok, but unfortunately also takes a very long time – and much longer in South Africa than in comparable studies in Europe, for some reason. Some of the TikTok data also appear to be incorrect: they contain likes for content that users did not watch, for instance. Following some 20 smaller-scale influencer accounts is also a tedious task to participants to accomplish.
Further, in the South African context the approach also needs to grapple with different technological contexts: capped and costly mobile data; older phones and mobile-only use; and the impact of slower export timelines on potential drop-out rates. Informed by this, the project expects a high drop-out rate, and incentivises participants by offering payments; however, South African users are generally more likely to participate in such studies.
