And the final speaker in this session at the 5th Data Donation Symposium at the Weizenbaum-Institut in Berlin is Ericka Menchen-Trevino from BRIC, who will present a browser extension framework for capturing chatbot conversations that seeks to circumvent some of the issues with formal data donation processes. Data donations build on the right for users to download their platform history data, of course, but this is often made deliberately complex and time-consuming by the platforms; the data contained in such packages are often incomplete and poorly documented, too, and contain highly sensitive personalised data that researchers need to strip out from these packages.
A different approach to data donations, then, avoids the use of data download packages, and instead works with a browser extension framework for Chrome and Edge that records what participants see on the Web. Doing so is complicated for chatbot interactions, though, as much is generated on the fly; the REX browser plugin, then, intercepts the chatbot interface’s own (undocumented) API requests, and makes its own requests of the API to gather the relevant interaction data. Importantly, this captures the hidden tool steps, external Web calls, timestamps, conversation tree data, and other aspects which are not directly exposed to the browser user.
There are further plans for this tool to engage in data minimisation, too: redacting certain details about conversation timings and titles, for instance, allowing manual redaction by the user, or also manual coding of conversations. The system is already available as open-source for use on researchers’ own servers, and as a hosted option.
