Abstract
Article 40.4 of the Digital Services Act allows ‘vetted researchers’ to access data and algorithms within ‘Very Large Online Platforms’ (VLOPs) and ‘Very Large Online Search Engines’ (VLOSEs), for the purposes of studying ‘systemic risks’ and their mitigation. The Delegated Act setting out the terms for this access has recently been published (see here). But the question of what studies should be conducted is still very much open.
The Social Data Science Alliance (SDSA) was created earlier this year, to bring together experts studying digital platforms, to discuss, and help define, the scientific research programme that will be conducted under the DSA’s new access provisions. Our aim is that the new provisions defined by Article 40.4 enable a structured programme of research, rather than a piecemeal collection of individual research efforts. Defining a research programme requires coordination amongst researchers. There are already several venues for this coordination, including the DSA Observatory at the Universiteit van Amsterdam, the DSA40 Collaboratory at the Weizenbaum-Institut in Berlin, and the European Centre for Algorithmic Transparency (ECAT) in Madrid. The SDSA’s focus, alongside these initiatives, is to help establish the broad outlines of the new area of science that will be founded on the access methods created by the DSA. We believe the DSA can bring into existence a new public science of tech platform impacts and governance, that allows for key research questions to be investigated with a variety of methods, without infringing user privacy or company IP.
Our first project is to survey SDSA members about what research questions should be asked, using the new access provided by Article 40.4. We received over 250 questions, contributed by 18 members (a mixture of individuals and groups). This report summarises these responses.
To structure our summary, we define a taxonomy of possible research questions, and position respondents’ questions at points in this taxonomy, to show where respondents’ attention is focussed. We explain the high-level structure of the taxonomy in Section 1, and introduce some further structure in Section 2, along with some initial respondent questions. In Sections 3–7, we summarise respondents’ questions for different platform types: social media platforms, retail/commercial platforms, mapping / geographical information platforms, search platforms and adult content platforms. In Section 8, we consider uses of generative AI methods in all of the above platforms.
Sections 9 and 10 cover responses in two categories that are adjacent to our taxonomy. In Section 8, we summarise suggestions about methods by which analyses of different kinds could be carried out, without infringing user privacy. In Section 10, we summarise respondents’ thoughts about the importance of ‘cross-platform research’, and make some suggestions about how research of this kind can be conducted under Article 40.4 of the DSA.
In Section 11, we list and number all the member responses we received. Our summaries in earlier sections are indexed into this list.
We circulated a draft of this report to all SDSA members, and provided an opportunity for members to indicate the questions they found most important. In Section 12, we describe the results of this second consultation.