Platform data accessed through the DSA: The foundation for a new science of AI impacts

Platform data accessed through the DSA: The foundation for a new science of AI impacts
Ali Knott, Centre for Data Science and AI, Victoria University of Wellington 
 

Abstract
We urgently need to understand the impacts of AI and related technology on our world. In this talk I'll begin by arguing that the best scientific methods for studying these impacts all require access to tech platforms of one kind or another. This access might be to data or code held privately by platforms. Or it might be to tools that allow experimental interventions in platforms, comparing the effects of different AI systems on individuals and communities. 
The EU's Digital Services Act permits external researchers to gain the necessary kinds of access. Article 40(4) of the DSA gives vetted researchers unprecedented powers of access, to study 'systemic risks' of the largest tech platforms, and the mitigation of these risks. I'll argue that this little clause is potentially a huge deal: it could in principle serve as the foundation for a brand new public science of tech impacts, similar in scope to macroeconomics. But to realise this possibility, the research community must organise itself. I'll describe a new organisation of researchers which we set up to support this work: the Social Data Science Alliance. Our first project was to map out the structure of the new science enabled by DSA 40(4), by asking researchers which questions they would ask with DSA access, and organising their responses into a taxonomy. Our second project, currently under way, is to suggest ways to streamline DSA access, by identifying the platform datasets which enable the broadest range of questions to be addressed.  I’ll describe the results of our first project, and ask for your help in defining the second project.