Brookings: “…The system of APIs set up by Twitter over the past decade—most of which were set up for business purposes as opposed to research—led to a flowering of academic research using Twitter data. And to be very clear, Twitter deserves kudos for making so much data available, including specialized collections around Russian Internet Research Agency (IRA) trolls and COVID-19. However, even the data that Twitter makes accessible to outside researchers has left out information that is crucial for academic research, such as data about which users have seen which tweets, known as “exposure data,” or even by how many users have seen each tweet. Therefore, understanding which tweets and associated news items reach which classes of people remains an area of inquiry that outsiders to Twitter cannot investigate. Moreover, some data, such as friends and follower networks, have become harder to collect at scale over time as new API access rules are rolled out with new rate limits governing how often researchers can access the API. Nevertheless, because more research has been performed on Twitter than any other platform, our understanding of the relationship of social media to online harms is highly biased toward what is occurring on that particular platform…To break through the logjam, we need federal legislation.  That legislation, such as the Platform Transparency and Accountability Act proposed by one of the authors (Persily), could come in many forms but it should have three essential characteristics.  First, a federal agency must be empowered to force the large internet platforms to share data akin to what firm insiders are able to access with outsiders not selected by the firm. Second, that agency (perhaps working with a nongovernmental organization or another arm of the federal government, such as the National Science Foundation) should vet researchers and research projects that will be given access to platform data Third, data should reside at the firm, and regulations should specify in detail the process for accessing data and publishing results in a way that would not endanger user privacy…”