Learning from Randomized Interventions in Social Media

How should we reason about the effects of interventions in social media? What effect sizes should be expected from such changes to algorithms and content? And, given the fundamentally social nature of these services, what conclusions can we draw from individual-level experiments? I consider two cases: effects of changes to algorithms and affordances on political polarization; and effects of digital ads on propagating low-quality content. For the former, I will comment on the published results from prominent, recent experiments on Facebook and Instagram conducted during the 2020 US Elections. For the latter, I will share two new field experiments on Facebook (N>33M) and Twitter (N>75k), each randomizing exposure to advertising featuring content-general messages reminding people to think about accuracy.