Studying the Quotidian Web
There are about 20 billion videos on YouTube, with a median view count of just 41. 800 million have never been seen by even one person. For all the press and politics about Internet platforms, basic statistics like these are hard to come by. We typically don’t know how large platforms are, what languages their content is in, and what regular people use them for because – especially in the “post-API age” – most of what we do know is filtered through opaque, attention-optimized recommendation systems. But it’s a project worth doing, not just for the sake of transparency and auditing, but because they are also rich, global repositories of everyday life and culture that are deprioritized in favor of MrBeast.
This presentation will explain the research program that has been built at the University of Massachusetts Amherst dedicated to the production and study of representative samples of social video sites. More important than what’s already been done is what’s still left to do. The purpose of this talk is to get people excited about solving the technical challenges associated with random sampling. YouTube and TikTok have been figured out, but, absent meaningful legislation to mandate platform transparency in the public interest, there are big open questions about most sites, especially smaller platforms and those that are less popular here in the U.S.
Saturday 2000 Crystal Ballroom