The Google “Study” that Crashed the Copyright Office

Ellen Seidler spotted a Google funded “study” from the ever-pliable professoriate entitled “Notice and Takedown in Everyday Practice“.  This is the typical kind of academic astroturfing we’ve spotted from Google in the past.  This particular study happened to be released a couple days before the deadline for comments on the notice and takedown process were due at the Copyright Office.

The study completely ignores two fundamental facts:  First, it focuses solely on notices sent by “major entertainment industry” corporations and never considers SMEs or individual professional artists who also send notices.  Second, the reason Google receives so many notices from organizations like BPI is because that’s who Google gives the tools to automate the process.  So it’s a fallacy of composition at the heart of the study.  In fact, the study pretty much acknowledges that the dataset American Assembly used was flawed because it was 99.4% based on Google search.  (This would make sense because Google financed the “study”.)  The study disclosed this flaw (at pp. 78-79):

The data is also significantly skewed by the fact that one entity, Google, received 99.4% of the notices in our set [from “Lumen” which is the renamed “Chilling Effects” site that Google financed, so no wonder]. Twitter was the next most prominent recipient, with only .3% of notices. This effect is further pronounced when looking at individual takedown requests rather than notice count: fewer than .008% of the requests in the dataset were directed to entities other than Google. Further, although Google reports to Lumen [the company who collected the dataset used by the “study”] on a wide range of its services, it does not include all.  Most notably, it excludes takedown requests for its YouTube service.

The dominance of Google notices in our dataset limits our ability to draw broader conclusions about the notice ecosystem. Google has characteristics that set it apart from many other services, including the use of form notices, automated triage systems, and a “trusted sender” program. Google’s dominant position in search and the extraordinary number of notices it receives also make it unusual. This makes the Lumen dataset [the one used by the American Assembly] useful for studying an important part of the takedown system, but also means that the characteristics of these notices cannot be extrapolated to the entire world of notice sending.

Of course, the “academic study” continues the revisionist history about the reliability of the takedowns.  According to Politico’s reporting:

Google said Friday it has received more than 4.3 million copyright removal requests in the past month — about 97 percent of which are valid. Many of the domains that are targets of the most requests are file-sharing and torrent sites. (emphasis mine)

Ellen’s post is a must read, particularly because the argumentation is clearly replicated by the comment bots that crashed the U.S. Government servers.  Sounds like a job for the U.S. Digital Service in the Executive Office of President Obama–oh, sorry, the Administrator of the U.S. Digital Service used to work for Google.  Forget that idea.

Read Ellen’s post here on her excellent VoxIndie blog:

Google Funded Report on Copyright Takedowns Drops The Ball

Google-funded report generates desired headlines and conveniently downplays the role of DMCA counter-notices–ignoring fact the system is weighted against rights holders.