Thread by @DavMicRot

.@SocSciOne is incredibly exciting & important initiative to package & make available to qualified researchers secure & privacy focused data from private companies like @Facebook, so researchers can explore what is being produced, distributed, consumed on these platforms 1/8

In this paper misinforeview.hks.harvard.edu/article/resear… w/ @_JenAllen @duncanjwatts @markusmobius we explore its 1st major data release, testing how key outcomes researchers want to test (eg fake news) change due to introduction of differential privacy & censoring to URLs w/ 100+ public shares 2/8

We find that censoring the data to 100+ public shares overestimates relative share of news clicked on Facebook by nearly 2x, fake news by 4x. Fake news URL, conditional on being consumed, are engineered to be shared and shared publicly, at much higher rates than other content 3/8

Thus, a naïve reading of this data would lead researchers to a large overestimation of the size of the fake news problem on Facebook. We are able to address this by comparing this incredibly large data, with a much smaller representative sample of engagement 4/8

This research speaks to public debate over what journalists are seeing @crowdtangle which is also built on (even more stringent definition of) public shares: fake news & right-wing clickbait way overrepresented in this data versus what is actually being consumed on Facebook 5/8

We urge future data releases to focus more on smaller representative data and/or stemming URLs to avoid the need for high-threshold censoring of the data to protect privacy. More details in paper misinforeview.hks.harvard.edu/article/resear… 6/8

Important to note that @Facebook is releasing two different peaks into what is being consumed on its platform (w/ @crowdtangle & @SocSciOne) and both give a distorted view that make Facebook look much more filled with fake news & right-wing clickbait than it really is 7/8

It would be easy for Facebook to release data securely w/ proper privacy constraints that would give more realistic view of consumption of news on Facebook. Just perplexing why they don't produce more representative data of their platform: obvious win-win for them & research 8/8

@threadreaderapp unroll