That “Influencers” Preprint

“Influencers” and their lucky socks — why correlation is not causation

If you’re seeing the image with this post, the propensity of US flags has some bearing.

A number of readers forwarded me a preprint on arXiv covered by IEEE Spectrum under the headline “A Few Social Media Influencers Are Shaping AI.”

In the IEEE Spectrum article, the preprint’s authors praise the “influencers” in a rather sycophantic manner, and there is more than a little boasting by the “influencers.” The headline of the article itself is far too definitive given that the writer is covering an unreviewed preprint with what I feel are some obvious problems.

Nobody seems to have critically scrutinized this preprint or asked hard questions — but that’s par for the course in the exciting new world of “say whatever you want as long as you plop an abstract over it” science publishing . . . sorry, press releases disguised as preprints.

  • The authors of the preprint relied on a service run by one of the “influencers,” lending a slightly different feel to the term.

But a look at the data shows the “influencers’” trick, and it’s rather simple and obvious — over-index research from the US, mostly ignore the stuff from the EU and China, and you’re suddenly a genius.