Thread by Emilio M. Bruna

twitter.com
1 Recommender
1 Mention
Recommend
Post
Save
Complete
Collect
Thread
You may have seen Rufo gleefully reposting a report from an organization called the @NASorg (I'm sure the name's similarity with the other NAS - National Academy of Sciences - is purely coincidental, wink wink) It's on the "takeover of the academy by DEI"

(stay with me here..)
@NASorg They sell the report as a bunch of quants letting the numbers and data speak for themselves. One of the results that caught my eye was that university twitter feeds had had a *huge* increase in race-related tweets... look at those huge increases...the lines go up up up!
A few things caught my eye...for starters, they don't present any of this as a percentage of the overall tweets...or even graph the total tweets over time. Twitter has become much more important, so my guess is that this is a miniscule fraction of the overall twitter feeds.
Of course the other reason it looks so impressive is because they let the y axis float for each graph. So it looks like everything is going up massively, even if relatively speaking most panels are a small number of tweets.
To their credit - and I mean this sincerely - they actually do something very unusual. They posted all their code on github and the data on Zenodo. This kind of transparency is very unusual. So I went and grabbed the repo and data, and...
go figure...it's hard to get lathered up over a bunch of (mostly) flat lines. The big spike in "racism" after GF's murder is why the bug driver.

Now coincidentally, I'm doing some kw analyses myself, and I thought to myself...
"I wonder how good their triage was? Because some of their search terms were (imo) way to broad...."

People, I can't stop laughing...
their "tweets_clean" csv has 151284 tweets. I figured I'd start with the term that serves as the foundation for this whole report - RACE - and sure enough....
After only 100 tweets, 16% of the tweets about "Race" are about....

a nascar...race
robots and AI taking over the human...race
the Heisman Trophy...race
the boston marathon road...race.
the great food truck...race
swimmers finishing second in their...race.

hahahhahhahaha.
If you really want links to the report and repos I'll share here. But only if you want examples for class of

'quantitative' != 'rigorous'

Enjoy your weekend!
Aaaaand typo. This should obviously say “after…is the big driver”.
Another hilarious "we quantified DEI taking over campus" fail:

"Diversity" tweets about ecological research on SPECIES diversity...


Promised I would put exactly 60 minutes into reviewing the results of their tweets by searching for what should be obvious potential false positives that should be eliminated from their "clean_tweets.csv" dataset.

Buckle up:
but before going further, I literally set a time for 60 minutes. This is the first pass...low-hanging fruit, easy and obvious stuff. within the tweets for each of 4 terms - Race, Diversity, Equity, and Justice - I searched for tweets with / without combinations of terms.
for instance, for all tweets with the word "Diversity", I then filtered the tweets about "species diversity" or "biodiversity" so long as they also *didn't* have words like "social justice" or "racism" and sure enough...

170 of those (about 0.54% of diversity tweets)
"Equity" may be my favorite fail.

Turns out 270 of those equity-related tweets were about...wait for it..."Private Equity". As in "investors and hedge funds."

that's a 1.64% of the "equity" tweets. I'm sure there are more errors, but I was laughing so hard I quit looking
All those tweets about "Justice" that prove DEI has run amok on campus? Turns out 10.5% of them were about...

Justice Sandra Day O'Conner. Or Justice RBG. Or that well-known SJW Antonin Scalia. Or the School of Criminal Justice increasing in the national rankings.
(again - it's not the final error rate, just a first pass based on quick scan of their 'clean' dataset. There's nothing really sophisticated here; with more time I would have tried some co-associations of words within tweets to look for other potential errors)
So that brings up back to what started it all - tweets about RACE. This is their big concern - that DEI efforts have permeated and corrupted the academy. So you'd think that they'd be especially careful about analyses related to this fundamental issue/term.

You would be mistaken
fully 20.5% of the 25,167 race-related tweets were actually about:

the space race
the human race
the indy 500 race
the race for the acc golf title
olympic races
robot races
the amazing race

I'm sure there are more. I just quit looking because it was too embarrassing.
the snippet of code used to do this is at my fork of their repo github.com/embruna/quantdei_nas, but you'll need the data from zenodo to use it. I'm, happy to share if anyone wants to keep going.

Not me though, because tomorrow I'm going for an early ride.
not a race.
just a ride.
This is the worst hobby ever....

"Advocacy": another term whose increase @NASorg attributes to DEI's takeover of STEM & Universities...

...Except that 10.2% of "advocacy" tweets are actually about law school mock trial teams, veterans advocates, and meetings with legislators.
...Also "patient advocates" in hospitals.
(accidentally broke the thread, re-linking)

Mentions
See All