Thursday, May 26, 2016

The New Demos Study On Twitter Misogyny. A Procedural Fail.



Demos in the UK tells us that it has carried out a new study on Twitter misogyny (the hatred of women and girls).  Newsweek headlines its summary of the study findings like this: 

Half of Misogyny on Twitter Comes From Women


Which is interesting.  Another summary tells us a little bit more of the study, such as the words it picked for measuring misogyny: "whores" and "sluts".   That "cunt" was left out is odd.  I would have thought that it would be the very choicest of words to measure woman-hatred.

The  study found 1.5 million tweets with either one or both of the search words "slut" and "whore" during a three-week period.  That sample was then winnowed down by using a Natural Language Processing Algorithm to remove from the original sample all pornographic marketing tweets (which were 54% of all tweets with those words!)*, and all other tweets which do not express anger in the use of those terms.  The remaining 10,000 aggressive tweets are the focus of future analysis, such as the sex of the tweeter.

Now, how did Demos decide if a tweet was sent by a man or a woman?  In my experience few people (at least in politics) tweet under their own name or any normal type of name.  Most people have Twitter handles, and many of those are very hard to attribute to either men or women.  Pictures?  I know several people on Twitter who have the picture of some famous man or woman whose apparent gender is not the same as the tweeter's gender.  And I, for instance, have a picture of embroidered snakes on my Twitter page.

After thinking all that, I was eager to learn how all that was done in the study.  But guess what?  There is no written study anywhere on the wide and varied Internets!

I then e-mailed Demos to ask for the url of the study.  The rapid response I received (thanks, Demos!) told me --  and here comes the fun part! -- that there IS NO WRITTEN REPORT THAT PEOPLE COULD ANALYZE.

That is bullshit.  Absolute bullshit.  Just think about what that approach means for scientific inquiry:  Nobody, but nobody, can tell if you have done the research correctly and nobody, but nobody can post a criticism if your study is a bouquet of nose hairs with dried snot on them.
-----
* How can we be sure that these tweets aren't based on misogyny?  Do corporations get an automatic pass?

Added later:  That there is no report does not imply that the results are incorrect, only that we cannot tell if they are correct or incorrect.  But a written report is very important.  The reason that researchers write their studies up is so that others can see what they did, how they did it, and also so that others can judge whether the study was done right or not.