Tracker update: whatcha doing, Google?
Well, I did it. I finally switched completely to a “dark mode” dashboard. I sort of had to, after downloading this absolutely gorgeous RStudio theme. Regardless, there’s something about seeing things slightly differently that affords you different perspective and I’m starting to get the impression that there is something truly, terrifyingly wrong with the way Google serves and curates news items in the index.
When I search for “google news content guidelines” I get the official Google News content policy where they claim that Google will not provide content for hate sites or sites that peddle disinformation (written here as misinformation). But is that really true?
During the 1 month time period of the scraper applet running, among top search results the Daily Wire was served 19x times, sibling publication the Daily Signal 29x, and The Blaze 17x. Tabloid Daily Caller was served 55x.
If somehow the Daily Wire has escaped your attention (honestly I’m so proud of you!) they recently made international headlines when one of their flagship staff Michael Knowles called for the “eradication of transgenderism.” Matt Walsh has gone on to clarify that this is exactly what they endorse, and that they will not be satisfied until “gender ideology” is abolished. The Daily Wire has previously called for the execution of teachers and doctors. So much for that content policy.
The scraper continued to serve up Daily Wire as a top result even after the CPAC comments. Including an article where they mocked trans people for being outraged about comments regarding our eradication. Classy, Google.
Google News result for “transgenderism”
In light Knowles and Walsh’s recent outbursts, a quick search for “transgenderism” shows that when it comes to trans people, anything still goes.
So much for news: by contrast, MSNBC, one of the largest liberal news outlets in the entire US was only served 8 times. That’s just 1 more headline than Christianity Daily, which was served 7 times during this period. Focus on the Family’s blog Daily Citizen was served in top results 12x, 50% more than MSNBC. What’s happening?
Google seems to have no uniform definition of what constitutes a “news item”. It would seem that Google’s definition of “news” entirely depends on the type of platform serving the content, not the actual authorship or organizational affiliation. A quick look through what Google News returns when I search for “transgender” reveals a disproportionate number of articles focused on items that are not news at all, but think pieces where cisgender people come up with bizarre theories about the lives of trans people, and how to “solve the problem” of trans people’s access to public life.
Even more worrisome, are the number of items from hard right-wing outlets like the National Review, and a number of non-news hyper-conservative organizations like the Catholic Medical Association, or Parents Defending Education. For the entire month of February, what did not appear in the dataset, while still making sure every PDE and CMA “article” was front and center were long-respected left-leaning outlets like Media Matters for America or Democracy Now.
Search results for keywords related to trans people tend to favor salacious headlines like “Pope Francis: ‘Gender ideology’ is one of ‘most dangerous ideological colonizations’.”
As an aside: it’s little rich for the Catholic Church to be talking about “colonization”.
But aside from the pot-stirring headlines that generally support an anti-trans bias in the average reader, there are much more extreme headlines in top news results that are driving much of the moral panic. For this, we’re going to try an experiment.
Content warning :
Beyond this point is description of search results related to sexual violence.
Within the top results of the dataset are articles with eye-popping, nauseating sensationalist headlines like:
“Violent pedophile who raped three-month-old baby as a teenager is quietly moved to female prison with mother-and-baby unit after having transgender surgery in Canada” This is obviously accelerated if we search for “trans prison”. The same article still shows up on the front page, between position 3 and 5.
So what if I just type “trans women”? This time, that specific Daily Mail article doesn’t show up, but two other Daily Mail articles, also about Isla Bryson, do show up within a -1.5/+1.5 margin of the same search result, this time is positions 3 and 7. On headline reads: “Trans woman found guilty of rape moved to men’s prison”.
While I understand the desire to cover Bryson’s story, there are have been more than 100 articles on her in the past month, and the story in question is fromover a month ago and many of the news stories returned are from weeks ago despite the fact that Google News is engineered to prioritize new stories with the most recent timestamps. More to the point, contrary to the suggestions of the headlines, Byrson was still clearly identifying as male at the time of her crimes. She was using the name on her birth certificate, presenting as she had always presented and had never told people around her that she was trans.
Surely, Google, more has happened surrounding transgender women and girls since early February, so is there a reason this is a top story?
What else is there on that first page for “trans women”? Another of the first hits is an article entitled “My son transitioned into my daughter. I love her, but I'm still mourning the loss of my son.” We’re being hit with the biggest assault on trans lives in US history, but let’s dedicate the 4th highest search result in “news” about how the real victims in all of this are the family members who have to put up with us? Two more results are about our “intrusion” into women’s sports and bathrooms.
Another result includes one eye-catching headline “Uterus transplants are already a reality. What does it mean for transgender women getting pregnant?” While if this were true, it would be a really large news item, the fact remains that while almost certainly medically possible under the right conditions, to date there have been no successful uterine transplants for transgender women, and only a handful of successful uterine transplants for cisgender women. Aside from a handful of physiological issues that need to be sorted out, unfortunately, the anti-rejection medication cocktail required for tissue donation from something like a uterus transplant makes it extremely impractical.
And while I have to address it may be muddying some of the results by using the same OAuth token for all of these results in my dataset, it still shows that from a user experience perspective, overwhelmingly Google is reproducing the extremely negative valence of the current discourse on trans lives.
The reality is that not only are google’s responses to those keywords worrying, but a simple natural language processing of the shows that the topics of Google’s News Index vastly skew towards higher contention topics where the sentiment of the average reader is more likely to harbor anti-trans bias.
If you are structuring your machine learning algorithm so as to try and give a balanced account of a given topic without negatively influencing public sentiment, you wouldn’t be seeing sports take up the top subcategory in the news index, a topic which effects the least amount of people, cis and trans, possibly out of almost any given keyword. And the tone and tenor of these articles is extreme, such as “Female swimmer denied trophy in favour of transgender male”. Western Standard is a far-right blog with anti-vaccine and COVID conspiracy articles, as well as a surprising number of articles about the dangers of trans people. But the headline should be a clue.. “…in favor of a transgender male”? Not to appear pedantic, but the choice to use the wrong term (in this case should say transgender woman) is a deliberate one.
And while it may be a topic the fewest number of people actually care about, what that topic does do is highlight the talking points that groups like the Alliance Defending Freedom have been planning to focus on for years as a way of driving public sentiment against trans people — as shown in recent leaked emails.
There are ultimately two possibilities. Either the index has no breaks on it, no safeguards at all (which seems most likely), or Google’s claim to have engineers working on solving this problem turn out to be a bit of a fib, like when it turned out that Google’s search index hate speech detector was dripping in anti-Black bias?
The more I work on this project (which really, truly was supposed to be a day’s distraction), the more clear it is that anti-trans biases built into news aggregation and distribution platforms — namely, the single most used news distribution platform in the world — are amplifying the already raging anti-trans sentiment tumbling out of control at an alarming rate.
Update: after sleeping on it, I decided to test a theory.
You’ll recall from my original post detailing the experiment that I had begun by creating three sets of RSS feeds, each with a google alert tailored to region and a keyword chosen for its perceived valence.
Keywords:
“transgender” (positive valence)
“gender identity” (neutral valence)
“biological sex” (negative valence)
Now before you fill my contact form with hate mail, the reason I chose “biological sex” as a negatively-valenced keyword is not because I don’t believe that biological sex exists. Hold your horses, Joanne.
I chose it because part of the rise in anti-trans sentiment corresponded in the UK with an increased emphasis on that phrase. News outlets sympathetic to trans people tended to use transgender, since that’s the correct term. Neutral outlets, (or those that are casually bored by us, but without malice) tended to use “gender identity” as a fence-sitting choice. But “biological sex is real” has become a rallying cry for everyone who wants to prevent people like me from using the bathroom or filling a hormone prescription.
And while there seems to be some semantic drift — namely that the US is now using the term more, with the UK favoring the first — the practice of using “biological sex” as a dog whistle has been imported to the US. Anyone who has read the current bills knows this to be true.
So how accurate was my hypothesis? Depressingly so.
What does this mean? It turns out detecting anti-trans valence is actually not all that hard. Google would just have to start giving a damn, first.
