Hate in a post-information age
A week ago, I posted an article about a dataset that had been compiling over the course of a year based on a series of RSS feeds I’d set up and mostly forgotten about. Most of all, what I can’t get out of my mind is the sheer volume of articles.
At first, I thought it was an artifact of the data collection process. Setting the alerts to any news article using the “as-it-happens” frequency setting means that the second Google indexes a news story, it gets pushed to the feed, and that is understandably a lot given the sheer size of the internet. While I have no doubt Google would dispute this, the evidence appears to suggest that pretty much any website can be— given enough traffic and/or paid advertising— listed as a news story. But that’s only a tiny portion of the story. While large language models and generative AI is almost certainly replicating and even fabricating news stories as clickbait while artificially driving up trends, to me the problem remains when you look at individual outlets. Let’s start by looking at the ones that we can all agree, begrudgingly, are “news” of a sort, in the same way that a hot dog is a sandwich.
An overview of the numbers
When restricting the dataset to only a handful of the top conservative news sites in the United States, those 6 sites still account for a minimum of 200 articles every month, with 320 in June alone. That month Fox ran 92 articles, compared to 34 from CNN. Only 4 were reported from MSNBC.
In just 9 months, the data capture system recorded 630 unique articles from Fox News, that’s roughly twice a day every day for the entire year. Between just those 6 sites, that’s 2,169 articles in 9 months. It’s important to note, by the way, that this is still only a fraction of the news articles posted by these sites. Across more keywords, using the coded language of “parental rights”, we would likely find a whole range of other articles.
Here’s where it gets wild. However, since the RSS feed is only based on what Google served as news alerts, we can also manually check our work on Google using advanced search operators. Below you will see 75 results for MSNBC, 147 for CNN and a whopping 690 for Fox using identical parameters. The true number is likely somewhere in the middle: Fox frequently repackages the same story, even using the same end URL, with a different headline and image to enhance SEO performance. The numbers provided in the plots on this page are filtered to ensure that these are unique URLs regardless of headline. This means that regardless of the true number of stories run, we still are seeing multiple times as many stories between liberal and conservative outlets. A manual search of The New York Times, which engages in similar SEO shenanigans and is notorious for running anti-trans misinformation, returns 340 results. That’s a ridiculous number of news stories, but it’s still a fraction of what we see from Fox.
Nevertheless, even using only the numbers captured by the EDC, the volume is still quite shocking. Using only my dataset, sites like Fox News still on average published greater than one new article on trans people every single day for the entirety of the data collection period. This amounted to a staggering number of news stories, and as is visible below, these were not restricted to search terms that are more inflammatory in nature such as “transgenderism”, made popular by people like Michael Knowles.
Given that trans people are at most 1% of the population, the volume is frankly bizarre. As mentioned in my previous post, research suggests that people who hate groups of people tend to overestimate the size of the groups they hate. For a long time, white supremacist groups have pushed “great replacement” theory, the idea that over time, as the census in the United States grows more ethnically and racially diverse, white people will become the numerical minority in the country.
Great replacement and adjacent ideas may also explain why despite evidence to the contrary, many people in the United States believed, despite no small amount of evidence to the contrary, that major cities which had seen huge demonstrations following the police murders of unarmed Black men like George Floyd had in fact (according to many people online) burned completely to the ground as pointed out by a Bluesky user. That this was not at all true appeared to have virtually no effect on the proliferation of this belief. The sheer number of news stories, with little to no content curation on the part of news hubs like Google News or MSN serves to convince people past the point of reason.
Trans panic & discursive drift
When I originally set up the RSS feeds, I chose the keywords based on the assumption that there would be clear differences in the type of results produced by each of the three primary keywords. I also assumed there would be changes in how those keywords performed, but I imagined it would take at least a few years before that shift was visible. While the data appears at this point to support the idea that there are changes over time in the types of results produced by various keywords, there appear also to be changes in the number of results per keyword even on a single platform over a much shorter span than anticipated.
One of the primary driving factors for this project had been reflecting about the level of discursive drift that had occurred in a short period of time around the words we use to refer to trans people. Discursive drift occurs when the understanding of a word is altered by its contextual use, such as when a word or phrase comes to denote its opposite. The term was popularized by Professor Deborah Cameron, who examined the inversion of the meaning of “political correctness” in popular language in 1995 from something generally seen as positive and progressive to something censorious. While she writes prolifically on the subject of gender, Cameron herself has been careful about defining her views on transgender people specifically, although she is often claimed by people like Helen Joyce. Regardless of Cameron’s specific views on transgender rights, discursive drift is an important contribution to the fields of linguistics and anthropology, as it offers us a lens to understand that not only do words change in their meanings over time, but this change in meaning also recursively shapes the discourse itself.
It is no secret that over the past ten years, a concerted effort has been made to join two previously mutually exclusive groups: trans-exclusionary feminists and the far-right. This merger was most visible when the Women’s Liberation Front (WoLF) began working in earnest with the Heritage Foundation. Since then, WoLF has submitted amicus briefs in many major court cases and no shortage of legislative efforts seeking to curtail transgender rights— sometimes not only against transgender rights but against those of the broader LGBTQ community as well.
This has created new and interesting contradictions in anti-trans campaigns, particularly around the phrase “biological sex”. Trans-exclusionary feminist groups favor the term as form of dialectical materialism, in which the material reality of sexual dimorphism has led historically to the subjugation of women as a reproductive class [1]. Conversely, conservatives have come to favor the term as evidence that there is a biological basis for traditional gender roles rooted in scripture. This odd-couple merger has occurred against a rapidly fluctuating backdrop of social and political forces, including among the more commercialized branches of the social sciences. Deb Cameron herself refers to this as “the new biologism”.
Fox News
Daily Wire
And while discursive drift is academically interesting, it has very real consequences when applied to the manufacturing of consent. While according to massive leaks from extremist, anti-LBGTQ group ACPeds, the group alongside the Heritage Foundation and Family Research Council have worked hard to promote the term “gender ideology”, something which can be seen above as a favorite of the Daily Wire. Religiously-affiliated groups appear to also prefer “gender confusion”, as it carries the inherent implication that being trans is a mental illness.
Anti-trans groups like Genspect have worked hard meanwhile to convince the public to shift away from the word transgender. And for much of the time, outlets like Daily Signal have followed suit using cumbersome phrases like “a man who thinks he’s a woman”. But the data suggests this trend may be shifting. As these plots show with articles matching the search term “transgender”, towards the end of 2023, that word appeared to be making a comeback. And while it might be tempting to see this as progress, I see it as reflective of the fact that the context around the term has become so fraught that it has lot its neutral valence. While groups like SEGM and Genspect will likely continue to shy away from the term as they need to at least casually feign seeing transgender people as human, news outlets like the Washington Examiner can use it in an overtly anti-LGBTQ+ context without financial consequence.
As is visible across the six news outlets displayed here, there are sharp changes in which keywords bring users to their sites. April 2023 starts with a staggering number of articles, 110 from Fox News in that month alone, with nearly half appearing under the search term “transgender”. This has the added effect of meaning that people who want neutral sources, or who are merely seeking information about the topic have no choice but see anti-trans outlets listed, if not have them served as the only options.
Daily Signal
Breitbart
Examiner
About that volume though
So how are there so many news stories? While there’s always been a small market for outrage pieces from professional pearl-clutchers every pride month, this past June was somehow more over the top than any before it. While the year had already been packed with escalating violence, the Pride month boycotts started two months earlier than usual with the Bud Light boycott.
The month culminated in what has increasingly become a staple of anti-LGBTQ protest, with bomb threats at the Anheuser Busch factory and videos of mid-90’s has-been musician Kid Rock firing round after round of automatic gunfire into several cases of the beer. After an unending stream of death threats, Dylan Mulvaney was forced into hiding for her own protection as the entire far-right appeared to lose its collective mind.
Meanwhile, a wave of similar threats and a concerted campaign from, among others, the American Family Association, prompted the store to remove merchandise just before June actually began. Bomb threats also followed, although exactly who sent them remains unclear. Just prior to the decision, several videos emerged of far-right livestreamers throwing tantrums in the stores, often knocking over merchandise and harassing workers. Among the companies targeted was Oreo.
Oreo has long engaged in Pride packaging and merch, as far back as least as 2012. Every year they create pride packaging, and every year there is an outcry of sorts. But like most outcries, they start during or immediately after Pride month. When on February 7th, I opened my browser and typed “trans” into Google News this hatebait article from Heritage front group Daily Signal popped up within the first page or two of results. At first I naively assumed Oreo had done something new. But the packaging of offense cited specifically in the Daily Signal’s article isn’t only from this year’s Pride, it’s from Pride 2022. The article was instead a press release about a video from the National Legal and Policy Center (NLPC) that would be released within a couple of days of the news story on right-wing YouTube clone Rumble announcing a targeted pressure campaign against the brand. But that same time, the story was replicated almost verbatim across dozen or so sites, including the Christian Post and the Daily Wire. [2]
Surely of all the things going on, there must have been other things their partner websites could get angry about, but replicating content and marginally editing with ChatGPT allows fake news organizations to rapidly churn out content [3], thus dominating the news indices so much that even Google, which attempts to customize its results to the user, served the story to me on multiple profiles across multiple machines. It would appear that there is not only no protection against saturation in Google News’ algorithm, but that it might in fact thrive on it.
Footnotes:
The language of “reproductive class” is typically specific to the intersection of Marxist and Feminist theory, especially within the extraordinary canon of Black Feminist theory. I imagine readers might bristle at this language, however dialectical materialism despite its faults is still an essential tool to understand the world. A more robust materialist analysis would also consider the social and racialized forces that shaped our current understanding of gender as being far more complex than the reductivist drivel pushed by many right-wing groups cloaking themselves in the thin costume of feminism.
The NLPC is not only anti-LGBTQ, but has also taken aims at corporations for DEI and BLM efforts. NLPC is also strongly anti-immigrant.
As journalist Mia Sato shares in her brand-new must-read article in The Verge “When a Funeral is Clickbait”, this phenomenon is not restrict to hate speech. Instead, Sato describes how sites run by generative models have created fake obituaries to generate advertising dollars. As a sidenote, many of those advertising dollars, incidentally, going to Google affiliated ad networks.
github.com/jessicakay/gayagenda
A couple of data housekeeping points:
You cannot separate data from the methodology used to collect it. Currently it appears that the time-based google search operators are working unreliably, so it’s difficult to ascertain exactly how much of the above data and trends generalize to the entire news ecosystem.
As you can see in the plots in my previous post, one of the feeds had disconnected over the course of the year leading to an incomplete dataset for the keyword “gender identity” set to UK regional settings. The data remains in the dataset, but anyone doing regional analysis should exclude that keyword/region combination.
