Cognitive Failure and "Misinformation" Research (Part 1)

In April, researchers reported that they had conducted a study of “misinformation sharing” by Twitter users.

From their abstract:

Analyzing Twitter suspensions shows that users’ sharing of links to misinformation sites was as predictive of being suspended as was the users’ political orientation...

Their central finding (stats removed — we’ll see them below):

"... we find that Republican users in our dataset shared news from domains that were on average rated as much more untrustworthy than Democratic users, based on either the fact-checker ratings or the politically-balanced layperson ratings.”

Their conclusion:

“In so far as conservatives share more misinformation, it is not possible to be maximally effective in combatting misinformation without preferentially taking action against conservatives. Given the widespread public demand for reducing misinformation online (e.g. Figure 1), some level of differential enforcement across party lines must be deemed acceptable if platforms are to keep misinformation in check.”

The lead researcher told a reporter:

"...platforms have been under a huge amount of public pressure to act on misinformation,” Rand said. “If they do that, then necessarily, they’re going to wind up sanctioning conservatives more, and it’s going to look like they’re biased against conservatives.”

Incredibly, none of the above variables exist in their study — misinformation, Republicans, or conservatives.

• Misinformation:

  • There’s no misinformation here. They didn’t identify or code any misinformation in users’ posts or linked articles.

  • Linking to a website is the misinformation act in the minds of the authors. Yet they didn’t assess any websites on their misinformation characteristics.

  • They chose the websites themselves. They didn’t take the sites users actually linked — the researchers chose a small number of sites in advance, arbitrarily, and looked only at links to those sites. They don’t disclose those sites, or users’ actual linking behavior. Think it through — if they chose the sites, we don’t have a study. (And guess what percentage of links their arbitrary set covered.)

  • Their “misinformation” measure: In an earlier survey, they had random people rate how much they trusted a domain on a 1-5 scale (e.g. conservativetribune.com). They used those ratings for this study. It’s an epistemic collapse to use opinions about domains as a measure of misinformation, far out of range of anything we could call science, or even reality.

  • They also asked if people were familiar with the domain (No/Yes). They kept the ratings from people who said No. (85% of the smaller outlets were unfamiliar to over 75% of the raters.) They don’t disclose this in their paper.

  • All smaller/unfamiliar outlets ended up at the floor of these pointless trust ratings, becoming “misinformation sites” for this study. All big legacy media outlets became non-misinformation sites. They don’t disclose this in their paper.

  • The authors rigged their set of smaller sites toward conservative outlets: 70% were conservative, 25% were leftist, one hard to characterize. They don’t disclose this in their paper.

• Republicans: They didn’t identify anyone’s political party. They only looked at a subset of users who had posted a particular Trump or Biden hashtag in October, 2020.

• Conservatives: They didn’t identify anyone’s political orientation.

If you think I must be exaggerating or there will be a catch, no — rigor is too important to me to play games. Everything is true as written, and as you’ll see, it gets somewhat worse. (Note also that all the big legacy media except for Fox News are leftist, and I’ll take various $1,000+ bets that they publish more misinformation than the smaller outlets (or Fox), as a percentage of headlines, claims in stories, etc., and are worse on various forms of bias we might decide to care about more than misinformation rates. Send me a bet proposal, especially if you have funding we can use. It’s long past time we found out.) Let’s throw some fun positivity into this madness: If 1) You’re skeptical at this moment, and 2) After you read the details you agree that I wasn’t exaggerating, didn’t make a mistake, etc. send me a joke or a haiku, a comment, whatever.

“Research” of this quality feels like the product of a crackpot one-man “think tank” subletting the back of a laundromat, yes? Baseline brain activity would rule out such recreations, to say nothing of intellect, integrity, scholarship, or god forbid, science.

One would think. This… artifact… comes from the Massachusetts Institute of Technology (MIT). The one in Cambridge. And Yale. This is MIT and Yale. This is who MIT and Yale are now. The lead author is David Rand at MIT’s business school (Sloan), formerly at Yale — a hyperpartisan biologist who does crackpot-layman-style “research” on misinformation as a tenured Marketing Professor (I’ve no idea; even biologist might be a reach — with academics, degrees and titles can be misleading, so you have to cut straight to their work and demonstrated ability and intelligence). He’s joined by two other partisans at MIT, activist Gordon Pennycook at Regina, and Tauhid Zaman, an MIT electrical engineering alum who is now a management professor at Yale’s business school (no idea).

The artifact is a working paper, not published or peer-reviewed by a journal. A working paper can be true and valid, but obviously this particular artifact isn’t true or valid, and by academia’s ethical norms no one should be touting a working paper. It’s been a working paper since April. In late March, Elon Musk polled Twitter users:

On April 14 it was widely reported that Musk was buying Twitter. Rand produced his “WP” on Twitter that very day.

Three days later, in a spasm of idiocratic irony, the artifact was gratefully reported by CNN’s Reliable Sources program. They billed it as coming from “professors at MIT and Yale”. Horrifying, painfully vacuous video here, littered with “Professor Rand says…” with images of Rand’s tweets. Figure 500,000+ viewers (a fairly low-rated show). Rand was duly proud of his maneuver:

Alright, this all needs to be undone, and these charlatans must be fired. MIT and Yale have profound ethical obligations here — this is, unequivocally, misinformation. Actually it’s “disinformation” in this new Orwellian framework, because it was intentional — it was a false and fraudulent rationalization for censorship presented under the banner of two of the most elite institutions in the world in the minds of the public (for now). The authors had nothing. They had nothing pertaining to anyone posting misinformation. They tried to marginalize — to justify censoring — 90+ million Americans because they don’t embrace the charlatans’ political ideology, defaming them as posting more misinformation than the charlatans’ own veracious tribe.

For decades academics have flooded our civilization with false claims about conservatives, seeking to marginalize and harm them, while avoiding and punishing critical evaluation of their own ideologies. Trying to censor them shreds sacred principles and norms of the open society, reason, inquiry, and the American experiment. Smug assumptions that leftist ideology is somehow grand, or even good, or superior to other perspectives are as irrelevant as they are likely to vaporize on contact with outside thinkers. Notions that leftists are more accurate in their beliefs, less prone to spread misinformation, are wholly without empirical support and collide with all the extreme recent examples of leftist misinformation, false conspiracy theories, and hoaxes. If this combination of malice and bovine “research” is what academia intends to spew into the world, we should pull the plug, re-approach the problems we’re trying to solve, and start over with solutions and scholars willing and able to move civilization forward, to learn, to understand, and maybe to know. If MIT and Yale don’t fix this and vigorously unroll the disinformation their charlatans spread, they shouldn’t receive a dime of taxpayer funding in the future, federal grants, etc. They are, in the parlance of the new leftist pathogen model of misinformation, disinformation superspreaders, not centers of inquiry and truth-seeking. Ethically, taxpayers cannot be expected to fund disinformation.

Academia is in a broad state of collapse, and MIT and Yale lead the spiral. MIT recently canceled a UChicago climate scientist’s lecture on climate science (the annual Carlson Lecture), because in the course of living his life he co-authored a critique of the new leftist “Diversity, Equity, and Inclusion” ideology, and its typical manifestation as enforced dogma. He offered an alternative values triplet: Merit, Fairness, and Equality. (I’d replace DEI with the volume of air the DEI enforcer drones displace, but if we must traffic in triplets, the Moulin Rouge values of Truth, Beauty, and Love would serve, or we could leverage the ancient dark art of mathematics to weigh relative standing — Truth⁷, Reason⁴, Rigor³, Courage², and Kate McKinnon².) MIT also hired six deans just to enforce leftist DEI ideology — an awkwardly dim cult ideology that assumes without arguments or evidence not only its own righteousness, but its supremacy and priority over all other cults, religions, value systems, ethical frameworks, political perspectives, triplets, and sadisms that humans have devised across time and place, saturated with malice and misinformation about invisible forces, the effects of “implicit bias”, and egregiously false claims that racism is the principal or a major cause of various social outcomes.

Yale’s scholarly environment is best captured here, and at least twelve federal judges now refuse to take law clerks from Yale Law School because of its disturbing cult-left intolerance of cognition and that it allows students to disrupt speakers such that they’re ultimately unable to speak. It probably didn’t help that Yale viciously retaliated against its own law students last year for not saying exactly what Yale wanted them to say about their meeting with professor Amy Chua in her home a year into the COVID pandemic. Yale wanted to hit Chua with some sort of COVID rule violation, and a leftist law student/hall monitor created a “dossier” on the scandalous meeting(s) she had with students. There’s no word on whether everyone was duly double-masked in her living room, or if Chua’s non-leftist books were visible to the students at any point. When I contacted former Yale professor Bandy Lee in 2018 for the data for her obviously false study, she refused, couldn’t answer basic questions about her methodology, and randomly insulted Steven Pinker, saying he, the man, “is a scam”. I’m not sure it’s theoretically possible for a hominid to be a scam, but Yale is an intellectual junkyard. I must have mentioned Pinker at some point, and she linked me to him in that leftist contagion framework of who you’re “with”, e.g. Crips, Bloods, Jets, Sharks, Science, Mensa. You should assume all her research is false, and the same for anything coming from Yale, especially anything that advances leftist ideology. It’s underexposed that with academic research, no one looks at the data, even to see that it exists. Peer reviewers don’t look at data, or check analyses. No one checks, and the word “audit” is unfamiliar to academics. (It’s not surprising that an entire package of ocean acidification studies tied to climate change is fraudulent.)

If you’re looking for research with epistemic standing, meaning that it carries at least trace levels of knowledge or insight into reality, you’re more likely to get it from the University of Tennessee at Chattanooga or Bob Jones University than MIT or Yale. At least research on political behavior, misinformation, or any topic that activates leftist ideology, e.g. freedom of speech. It’s an extraordinary reversal of past standing. For research that is likely to be true, and doesn’t suffer from catastrophic invalidity or logical errors, look to the private sector, not any academic institution. Otherwise you’ll have to design and conduct it yourself.

I won’t be the one to contact CNN to get them to loudly, affirmatively, and repeatedly retract their misinformation (a few straight days on TV and on top of their homepage seems appropriate). That’s for MIT and Yale. In my experience, leftist media outlets rarely correct false claims. It’s a broken monoculture that doesn’t police itself, and leftist gatekeepers like NewsGuard cover for them by pointing to an occasional correction as proof of an outlet’s credibility, instead of systematically assessing their accuracy and sending them a bunch of corrections (or calling for transparency on correction notices they receive, change tracking, etc.) Only NPR (see bottom — that was me) and the Guardian (bottom — me, but I didn’t realize at the time that the study was junk, so ignore it and more later) have corrected lately (without attribution). The Daily Beast, Washington Post, Axios, New Republic, VeryWell Health, FactCheck.org, and Politifact don’t correct. I got Politifact to correct in 2017 (bottom — me), but they no longer reply to emails or correct — more on them below. CNN is extremely partisan and misinformation-dense, and they don’t hire smart people (look at their site), so I’d hate trying to communicate with them. This case will take more than a few seconds to explain. It’s not a bogus inflation rate — someone will have to make them understand that the study never measured anyone posting misinformation, that it was just large incumbent media vs. lesser known outlets, that the charlatans chose the sites, covering maybe 5% of their unwitting participants’ links, rigged the set against conservatives, and that they didn’t know their participants’ political parties or orientations. The charlatans should explain it to them, or Research Ethics officers from MIT and Yale.

---

Censorship

For those who thought Twitter’s censorship was principally about misinformation, let’s pause. Twitter was supposed to censor misinformation, though its accuracy and performance are unknown. (Many factual posts were censored, and Twitter lacked the epistemic framework and staff knowledge to avoid censoring facts.) However, Twitter censors much more than purported misinformation — they censor and ban people for dissent from leftist ideology, especially the latest release of leftist trans ideology. The examples I know of are all feminist women. Musk seems likely to end that practice, though it still happens. Here are some examples from last month:

On the left is an article at a runners website arguing against the academic “gender” construct in sports on fairness grounds, given that men are stronger and faster than women even of the same size. On the right is an article at the Daily Wire reporting an Orwellian new JAMA report on “gender-affirming chest reconstruction” on adolescents (they mean cutting off girls’ breasts — double mastectomies — not reconstruction). Neither article contains misinformation, nor are they “violent” or “harmful” — they’re just humans disagreeing with leftist ideology or focusing on data that leftists for whatever reason wish to obscure. Leftists recently started using the terms “harm”, “safety”, and even “violence” to label and marginalize non-leftist discourse. That usage and underlying framework is exclusive to the left, so any censors applying such labels to discourse are enforcing leftist ideology specifically.

They’ve also censored and even banned doctors, scientists, and journalists for factual posts (falsely claiming it was misinformation), including this announcement from the Florida Surgeon General:

They restored this post hours later, without explaining the censorship. It’s just risk data, and it aligns with findings the Nordic countries acted on a year before. (Relative risks like these are often irrelevant and misleading, and I’m skeptical of the decision-making import of the cardiac risk levels here, but the CDC and leftist media pervasively traffic in acontextual relative risk estimates as well. See the CDC webpage on smoking, or really any topic.)

Censorship plays out along a spectrum, from disrupting people trying to read a forbidden article or trying to like a Tweet linking to a forbidden article, to fully silencing people by banning them. Two examples I saw recently, plus Amy Hamm:

Raicilla is a liquor my people make in Jalisco – I’m not sure if Murphy has an agave-related backstory, or if she’s appropriating my culture and our struggle to survive in the Arizona desert after crossing the “border”, with nothing but a pound of fentanyl and a bottle of raicilla wrapped in tortillas, but sure, raicilla for everyone. (Meghan, here’s a decolonized NFT authorizing your appropriation of Mexican culture – liquors and beers only – through Dec 31.)

I don’t know the details of Forstater’s ban — she’s a British feminist who recently won a legal case against her former employer regarding whether noting that there are two sexes is a view protected from discrimination by British law. I don’t know how else to describe it. In several English-speaking countries, there’s now Orwellian enforcement of leftist trans ideology that will be difficult for Americans to comprehend. In the UK, one’s tweets are a police matter and leftists will report you, and in Canada nurse Amy Hamm is on “trial” for noting — outside of work, maybe Twitter — that there are two sexes, or supporting J.K. Rowling, First of Her Name, Queen of the Angli, Breaker of Chains. Canada has a centralized, socialist healthcare system where occupational guilds/unions have licensing power, as well as the power to put people through kangaroo courts for dissenting from leftist ideology. Hamm could lose her nursing license, her career. The complainants are two outsiders, a Marxist and an anonymous person, in response to mundane views she expressed in her non-work life. The charges against her assert that noting that there are two sexes constitutes “erasure” of “transgender” people (the distinction between sex and the leftist academic “gender” construct that seemed important to academics has collapsed — I’m not sure why).

Canadian graduate student Lindsay Shepherd was targeted by a man who specialized in getting people banned from Twitter. He was infamous for demanding that women at women’s spas wax his scrotum and surrounding area, on grounds that he asserts that he’s a woman, and filed numerous complaints under Canadian law. He insulted Shepherd on Twitter, asserting (without evidence) that she had a “loose vagina”, and mocked her septate uterus condition, saying “I heard Donald Trump is building a wall inside of your uterus”. (She had spoken of her condition in a podcast, invoking a “wall” visual.) She retorted by calling him a “fat man”.

She was banned, for calling him a man. (Though soon it could be the fat. Censuring people as “fatphobic” — an undiscovered pathological fear of fat people — is on the rise.) It was perhaps foolish of her to engage at all, but her retort was accurate, and it’s worse than foolish to embrace and enforce a noxious cult that demands that people deny reality at a sensory level, strips women of their identity, and rewards awful behavior.

[Shepherd is the grad student/TA who was persecuted and bullied by two Marxist and leftist faculty members and a “diversity” apparatchik for playing her class a video of a debate including clinical psychologist Jordan Peterson. Audio of the "meeting" here. Yes, she’s lamentably bad at defending herself and making her case in that ambush, but what can you do — her first rodeo, early 20s, she’s in distress, crying, and being bullied by vicious, witless men playing at scholars.]

As far as the doctors, scientists, and journalists, Twitter banned them for alleged COVID misinformation, but I don’t know any actual misinformation cases (though there must be plenty). It’s hard to know why they were banned because Twitter never specifies, just pastes their policy. Twitter’s censors are, like their peers elsewhere, lay leftists and quite extreme. Being here in San Francisco ensures a significant divergence from mainstream American culture. On COVID, my best description of the nature of the censorship is:

C1: It’s based on a lay leftist interpretation and transformation of the recommendations of government agencies into broad, inarguable, and fixed scientific truths on topics like vaccine efficacy and side effects, and real world mask efficacy, including cloth masks.

—or—

C2: The censorship sometimes seems to target non-positive valence of any data, claims, etc. Call it the Lego Movie bias — everything is awesome (re: vaccines, masks), and if you’re saying different, misinformation! No soup for you!

[We have no mask trials in the US, and none in Western countries that found benefits for any type of mask. The obvious reasons why community masking might not reduce infection rates are not discussed, and many academics falsely suggested that an arbitrary reduction in pathogen population (pick a percentage) necessarily reduces infection risk for humans, or that masks “have to work at least a little” — nothing has to work, and any scientist talking like that is a disgrace. If you thought there was other, non-trial research that was useful, read it before sending it to me — I’ll write up that topic separately soon. Twitter might also have censored discourse on non-physical harms of masking, like social, communication, cultural, psychological, educational, and child development harms — I’m not sure.]

Americans are in an awkward bubble on the vaccines vs. other countries, in part because no one in the media does any actual work or deep dives (go look for one), reads studies, or critically examines government agencies’ claims. I doubt there’s anything huge to know about the vaccines — it’s more about rigor, severe misinformation from leftist media/politicians overstating efficacy, and the vaccines’ unknown long-term safety (the obvious reason many people choose to pass). If you were looking for beliefs in vaccine misinformation, you’d find lots of leftist beliefs to work with, driven by their media: that we (somehow) know the vaccines’ long-term safety (or that of any new pharma), false beliefs about efficacy (quantify, vary, and find the overestimate), transmission (same), that natural immunity is less effective than the vaccines (or even equal), side effects, that the CDC didn’t overestimate deaths (several times), evidence for vaccinating children, lots of cost-benefit beliefs, strong claims about mask efficacy proven by science (maybe falsely claim that randomized trials in the US found benefits, see if they agree), that ivermectin is unsafe or only for horses, that it doesn’t improve COVID recovery, that mRNA and spike proteins don’t last long in the body (perhaps unfair since the CDC just recently withdrew that claim — it was their third major bullet after “…does not affect DNA”).

Contributing to the bubble is that leftist “fact checkers” recently started tagging new peer-reviewed scientific journal articles “false”, getting them flagged/censored by Facebook and ignored by leftist media. They swarm every time science conflicts with their growing list of mandatory beliefs — often within 48 hours of people citing a study or disclosure. These “fact checks” usually just quote a random academic’s criticisms or opinions, as though the peer reviewers don’t count (and they only cite critics, never defenders). It is possible to truly debunk peer-reviewed research, depending on the details, but these quotes generally don’t contain a debunking (and the design of these “fact checks” makes it impossible to know anything without reading the paper, which is likely true regardless). The “fact checkers” themselves don’t read the papers, or lay out all their claims (most papers/studies have more than one independent claim). This is a wonderful development for a purported scientific civilization, and in the “war” on “misinformation”. Fortunately, I think the “fact checkers” and social media corporations have inadvertently exposed themselves to civil liability in that their suppression of facts and data might have straightforwardly harmed people — liability that can be enhanced by legislation.

For example, you can see the problems here right away:

Nothing causal? (Link) That’s an invalid complaint for this type of science — the researchers can only report what they have, not provide the ultimate causal story (are the alternatives to “nothing causal” something causal, a whole lot of causal, etc?)

And “not verified”? By whom? In that one, it’s not clear that the “fact checker” understands what the study says, or its various elements. He seems to think he’s debunking one of the sentences in the abstract, but it’s hard to say. He doesn’t bother quoting any academics’ opinions this time, just pastes from government webpages. This is a recurring pattern. The people “combating misinformation” think they can debunk new scientific research by quoting from arbitrary government webpages — content written (by anonymous employees) before the new science was released, sometimes more than a year before. This is a stunning failure to understand the nature and flow of spacetime, and discovery itself (and the missions of those agencies). They also clearly do not understand the scientific research they aim to censor, its methods and specific claims, and that one cannot simply compare the wording of scientific journal findings with someone else’s words to determine whether they’re “true” or “false”. They even think they can do this in reverse spacetime order, debunking new findings with prior beliefs (written by an anonymous GS-7). These morons are going to get people killed, and it’s important to linger on the reality that they genuinely believe that this is how one can evaluate scientific claims, that this is how knowledge of reality works — this savage stupidity is pervasive at leftist “fact check” operations like Politifact, FactCheck.org, AP, et al.

If we want true fact checking in this civilization we’ll need IQ requirements (≥125) and education/knowledge testing, and they should permanently forget about “fact checking” scientific questions and journal articles. We have decent scientific methods for assessing bodies of research and competing claims (e.g. meta-analysis). Quoting the opinions of one or two leftist academics — out of a pool of ≈100-5,000 similarly credentialed English-speaking “experts” — is not a valid method of knowing anything about anything.

Try posting the above journal articles to Facebook and see the flags and promised punishment — the Nature article (left) and the one at Science Direct (Food & Chemical Toxicology). I’ve posted so much science on Facebook that I’m now invisible on friends’ feeds, and they have to go to my profile to see my posts. That’s what America has become, and when the editors of the BMJ wrote an open letter to Mark Zuckerberg about this, there was no response and no leftist media coverage. Twitter doesn’t empower outside leftist activists to tag science as “False” like Facebook does — their censorship decisions are in-house, and they banned two of the authors of the F&CT paper above. (Note that I’m not endorsing that paper or the authors, haven’t read it, and I’m skeptical of strong claims about the vaccines.)

Finally, the MIT and Yale charlatans didn’t mention the surprising story of journalist Alex Berenson. Twitter banned him last year for COVID-related blasphemies. He sued Twitter, and, incredibly, won. He’s now back (this all happened before Musk’s purchase).

It’s hard to fathom how he could win if he was in fact guilty of posting misinformation (or much of it), but I don’t know the details. Legal discovery being the beautiful thing it is, he discovered that a Pfizer director who was then with the US government and leftist activists at CNN (Oliver Darcy) and Axios (Ashley Gold) had emailed Twitter to censor him. Like the MIT and Yale charlatans, the emails don’t identify any, like, misinformation:

She’s typical of what we’re working with in leftist media — no scientific background, no rigor, no curiosity to, like, know or uncover anything, just a random leftist activist with, like, a BA. Axios is a low-calorie leftist website that doesn’t correct false claims (or respond), and she and her colleagues write entire articles (very short James Patterson style) on “misinformation” without citing a single example of the dark art. Like MIT and Yale, they’re layers and layers removed from reality, from direct engagement with scientific research, terminating a long chain of assumptions, gists, authoritarianism, mockery, and reactionary bias. Nowhere at Axios, CNN, the Washington Post, the New York Times, etc. will you find smart, well-educated people who can read papers, ask questions, and conduct deep dives, or just do their jobs without caring about political parties or who said what and which tribe they belong to. Smart people have been forced to go elsewhere, and I concur with UCSF professor, hematologist, oncologist, and man of letters Vinay Prasad:

The COVID19 pandemic reveals the limitations and arrogance of science influencers. Pre-covid they were often on target. In a world of reiki, jade eggs, supplements and cupping, it was easy for health misinformation ‘experts’ to make their mark debunking popular, unproven remedies. Why? Because these things are obvious nonsense, and you don’t need to know much to know that.

But these ‘experts’ were woefully unprepared for the pandemic. Lockdowns, the possibility of lab leak, school closure, masking adults, masking kids, 5 vs 10 days of quarantine, and who should get how many doses of vaccine and when— are complex technical questions that require deep knowledge of biomedicine, trials, trade-offs, statistics, and more.

Science debunkers and influencers—who often don’t work at universities, don’t publish research, don’t understand statistics, don’t peer review for journals, don’t have a good sense of the pre-test probability of interventions, and/or don’t have deep technical understanding of drug regulation—tried to crusade against misinformation, but they made many mistakes. This would be okay if they were merely debating, but repeatedly they sought to use the tools of the platform to extinguish ideas they disliked.

Interestingly, influencers almost universally share the same political views— mixing debunking with far left socio-political positions and open endorsement of Democratic candidates. They draw upon tactics used in politics and the culture war and import them to medicine: opposing ideas are not just wrong, they are dangerous and harmful. We can’t fight speech with speech, we have to censor, label, de-platform, and down-throttle.

Science influencers have chosen to label fellow doctors and professors who disagreed with them as “anti-science,” when, ironically, it was their rhetoric which reflected an inability to understand the nuance inherent in science. Science influencers, in their zeal, stifled debate on how to give vaccines more safely. They thwarted efforts to run cluster randomized trials of masking.

…

The Venn diagram of science influencers, far left socio political views, and a desire for canceled culture tactics have coalesced to create #TheScience — a bizarre group of individuals who insist on policing information they barely understand.

It would be one thing if their preferred tactic was merely to offer critical speech. People are free to shout about all sorts of stupid things in our society. But their preferred tactic is to restrict the speech of others and to pressure employers to fire their employees, as occurred in the case of the myocarditis preprint. Both of these actions make the actual conduct and debate of science difficult and thankless.

MRNA vaccines for COVID-19 represent one of the greatest societal successes of the last quarter century. The time span in which they were developed and the initial efficacy in the elderly was so remarkable that even the optimists look close-minded. But these vaccines have been misused in low risk populations. The data for children, particularly those with prior immunity, is woefully inadequate. It is woefully inadequate for a widespread vaccine program, as I detailed in Bari Weiss’ substack.

To be clear, I don’t endorse any particular claim of Berenson’s — I don’t know enough. He’s just an example of the banned, and it’s strange that leftist outlets like CNN or the New York Times (his former employer) haven’t jumped on the fact that a Pfizer director urged Twitter to censor an indy journalist who published unfavorable claims about Pfizer’s product, or Berenson’s remarkable journey of being banned, suing, and winning.

Back to the MIT and Yale artifact…

---

Assuming

The MIT and Yale artifact is about 4,500 Twitter users who shared the #Trump2020 hashtag in October of 2020, compared to 4,500 users who shared the #VoteBidenHarris2020 hashtag. That’s it. There’s no explanation why those specific tags are important, or if they were the modal tags for supporters (note their structural differences). The authors used strange, asymmetric filtering steps to produce their sub-samples, and removed 426 “elites”, but I’ll set that aside because we don’t have a study here anyway.

Incredibly, the charlatans call the Trump taggers “Republicans”. Sometimes they say “conservatives”.

They call the Biden taggers “Democrats”. They assigned political parties to their unwitting participants. I’ve never seen anyone do this, and it’s wildly unacceptable.

As well-educated people know, the Trump taggers won’t all be Republicans. Some will be independents. Some Democrats. Those who are Republicans might not be representative. And the Biden taggers won’t all be Democrats. Baseline political literacy would give you the basic outlines. What percentage of voters voted for Trump in 2016 and 2020 (or any major party candidate, win or lose)? And what percentage of voters are Republicans? If you know those numbers, ballpark, you’d do the obvious math in your head in the time it takes to finish this sentence — expect about a third of anyone’s voters to not be of their party. You’d also know that most independents voted for Trump in 2016, and 41% voted for him in 2020.

Moreover “conservative” is a political orientation that may or may not describe people who used the Trump tag — it’s implausible that they’re all conservatives. Note that humans don’t need to have a political orientation — that’s optional, and they certainly don’t need to select from only conservative/leftist options.

So the study’s dead from the start, as soon as they call these people Republicans and Democrats and conservatives, assigning them political parties and orientations. Humans don’t have to fit the crude otherizing box that leftist academics draw for them. This is worse than a random layperson’s spitballing. Academics are generally too poorly educated, biased, and culturally prejudiced to conduct valid, accurate research in the political domain, especially when it comes to outsiders, non-leftists, conservatives, libertarians, Christians, people who don’t vote for leftist politicians in a given election, etc. Telescopes are good for stars, not people. If academic institutions want to do credible research on politics, they’ll need to bring in lots of non-leftists — it’s not possible to validly research other cultures and perspectives when you don’t include any humans representing those cultures and perspectives, at least not given the prejudices and lack of education typical of academics. Here, the MIT and Yale charlatans didn’t know the basic party numbers for their own country, or about independents, or the many reasons people gave for supporting or opposing Trump or Biden.

There are countless reasons why people might vote for someone, and in a perceived forced-choice task like American elections, their decisions might be more about the other candidate/ideology. If academics want to learn about and accurately characterize humans who disagree with their political ideology or candidates, they’ll have to read their discourse and talk to them, and let them speak in their own words (rather than researchers’ items consisting only of conspiracy theories and “misinformation” they attach to conservatives, strangely missing conspiracies and misinformation popular with leftists). There were many victims of nationwide leftist violence in 2020 specifically, and even more victims of the crime wave leftists sparked by degrading police performance and prosecution in their war on policing as such, defunding, etc. Some of those Trump taggers might have had their businesses burned down or cars smashed in the riots, or known people who had. Many communities lost stores, closed due to the rise in smash-and-grabs and brazen bulk thefts that started with the riots (we lost another five Walgreens in San Francisco last year — note this is where people get medications.) They could also be current or former leftists alarmed by the censorship, child-harming authoritarian policies, and lack of scientific rigor they see in the left, or had any other reason(s) for their decisions. You’d have to ask them. (The linked examples might have broken from the left after 2020 — they’re just examples of the phenomenon.)

---

Rigging

We would only proceed if we were interested in the misinformation-related behavior of this strangely filtered subset of Trump taggers and Biden taggers, but we’re derailed within seconds. The authors chose the websites in advance, arbitrarily.

This study is about link posting. Posting a link from a given website is the behavior coded here (e.g. huffpo.com, latimes.com). To assess that behavior, you’d have to take the links/domains users posted. The authors didn’t. They chose the websites themselves: some based on general popularity, and some from much longer lists of “misinformation”, “hyperpartisan”, or even “parody” sites created by lay leftists, filtered by general Twitter popularity in early 2018 (two and a half years before their study). There were sixty sites, maybe thirty that would’ve been linked by more than a few people.

We have no idea which websites users linked, how often, how many were not in the authors’ arbitrary list, or the misinformation-related characteristics of those or the authors’ websites. There’s no way to know who is linking to more “misinformation sites” using this method — it’s invalid.

They also don’t disclose the websites they used — you’d have to read their PNAS paper to see.

In PNAS, we also learn that they overrode their own filter methodology and added two conservative domains post hoc: dailywire.com and breitbart.com. They say they added them, but don’t explain why.

It’s easy to achieve their results by excluding some “misinformation sites” that Biden taggers might have frequently posted — serial hoax publisher Rolling Stone, Alternet, The Daily Beast, Jezebel, and Buzzfeed stand out as missing (but it could be anything since they won’t all be leftists or Democrats, and properly characterizing these sites will require systematic, impartial assessment). Moreover, given a continuous scale score for domains, people posting links from reliable sites excluded by the authors would get no credit for doing so. They excluded National Review, likely the most popular conservative site in 2020, Townhall, The Federalist, City Journal, The Post Millennial, and Quillette (not conservative, but certainly left-critical). But again, the Trump taggers won’t all be conservatives, and the characteristics of these sites are empirical questions, as are the political orientations — if any — and actual linking behavior of the users.

I expect some readers are fidgety with annoyance because you’ve already seen why we’d never be able to assess misinformation based on links to websites:

Even unreliable sites are mostly reliable.

You’d have to assess them to find out, but I predict that even the worst sites will land under a 15% misinformation rate if we code them by article (and thus link) using a rigorous epistemic framework. There would be no way to assess misinformation by site if the accuracy rate is 0.972 for Fox News links, 0.903 for the New York Times, 0.891 for the Washington Post, 0.981 for The Post Millennial, 0.911 for the CDC, 0.989 for the Congressional Budget Office, etc. Even with weights for severity (like the latest trend in leftist outlets like the NYT and NBC News of citing unnamed experts) it’s hard to imagine being able to do anything with website alone. The potential dynamism of such realities, and the inapplicability of a site’s overall score to any given article, and to any person’s pattern of linked articles, also rule out this approach. If we aim to study misinformation, we’ll need to identify it.

---

Epistemology of the Mob

MIT and Yale had random people rate domains on one item: “How much do you trust each of these domains?”, on a five-point scale: Not at all, Barely, Somewhat, A lot, Entirely.

They rated literal domains, like conservativetribune.com, patriotpost.us, redstate.com, and commondreams.org. (Yes, the domains leaked political orientation, but that’s the least of our problems.)

There’s no universe in which scholars or scientists could characterize a website or publisher as a “misinformation site” by having people rate how much they trust the domain/site. If “misinformation” means false claims, opinions about domains are irrelevant — you have to do the usual work of scientists and go research the phenomenon you’re interested in, objectively.

They separately asked “Do you recognize the following websites?”, with a No/Yes option.

They kept the ratings for participants who said they didn’t recognize a site. All of them. Most of the small sites were recognized by fewer than 25% of participants. (They don’t disclose the rates for any sites, just the 25% threshold in their PNAS paper.)

There’s also no universe in which scholars or scientists could characterize a website or publisher as a “misinformation site” by using floor-level trust ratings of people who said they were unfamiliar with those sites. Even if humans had remarkable powers to characterize familiar websites as misinformation-or-not sites by saying how much they personally trusted them on a five-point scale, such powers surely don’t extend to unknown websites.

The method is rigged to score large incumbents as non-misinformation sites, and like all the rest, the reliability of those incumbents is an empirical question for which the charlatans did no work.

The result: All smaller sites were scored as “misinformation sites”.

Most of the small sites were conservative. They rigged their arbitrary set — only five of twenty were leftist.

This is Study 2 from their PNAS paper, which they claim they used for the present paper. All the smaller sites are at the floor. Republicans have them there too, even conservative sites, which many of them were unfamiliar with. Note that in Study 1, Republicans trusted PBS the most, a hair above Fox News, and trusted the other big outlets more than they do above. The authors chose to use Study 2 instead, which excludes PBS.

As you can see, Republicans were in the A Little-to-Somewhat ballpark for most big outlets, and were measured even for Fox News. Depending on what the empirical reality for those outlets turns is, Republicans might more accurate, less gullible, arguably wiser in their restraint compared to Democrats — we’ll have systematic, high quality data on those outlets in a year or so. Unlike the authors’ implications, Republicans did not agree with Democrats. The scores were correlated — that’s not the same as agreement. Same floor, different ceiling. I couldn’t explain why, but academics are extremely incompetent in their use of inferential statistics, and often conflate correlation for substantive standing on a variable, or agreement/disagreement.

Public Service Announcement: This error is the source of a great deal of misinformation that leftist academics have spread about conservatives, such as that they’re high in “Right-Wing Authoritarianism” (RWA) and “Social Dominance Orientation” (SDO). They’re not — conservatives don’t generally endorse or agree with the items on those scales. They cluster around, or just below, the neutral midpoint (note that below the midpoint marks substantive disagreement for those scales). Leftists (and probably libertarians) are at the floor, thus you get the correlation. It was unnoticed because academics typically conceal their descriptive data — where people scored — only reporting their inferential statistics. They’re the opposite of Gallup and Pew — who only traffic in descriptive statistics — and much less rigorous, useful, and trustworthy. The SDO and RWA items are absurd caricatures written by leftist academics, embed leftist ideological assumptions that humans are principally members of groups locked in struggle with other groups, and don’t map to conservative discourse. For example, from SDO: “In getting what you want, it is sometimes necessary to use force against other groups.”, along with mundane non-leftism: “Group equality should not be our primary goal.” Note you’d score high in something called social dominance orientation for agreeing with that one, for not being a leftist. For decades, academics have reported “conservatism predicts Social Dominance Orientation/RWA”, or even that they’re high in SDO/RWA (false), without disclosing that conservatives don’t endorse the items. See Samantha Stanley’s recent doctoral dissertation for more, pages 274-276.

The authors also had eight leftist activists — at least two of whom were foreign — rate these American domains on the same five-point trust item. (They might all have been foreign — see the PNAS paper, last page). We’re miles away from that mattering at this point, and the activists were predictably more partisan than regular (American) people.

---

Let’s pause on epistemology, which will be central to an account of misinformation, since it refers to false claims. We’re having to piece together leftist epistemology from their behavior — I haven’t found any formal discourse where they lay out their epistemic framework.

Trust: This seems to come up a lot in their discourse. Here they specifically asked people how much they trusted the domains, not how much misinformation they thought those domains publish (either question is a waste of everyone’s time). This trust idea seems to imply passive consumption, where you either believe something or not based on whether a site “can be trusted”. That’s a framework destined to collapse on examination. Reality is a lot more contextual, different types of claims have different trust implications, and humans can think — or should be encouraged to do so. In my experience, quotes and statistics are almost never fabricated — the issue will be the details of what a stat actually means, the tricks the writer plays, whether the quotes are microquotes sandwiched in the writer’s paraphrasing, broken logic, whether someone’s political orientation is outed in advance of their claims (popular with “fact checkers”, heavily biased toward outing non-leftists), etc. I predict that media deception works mostly due to readers’ lack of reasoning skills, basic statistics training, and baseline attentiveness. Suggesting that people trust any media outlets is reckless if we haven’t rigorously evaluated what those outlets are doing.

Here’s an example from FactCheck.org:

If I see that, I know I’m not dealing with professional fact checkers — what does her opposition to that law have to do with the fact check? (And they out everyone else in their artifact as “right-wing” or “conservative”. With Dr. Victory, I guess all they had was the “dirt” they dug up on a past policy stance that would peg her as the Other.)

---

Here’s a baseline attentiveness case — you just need to not be asleep. A Politifact fact check of Sen. Chris Murphy (D) (ignore the graphic about ABC News in the middle):

Don’t ask me. This is normal for Politifact — they’re morons and leftist activists. Truly morons. (And they switched from killers to shooters — they often switch variables.)

---

Here’s an example that might require slightly more cognition than grasping that if most of a group are under 18, then most can’t be 18-19. Politifact and its parent Poynter did this:

His claim was true, and critically depended on the word and, as humans are wont to do. No President had ever specified both race and gender requirements for a nominee (a black woman in this case). Politifact asserted Hannity’s true claim was false because some Presidents had targeted or mentioned gender, and some Presidents had targeted or mentioned race (different nominees — never the combo). As I said, they’re morons.

Then Poynter did this:

See what they did? The and is gone. As I said, the self-styled “fact checkers” are the least reliable sources of facts or truth right now.

All these issues I see most on leftist outlets like the Washington Post, Axios, Politifact, the Atlantic, FactCheck.org, etc. compared to National Review, Quillette, Reason, Fox News, Wesley Yang, random academics like Mike Huemer and Bryan Caplan, etc. I only see the severely broken logic cases on leftist sites, and it feels like a major new development. I’m willing to take all sorts of $1000+ bets on leftist outlets collectively being less reliable than conservative outlets, or non-leftist outlets more broadly (there are lots of interesing and not politically simple publishers these days). Think through your variable/measure and let’s see if we can answer it in the next year or so.

Conditioning people to navigate the world with a binary and mindless trust vs. don’t trust frame would lead them to deny reality and facts when they come from sources they’re told can’t be trusted. This is not wise, and not good for a civilization, especially given the coverage biases of leftist, conservative, libertarian, foreign, religious, and other media — if you don’t read from a broad set of sources, you’ll know little and might miss serious problems in society.

Mass opinion as the arbiter of truth: This seems like a consensus bias, or a what other people think bias, but an insane collapse-level version. We’ll never be able to know a reality like the misinformation risk of a website based on anyone’s opinions, much less those unfamiliar with the site. Moreover, it’s a logical error to treat a 92% accurate site as a “misinformation site” because of those trust ratings, and a 94% accurate site as a non-misinformation site, also because of those trust ratings. This is nonsense.

Links: People can have and share their own thoughts, so focusing only on links to other people’s thoughts again points to a bias toward conformity, authority, leaning on “credentialed journalists” (like the imbecile activists above). And in this case only linking to big corporate legacy outlets that also happen to have a profound leftist bias.

Scientifically, we have no reason to assume that misinformation on Twitter only or mostly comes in the form of links. Humans can have their own thoughts and words — they don’t always need to lean on other people’s. They also have memes. No links necessary.

---

Note that most human discourse is not subject to classification as misinformation or accurate information. The authors assume that Twitter is right to ban people for “misinformation”, as if this were self-evident. If we were interested in judging people’s online behavior — a big if — we’d have to be systematic. We’d quickly realize that there’s much more going on than making false claims at some weekly, monthly, or annual rate. For one, there are positive contributions, not just negatives — the accurate claims and news people post, rich, smart essays, stories, music, art. We might prize the intelligence or humor in someone’s posts. Behavior online can include providing emotional support, or attacking others in countless contexts. People can revel in violence against other-side politicians (see below), or shame a person into committing suicide. Or they might share haikus.

A singular focus on misinformation is a philosophical decision, and if leftists want to judge and marginalize humans who transgress in this particular respect — ignoring everything else about such persons — perhaps they should make their case before trying to rationalize such marginalization post hoc.

The authors cite no examples of banned users — they didn’t talk to any, or cite any accounts. Nor do they refer to Twitter’s censorship and banning policies of the period — that leaves us blind. They betray no awareness of the fact that many people were banned by Twitter not for misinformation, but for dissent from leftist ideology, especially the new trans ideology.

---

Misinformation Examples

There are plenty of examples of leftist misinformation for the curious. I searched for #BidenHarris2020 in the Twitter app on my phone, and gathered four of the tweets below in less than one minute. The other examples are just well-known misinformation cases. From the Biden tag search I learned that:

1. Tulsi Gabbard is a notorious Russian asset/agent.

2. The vaccines were almost two years away (as of Sep 17, 2020), because the Pfizer study was set to be two years long…

3. The vaccines wouldn’t be distributed before the election because they required specific bottles, made from special sand, and we were short on that sand due to Trump’s incompetence. (They weren’t distributed before the election, if it matters.)

4. A setup video of Republican Senate candidate Blake Masters shows that he empathized with the undercover leftists’ compelling tale of killing illegal immigrants on their border property. (The video is hilarious — they think that’s how conservatives talk, and the “right to a militia” is an amazing misread of the Second Amendment. I felt bad for Masters — he was friendly and approachable, and they kept trying to stick a knife in him.)

5. Something about conservatives being the product of alien demonseed, drinking bleach, and a Russian takeover of the United States. Hard to follow.

6. Some sort of conspiracy theory about Ivanka Trump dining with Chinese leader Xi Jinping on Day X, and doing other things on other days, like getting patents and trademarks.

That’s it for Part 1. Part 2 will dive into a historically unprecedented fact check from the Washington Post, and provide tools for training your students and other sentients using the Rand, Pennycook, et al papers.