Opinion: Fact-Checking Facebook’s Fact Checkers

***

The Facebook label links to the third-party site Health Feedback, a member of a World Health Organization-led vaccine project and an affiliate of the nonprofit Science Feedback that verifies scientific claims in the media. Another Science Feedback affiliate fact-checks climate-related articles in predominantly conservative media.

“Misleading Wall Street Journal opinion piece makes the unsubstantiated claim that the U.S. will have herd immunity by April 2021,” Health Feedback’s “fact-check” says. “Three scientists analysed the article and estimate its overall scientific credibility to be very low.” This is counter-opinion masquerading as fact checking.

Dr. Makary didn’t present his opinion as a factual claim. He argued, based on studies and other evidence, that Americans would have enough immunity from vaccination and natural infection by early spring to sharply reduce the virus spread. He essentially made a projection, much like the epidemiologists at Imperial College and University of Washington do.

But the progressive health clerisy don’t like his projection because they worry it could lead to fewer virus restrictions. The horror! Health Feedback’s fact checkers disagree with the evidence Dr. Makary cites as well as how he interprets it. Fine. Scientists disagree all the time. Much of conventional health wisdom about red meat, sodium and cardiovascular risk is still fiercely debated.

The same goes for Covid-19. There’s still much we don’t understand about the virus and its transmission and immunity. Yet Facebook’s fact-checkers “cherry-pick,” to borrow their word, studies to support their own opinions, which they present as fact. So let’s fact-check Facebook’s fact checkers.

***

A Kent State epidemiologist quibbles with Dr. Makary’s claim that “when [reinfections] do occur, the cases are mild.” Her evidence? A single case report in the Rhode Island Medical Journal that cited a handful of severe reinfections globally since April. Yet that same study noted that “there have only been a few reports of reinfections in COVID-19 patients.”

A New England Journal of Medicine study last month identified only two asymptomatic possible reinfections among more than a thousand U.K. health-care workers with anti-spike antibodies from prior infections. Another study last week in the Journal of the American Medical Association found antibodies from prior infection conferred a degree of protection that “appears to be comparable” to mRNA vaccines in clinical trials.

Another fact checker says Dr. Makary incorrectly extrapolated the share of Americans who have probably been infected. We will never know this number with certainty because most people with no or mild enough symptoms don’t get tested. Antibody surveys can help extrapolate infections, but antibodies wane over time and more quickly in mild and asymptomatic cases. So they may underestimate infections.

A study in the journal Clinical Infectious Diseases found that 2% of American blood donors had antibodies in mid-December 2019. Early in the spring, testing was only picking up one in 11 cases. The ratio fell by the fall to around 1 in 4 as testing expanded. Dr. Makary applied what he called a “time-weighted case capture average of 1 in 6.5.”

A disease ecologist at the University of Santa Cruz says he should have used 1 in 4, which would have extrapolated a lower estimate of infections. This is an arbitrary number and would underestimate the number of infections during the spring. She also claims the 0.23% “infection fatality rate” that Dr. Makary cites is wrong and says it is around 0.6%.

Infection fatality rates (IFR) are based on models and vary by population demographics and the degree to which societies protect the elderly. Stanford epidemiologist

John Ioannidis

found the median IFR estimate across 51 locations worldwide was 0.27%, but many estimates were from the spring when there were few treatments. The infection fatality rate has since likely fallen.

A recent study in the journal Science estimated an infection fatality rate of roughly 0.23% in Manaus, Brazil. Speaking of which, a Harvard epidemiologist takes issue with Dr. Makary’s citing Manaus as an example of “herd immunity.” Epidemiologists estimated 52.5% of Manaus had been infected by June at the same time hospitalizations plunged and stayed low for seven months despite relaxed government restrictions.

The Washington Post in August pointed to Manaus as a possible real-world example of herd immunity. Yet hospitalizations inexplicably shot up in December. A recent Lancet paper offers several hypothesis for this surge, including new virus variants that evade antibodies and flawed models. These hypotheses deserve to be investigated.

Yet Facebook’s Harvard epidemiologist concludes Manaus should “be more than enough to demonstrate the dangers of trusting to ‘herd immunity’ from infection for protection.” It’s always a red flag when a scientist proclaims that a single piece of ambiguous evidence is enough to demonstrate anything. These Facebook fact checkers aren’t acting like scientists.

***

Scientists often disagree over how to interpret evidence. Debate is how ideas are tested and arguments are refined. But Facebook’s fact checkers are presenting their opinions as fact and seeking to silence other scientists whose views challenge their own.

We’ve been leery of proposals in Congress to modify Section 230 protections that shield internet platforms from liability. But social-media giants are increasingly adding phony fact checks and removing articles flagged by left-leaning users without explanation. In short, they are acting like publishers in vetting and stigmatizing the content of reputable publishers. The legal privileges that enable these companies to dominate public discourse need to be debated and perhaps revised.