One day last August, as they struggled to figure out whether to lift Covid-19 restrictions, the supervisors of Placer County, California, convened a panel of experts. It was a reasonable move. If being a local official could be thankless in normal times, the pandemic had made it nearly impossible. Federal messaging had been hopelessly muddled. Rules meant to stop viral spread came with painful side effects. One constituent insisted the sheriff enforce lockdowns; another called stay-at-home-orders an economic death sentence. Wanting advice from doctors and professors was hardly surprising.
What was surprising was that the first invited speaker had chosen to frame himself as an authority on Covid-19 at all. His name was Michael Levitt. His credentials were stellar — an endowed Stanford professorship, one-third of the 2013 Nobel Prize in Chemistry — but utterly unrelated to infectious disease outbreaks. He’d won his honors with the computer-programming work he’d done in the 1960s and 70s, revealing the intricate origami of proteins, modeling how they fold and form the tiny machinery of life. Prior to those papers, the chair of the Nobel selection committee had said, studying chemical reactions was “like seeing all the actors before Hamlet and all the dead bodies after, and then you wonder what happened in the middle.” Levitt and his colleagues had described “the whole drama,” showing how each character died.
Now, though, dead bodies weren’t a metaphor. They were horrifyingly literal, their bagged bulk filling hospital morgues and refrigerated trucks. Public health specialists begged politicians and citizens to do what they could to slow transmission. Levitt, a biophysicist, had different ideas. He derided policies proposed by the vast majority of epidemiologists as “politically correct.” He liked to say he wasn’t against lockdowns, but against stupid lockdowns. Almost as soon as his face filled the Board of Supervisors’ screens, he acknowledged his mismatch in expertise: “I have been looking at Covid for a very long time, but I should say, as has been said about me numerous times, I am not an epidemiologist, I don’t have an MD.”
Yet this wasn’t his first time advising politicians on their pandemic response, and it wouldn’t be his last. By his own account, in March 2020, he’d spoken directly with the Prime Minister Benjamin Netanyahu of Israel and said he’d be surprised if the country saw more than 10 coronavirus deaths. (According to STAT’s Covid Tracker, the count is now over 6,400.) A month after appearing at the county supervisors’ meeting in August, he would take part in a roundtable on reopening, hosted by the governor of Florida, where he’d suggest that it was common sense, “for social reasons and for herd immunity,” to let young people be exposed to the coronavirus.
It’s become a pandemic truism that everyone — economist, tech bro, joe schmo — thinks they’re an epidemiologist, or at least wants to contribute their expertise. That war-effort way of thinking has yielded some of the most reliable sources: Look no farther than the Covid Tracking Project, a data repository spearheaded by journalists and coders and students. Some of the most respected virologists and epidemiologists are distilling science and guidance, tweet by tweet. But it’s easy for an impressive CV to make someone sound more credible about Covid than they should be. To those who’ve spent careers studying infectious disease, Levitt is a worrisome case study, a source of headline-grabbing false balance, and a conundrum for universities. How to uphold academic freedom without lending your brand to misleading, potentially harmful claims?
Levitt’s messaging has wavered slightly. He now admits he was wrong about the number of deaths in Israel, and wrong that the outbreak would be over in the United States by August 2020. He’s broken ties with the anti-lockdown group PANDA “due to their anti-vaccine stance.” He’s hardly the only modeler whose models have fallen short. His critics fully agree with him that imposing Covid restrictions comes at a great societal cost. What they worry about is his tendency, in conversations with politicians, in media interviews, and in tweets to his 90,000-some followers, to downplay the coronavirus’ impact.
US COVID19 will be done in 4 weeks with a total reported death below 170,000. How will we know it is over? Like for Europe, when all cause excess deaths are at normal level for week. Reported COVID19 deaths may continue after 25 Aug. & reported cases will, but it will be over. https://t.co/lnpxZ3bHIy
— Michael Levitt (@MLevitt_NP2013) July 25, 2020
“Michael Levitt has a huge, huge following, so this creates lots of problems when he’s tweeting something that may be misinformative,” said Maimuna Majumder, a computational epidemiologist at Harvard Medical School, who specializes in infectious disease modeling and pandemic response. She cited a recent case in which Levitt had incorrectly tweeted that India’s Covid-19 case fatality ratio, an estimate of the proportion of deaths among confirmed cases, was low. “We don’t want people to think of this as less deadly than it is,” she said.
Levitt, meanwhile, sees himself as an unbiased analyzer of data, an apolitical voice being outscreamed by partisans. “We were basically called the ‘Stanford terribles,’” he said, of himself, epidemiologist John Ioannidis, physician and health economist Jay Battacharya, and Scott Atlas, the radiologist who served on President Trump’s coronavirus task force. To Levitt, that moniker was a good thing, a sign that his university is a bastion of open-mindedness. “The diversity of discussion at Stanford has probably been the most of any place in the country,” he told STAT.
Many of his colleagues share the ideal, but say he hasn’t acted as a responsible member of the academic community. The question is how the institution should respond. “It’s not like we had a plan: ‘So imagine we have this type of emergency. What happens if people who don’t know anything about that domain start spouting opinions?’” said James Holland Jones, a Stanford researcher of human social behavior and infectious disease. “We’re kind of making it up as we go along.”
“Michael Levitt has a huge, huge following, so this creates lots of problems when he’s tweeting something that may be misinformative.”
Maimuna Majumder, Harvard Medical School
There is little as powerful at conferring near-mythical status as the Nobel Prize. MacArthurs, Pulitzers, and Fields Medals all have pull, but for the general public, they don’t have the dazzling name recognition of the Nobel. That sort of reverence is laden with risk. Leave aside the war-mongering Peace Prize winners and the fascist, colonialist littérateurs, focus only on those objectivity-loving scientists, and you’ll still find a number of laureates who’ve gone off the rails.
Kary Mullis, who won for co-inventing PCR testing, went on to deny that HIV causes AIDS, helping to sway South African president Thabo Mbeki into rejecting antiretroviral therapy, costing hundreds of thousands of lives.
Linus Pauling went from two-time Nobelist to full-time quack, a brilliant peace activist and investigator of chemical bonds who became convinced that everything from colds to cancers should be fought with vitamin C.
It was into this troubling tradition that Randy Schekman now saw Levitt inserting himself. Schekman is a Nobelist as well, but in the category of physiology and medicine, and he knew Levitt to be a world-class computational and data scientist. But as cases and deaths rose beyond Levitt’s predictions, Schekman increasingly felt Levitt was “on the wrong and possibly dangerous side of the facts.” And when Levitt signed the Great Barrington Declaration, which advocated for those at low risk of death from Covid “to build up immunity to the virus through natural infection,” Schekman felt his colleague had “gone off the deep end.”
“Michael absolutely deserves the freedom to pursue his own ideas and to develop even controversial positions to the best of his considerable ability. However, in this instance, I believe he crossed a boundary from data to public policy where the impact of his word as a Nobel laureate has undue influence,” he wrote in an email to STAT. He mentioned the tales of Pauling, Watson, and Shockley as “a cautionary note to those who may be tempted to use their past achievements to influence issues in the public realm.”
Influence on the world stage is what sets Levitt apart. Arieh Warshel, his old collaborator and one of his co-winners, wasn’t surprised to see Levitt turn his attention to Covid; he’d done the same thing. “Moving to hot problems has been what both of us did all the time,” Warshel said, expressing support for his colleague’s current work. Warshel, though, stuck closer to his old field, published his Covid research in peer-reviewed papers, and didn’t become a Twitter personality.
The way Levitt himself tells it, his pivot-to-Covid happened by accident. His wife, a curator of Chinese photography, has many friends in China. So, in late January of 2020, around the Chinese New Year, she’d reached out to them on WeChat, saying that she knew it was a frightening time and she was thinking about them. In their responses, she and Levitt could sense just how isolated they felt, facing this outbreak that nobody knew much about. “I thought, if my wife can do her share, I should start to look at the numbers,” Levitt said.
He’d won the Nobel for research he’d done in his twenties, and he was now in his seventies, a beloved mentor to younger biophysicists. At the award ceremony in 2013, he’d joked that everything he’d done happened 45 years before he got to go to Stockholm. “Which is quite normal,” he said recently, “it’s not a big deal.” By his own admission, he’d never studied anything like Covid before. But he had spent his career computer-modeling complex systems, trying to understand how proteins self-assemble. This application was a bit of a departure, but, he said, “maybe as you get older, you start to think about whether what you could do could be more useful to people.”
His reports began as informal messages to friends in China, comparing data from different parts of the country. But then around March 14 — he remembers the date because it’s Pi Day — a friend posted one of Levitt’s reports on “the Chinese equivalent of Facebook, and it got seen by 2 million people in the first day. So I was kind of cooked.”
This was right around the time that global anxiety was building up. Emails, he said, began pouring in. “I started to get the attention of the Israeli media,” he said — after his childhood in South Africa, he’s lived and worked extensively in Israel, as well as England and the U.S. — “and was asked to consult with Netanyahu.”
He began giving interviews. “Israeli Nobel Laureate: Coronavirus spread is slowing,” read one March 2020 headline from the Jerusalem Post. “Why this Nobel laureate predicts a quicker coronavirus recovery: ‘We’re going to be fine,” read another from the Los Angeles Times.
It was around this time that he began using Twitter in earnest. He’d had an account since 2016, but it was just a tool to nudge companies into responding to his queries. “I had used Twitter only to complain about service by American Express and CitiBank and things like that,” he said. “And that’s why my Twitter handle was a very outrageous MLevitt_NP2013” — NP for Nobel Prize — “just because that really got me customer service.”
Now, it got him a lot of followers. His Twitter presence looks the way you might imagine a scientist’s would, full of graphs and ratios and statistical terms. But nestled among the percentages, epidemiologists began to see questionable theories. Levitt, for instance, wrote last December that “society must assume only symptomatic infect,” though evidence of coronavirus spread from asymptomatic or presymptomatic people was clear by then. He’s claimed that countries were, “for mysterious reasons,” overcounting Covid deaths, and that people were dying with the virus but not of it, when in fact, medical examiners reported the opposite.
Colleagues raised questions early on. When Levitt sent out his analyses to the listserv for the fellows of the International Society of Computational Biology, Lior Pachter, an often-outspoken professor at CalTech, saw what looked to him like typos and issues with mathematical assumptions. He wrote back to Levitt and suggested he write up this research more formally, post it to a preprint server, release the code, and solicit feedback, without blasting it out. But Levitt kept on the same course, sharing his calculations with the fellows. In one case, he’d written, in all caps, “please distribute as you can.”
Deaths assigned to COVID19 are often connected to other conditions. For mysterious reasons, countries seemed to want to maximize their death counts. Many who died WITH corona were counted as dying OF corona. The only true death count is excess death in the particular period.
— Michael Levitt (@MLevitt_NP2013) May 19, 2020
For Mallory Harris, a Stanford Ph.D. student, what caught her attention most was a soliloquy that Levitt gave as part of a panel. Epidemiologists, he’d said, “see their job not as getting things correct, but preventing an epidemic. … We should never have listened to the epidemiologists.” He went on to say that they’d caused enough suffering and damage “to make 9/11 look like a baby story.”
Harris is an epidemiologist — she studies the effects of temperature on mosquito-borne disease — and she was startled to hear this from a professor who’d interviewed her for the scholarship program she’s currently in. He was comparing the impact of her entire field to a terrorist attack. “Public health scientists need to be really careful about the language they’re using, communicate uncertainty clearly, and acknowledge when they’re not an expert,” she said. “I study vector-borne disease. So when we’re talking about Covid dynamics, that’s out of my wheelhouse.”
If it was out of her wheelhouse, it was far, far outside of Levitt’s. At first, his power made her hesitant to speak up. But when he appeared at the roundtable with Gov. Ron DeSantis of Florida, where she had at-risk family members, she felt she had to, for the safety of her loved ones, and wrote an op-ed for her student newspaper, the Stanford Daily.
At the heart of many critiques was the issue of uncertainty. Covid models aren’t oracles. They aren’t made to predict the precise number of deaths or an outbreak’s exact end-date, because each of those outcomes is constantly being reshaped by human behavior and policy. Instead, these models give us a hazy picture of the spectrum of possibilities: Scenario X might produce around Y deaths, but if we do A, that might instead bring us closer to Z. Use a model for decision-making — or don’t — and some of the potential results it laid out will not become realities.
To epidemiologists, Levitt’s graphs were frustrating because they didn’t show this garden of forking paths. In May 2020, he and a few colleagues observed that the data from a number of different Covid outbreaks formed what’s known as a Gompertz curve in each place — a shape they thought would describe the rise and fall of cases over time, for a single coronavirus wave. So he followed this curve for each new hotspot, using it to make predictions. But as Joel Miller, a mathematical modeler of infectious disease at La Trobe University, in Melbourne, Australia, explained, those forecasts don’t take into account the array of variables that could still influence the different possible futures of the curve.
Levitt isn’t unique in this regard. “A lot of models were orders of magnitude off,” said Ellen Kuhl, a Stanford professor of bioengineering, who herself switched from modeling brain damage to modeling Covid-19.
What set him apart was his platform, his avoidance of the usual ways findings get vetted — and his certainty.
When asked to respond to those who saw him as a source of misinformation, Levitt’s answer, like much of his messaging, was hard to pin down. He seemed, at different times, both impervious and highly sensitive to critique.
“Anyone who is accusing me of fabricating data, or being untrustworthy about data, will receive the full wrath of my position if it becomes known to me,” he said. (Some critics have suggested the data sources he uses aren’t always reliable, but there is no evidence of Levitt fabricating data.)
His tiff with Pachter produced a similar reaction. Initially, Pachter had emailed the International Society of Computational Biology fellows, including Levitt, saying that Levitt’s remarks were “dangerous” and that the association should release a statement “refuting” them. The president responded that the group wasn’t in the business of reviewing, rejecting, or endorsing anyone’s personal comments.
When Pachter followed up with another email calling Levitt (whom he once again cc’d) “clueless” about Covid, and then a critical blog post, Levitt’s response was less public: He privately wrote to others he knew at CalTech. Two faculty members confirmed this was the case. One said that Levitt had suggested Pachter’s funding be withdrawn. The other forwarded the email to Pachter, who forwarded it to STAT. In it, Levitt complained of Pachter’s “insulting Tweets … 129 at last count,” and his “very carefully researched report that is likely accurate in detail but totally misleading in outline.” And he asked whether he should write to the CalTech president and the board of trustees “to try to stop this persecution.” What can I do, he wrote, “that will be effective but not necessarily damage Lior Pachter’s career in an irretrievable way?”
Pachter has tweeted that Levitt reached out to the president of CalTech. (The president did not respond to requests for confirmation.) But Levitt says Pachter’s account is false: “I never did this. But maybe I should. But I won’t go to the president, I’ll go to the rich Jewish guy who is funding the institution. But I’m not going to do this. I don’t do these things. …” (Levitt is himself Jewish.)
At other moments, though, Levitt said he didn’t care about such critiques: “By and large the people who said I was doing wrong will be seen to have made fools of themselves. And that’s fine. Einstein was ridiculed for decades.”
“By and large the people who said I was doing wrong will be seen to have made fools of themselves. And that’s fine. Einstein was ridiculed for decades.”
In theory, Levitt likes the idea of being wrong. To him, it is a fundamental part of science: If you’re not wrong sometimes, you’re not taking enough risks, you’re not doing science right. He likes to say that if his nonscientist wife told him that the third page of his most recent paper — chosen at random — had some mistakes, he would scour it and sure enough would find something that could be done better. And he’s very open about errors he’s made, in both the content and tone of what he’s said.
“The biggest one was declaring, in a somewhat belligerent way, that things would end in the United States by the end of August. I feel really bad about that,” he said. “I don’t think scientists should make declarations like that. It was a pugilistic, a combative argument. People said to me I was wrong, and I thought, I’m not going to back down. And it’s stupid. It was not the right way to behave.”
But part of him is also defensive, insisting that more has been made of his mistakes than epidemiologists’. Such retorts don’t always hold up to scrutiny. In some cases, he seems to be projecting onto others his own model-as-oracle fallacy. He likes to point out that Neil Ferguson, of Imperial College London, “said that there was going to be 2.2 million deaths in the United States” — when in fact Ferguson’s paper had suggested that if no preventive actions were taken, there could be as many as 2.2 million deaths. The number is the same, but its meaning changes drastically. To accurately interpret the modeling is to express its uncertainty.
Lurking underneath these cases of he-said-she-said is a divide that affects the entire country — but is especially thorny for universities. When Crystal Lee, an MIT Ph.D. student in history and anthropology of science, started studying how the pandemic was understood across social media spheres, she was surprised to find that different social pockets with differing politics had their own data visualizations and analyses. They all professed trust in data, but depending on how you sliced it, a graph could be found that backed almost any conclusion or policy. “There did not seem to be a shared reality,” she said.
You can see that even in Levitt’s assertion that none of his critics is truly looking at his numbers. He says Ferguson has refused to engage with him, though Ferguson and his colleagues responded to a number of his emails in March 2020, pointing out again and again the specific equation where they saw his mortality calculations going astray for the Diamond Princess cruise ship. Mostly, Levitt just heard rejection: “These people had the stage, I was the outsider, they were advising the British government. Engaging is more than just saying, ‘You’re wrong,’ in my opinion.”
It’s easy to imagine an actual, more public rejection — Stanford officially distancing itself from his statements, for instance — heightening that sense of exclusion and so entrenching his belief that his work is being dismissed cursorily. When a biodesign conference, organized in part by Stanford, did cancel Levitt’s keynote because of his questionable Covid claims, he tweeted, “New Dark Age Cometh,” and his followers made the predictable, Trumpian comparison to witch trials. His talk was later un-canceled, and took place as initially planned.
How to respond is both practically and philosophically fraught. For Levitt’s colleagues at Stanford, the question is uncomfortable. None of those interviewed wanted the university censoring professors; everyone holds academic freedom sacred. Then again, they don’t necessarily think that the right use of an academic position is as a megaphone to talk over public health guidance, especially during an emergency.
Julie Parsonnet, a Stanford infectious disease physician and epidemiologist who also advised Placer County, emphasized the pressure that cash-strapped public health departments are under. There is room for academic debate that doesn’t undermine their messaging. She didn’t want to discuss the work of specific Stanford colleagues, but she wishes that there had been a sense of shared purpose, everyone against Covid: “What we should have been seeing all along is people going to public health departments and saying, ‘How can we help you make the best decisions that you can?’ not, ‘Ah, the public health department is not supporting my model, and my model is the best because I am the best.’”
To her, there is a line between expressions of academic freedom and inappropriate uses of one’s professorship. She’s utterly ignorant about astrophysics, she said. “If I came out and I stood in front of a big S for Stanford and said, as a Stanford faculty member, ‘the Big Bang never happened,’ I think I should be chastised by my institution for that.”
Yet policing that line is a tricky proposition.
“Academics, clinicians, public health experts, science communicators — we want to empower them to counter misinformation, right?” said Timothy Caulfield, a Canada research chair in health law and policy at the University of Alberta, who studies misinformation and pseudoscience. But there’s an issue when the unreliable source is within the university itself: “You want to create these support mechanisms for those that want to be out there in the public sphere. But those same rules would apply to someone like Michael Levitt.”
In the case of Scott Atlas, only after around 100 faculty members signed an open letter about his falsehoods, and after Atlas himself encouraged Michiganders to “rise up” against the state’s Covid restrictions, did university officials speak out, saying his comments were “inconsistent with the university’s approach” to the pandemic. When asked about its response to Levitt’s statements on Covid, Stanford sent a very neutral statement: “Stanford Medicine supports the freedom of faculty to voice their position. We do not, however, endorse any individual views expressed by faculty, and their views do not represent an institutional position.”
Caulfield thinks institutions are going to have to start making hard choices, not stepping on individual professors’ freedom of expression, but being more careful about supporting those messages that are bolstered by a body of evidence.
The task is complex. It doesn’t mean blocking those whose work is at the intersection of different fields. After all, part of the reason that it took so long for the aerosol transmission of Covid to be accepted was that the engineers and physicists who saw the evidence were seen as epistemic trespassers. “The danger is that you’re going to have overzealous policing of members of a profession,” said Nathan Ballantyne, the Fordham University philosophy professor who coined the term “epistemic trespasser,” which to him, is someone who barges into another field without humility.
But going too far in the other direction holds a danger, too. As Joel Miller, at La Trobe University, put it, “If, in the next 10 years, there’s a new large epidemic, I worry somebody will take a Gompertz curve and say, ‘Look, the epidemic will go away very quickly and we don’t need to do anything, we should ignore it.’”