Consensus in Science: Is It Necessary?

An important but often misunderstood concept in science is the role of consensus. Some scientists argue that consensus has no place at all in science, that the scientific method alone with its emphasis on evidence and logic dictates whether a particular hypothesis stands or falls.  But the eventual elevation of a hypothesis to a widely accepted theory, such as the theory of evolution or the theory of plate tectonics, does depend on a consensus being reached among the scientific community.

In politics, consensus democracy refers to a consensual decision-making process by the members of a legislature – in contrast to traditional majority rule, in which minority opinions can be ignored by the majority. In science, consensus has long been more like majority rule, but based on facts or empirical evidence rather than personal convictions. Although observational evidence is sometimes open to interpretation, it was the attempt to redefine scientific consensus in the mold of consensus democracy that triggered a reaction to using the term in science.

This reaction was eloquently summarized by medical doctor and Jurassic Park author Michael Crichton, in a 2003 Caltech lecture titled “Aliens Cause GlobaL Warming”:

“I want to pause here and talk about this notion of consensus, and the rise of what has been called consensus science. I regard consensus science as an extremely pernicious development that ought to be stopped cold in its tracks. Historically, the claim of consensus has been the first refuge of scoundrels; it is a way to avoid debate by claiming that the matter is already settled. …

Let’s be clear: the work of science has nothing whatever to do with consensus. Consensus is the business of politics. Science, on the contrary, requires only one investigator who happens to be right, which means that he or she has results that are verifiable by reference to the real world.

In science consensus is irrelevant. What is relevant is reproducible results. … There is no such thing as consensus science. If it’s consensus, it isn’t science. If it’s science, it isn’t consensus.”

What Crichton was talking about, I think, was the consensus democracy sense of the word – consensus forming the basis for legislation, for political action. But that’s not the same as scientific consensus, which can never be reached by taking a poll of scientists. Rather, a scientific consensus is built by the slow accumulation of unambiguous pieces of empirical evidence, until the collective evidence is strong enough to become a theory.

Indeed, the U.S. AAAS (American Association for the Advancement of Science) and NAS (National Academy of Sciences, Engineering and Medicine) both define a scientific theory in such terms. According to the NAS, for example,

 “The formal scientific definition of theory …  refers to a comprehensive explanation of some aspect of nature that is supported by a vast body of evidence.”

Contrary to popular opinion, theories rank highest in the scientific hierarchy – above laws, hypotheses and facts or observations. 

Crichton’s reactionary view of consensus as out of place in the scientific world has been voiced in the political sphere as well. Twentieth-century UK prime minister Margaret Thatcher once made the comment, echoing Crichton’s words, that political consensus was “the process of abandoning all beliefs, principles, values and policies in search of something in which no one believes, but to which no one objects; the process of avoiding the very issues that have to be solved, merely because you cannot get agreement on the way ahead.” Thatcher was a firm believer in majority rule.

A well-known scientist who shares Crichton’s opinion of scientific consensus is James Lovelock, ecologist and propounder of the Gaia hypothesis that the earth and its biosphere are a living organism. Lovelock has said of consensus:

“I know that such a word has no place in the lexicon of science; it is a good and useful word, but it belongs to the world of politics and the courtroom, where reaching a consensus is a way of solving human differences.”

But as discussed above, there is a role for consensus in science. The notion articulated by Crichton and Lovelock that consensus is irrelevant has arisen in response to the modern-day politicization of science. One element of their proclamations does apply, however. As pointed out by astrophysicist and author Ethan Siegel, the existence of a scientific consensus doesn’t mean that the “science is settled.” Consensus is merely the starting point on the way to a full-fledged theory.

Next week: How Elizabeth Holmes Abused Science to Deceive Investors

Corruption of Science: Scientific Fraud

One of the most troubling signs of the attack on science is the rising incidence of outright fraud, in the form of falsification and even fabrication of scientific data. A 2012 study published by the U.S. National Academy of Sciences noted an increase of almost 10 times since 1975 in the percentage of biomedical research articles retracted because of fraud. Although the current percentage retracted due to fraud was still very small at approximately 0.01%, the study authors remarked that this underestimated the actual percentage of fraudulent articles, since only a fraction of such articles are retracted.

One of the more egregious episodes of fraud was British gastroenterologist Andrew Wakefield’s claim in a 1998 study that 8 out of 12 children in the study had developed symptoms of autism after injection of the combination MMR (measles-mumps-rubella) vaccine. As a result of the well publicized study, hundreds of thousands of parents who had conscientiously followed immunization schedules in the past panicked and began declining MMR vaccine. And, unsurprisingly, outbreaks of measles subsequently occurred all over the world.

But Wakefield’s paper was slowly discredited over the next 12 years, until the prestigious medical journal The Lancet formally retracted it. The journal’s editors then went one step further in 2011 by declaring the paper fraudulent, citing unmistakable evidence that Wakefield had fabricated his data on autism and the MMR vaccine. Shortly after, the disgraced gastroenterologist’s medical license was revoked.

In 2015, Iowa State University researcher Dong Pyou Han received a prison sentence of four and a half years and was ordered to repay $7.2 million in grant funds, after being convicted of fabricating and falsifying data in trials of a potential HIV vaccine.  On multiple occasions, Han had mixed blood samples from vaccinated rabbits into human HIV antibodies to create the illusion that the vaccine boosted immunity against HIV. Although Han was contrite in court, one of the prosecuting attorneys doubted his remorse, pointing out that Han’s job depended on research funding that was only renewed as a result of his bogus presentations showing the experiments were succeeding.

In 2018, officials at Harvard Medical School and Brigham and Women’s Hospital in Boston called for the retraction of a staggering 31 papers from the laboratory of once prominent Italian heart researcher Piero Anversa, because the papers "included falsified and/or fabricated data." Dr. Anversa’s research was based on the notion that the heart contains stem cells, a type of cell capable of transforming into other cells, that could regenerate cardiac muscle. But other laboratories couldn’t verify Anversa’s idea and were unable to reproduce his experimental findings – a major red flag, since replication of scientific data is a crucial part of the scientific method.

Despite this warning sign, the work spawned new companies claiming that their stem-cell injections could heal hearts damaged by a heart attack, and led to a clinical trial funded by the U.S. National Heart, Lung and Blood Institute. The Boston hospital’s parent company, however, agreed in 2017 to a $10 million settlement with the U.S. government over allegations that the published research of Anversa and two colleagues had been used to fraudulently obtain federal funding. Apart from data that the lab fabricated, the government alleged that it utilized invalid and improperly characterized cardiac stem cells, and maintained deliberately misleading records. Anversa has since left the medical school and hospital.

Scientific fraud today extends even to the publishing world. A recent sting operation involved so-called predatory journals – those charging a fee without offering any publication services (such as peer review), other than publication itself. The investigation found that an amazing 33% of the journals contacted offered a phony scientific editor a position on their editorial boards, four of them immediately appointing the fake scientist as editor-in-chief.   

It’s no wonder then that scientific fraud is escalating. In-depth discussion of recent cases can be found on several websites, such as For Better Science and Retraction Watch.

Next week: Consensus in Science: Is It Necessary?

Corruption of Science: The Reproducibility Crisis

One of the more obvious signs that modern science is ailing is the reproducibility crisis – the vast number of peer-reviewed scientific studies that can’t be replicated in subsequent investigations and whose findings turn out to be false. In the field of cancer biology, for example, researchers discovered that an alarming 89% of published results couldn’t be reproduced. Even in the so-called soft science of psychology, the rate of irreproducibility hovers around 60%. And to make matters worse, falsification and outright fabrication of scientific data is on the rise.

The reproducibility crisis is drawing a lot of attention from scientists and nonscientists alike. In 2018, the U.S. NAS (the National Association of Scholars in this case, not the Academy of Sciences), an academic watchdog organization that normally focuses on the liberal arts and education policy, published a particularly comprehensive examination of the problem. Although the emphasis in the NAS report is on the misuse of statistical methods in scientific research, the report discusses possible causes of irreproducibility and presents a laundry list of recommendations for addressing the crisis.

The crisis is especially acute in the biomedical sciences. Over 10 years ago, Greek medical researcher John Ioannidis argued that the majority of published research findings in medicine were wrong. This included epidemiological studies in areas such as dietary fat, vaccination and GMO foods as well as clinical trials and cutting-edge research in molecular biology. 

In 2011, a team at Bayer HealthCare in Germany reported that only about 25% of published preclinical studies on potential new drugs could be validated. Some of the unreproducible papers had catalyzed entirely new fields of research, generating hundreds of secondary publications. More worryingly, other papers had led to clinical trials that were unlikely to be of any benefit to the participants.

Author Richard Harris describes another disturbing example, of research on breast cancer that was conducted on misidentified skin cancer cells. The sloppiness resulted in thousands of papers being published in prominent medical journals on the wrong cancer. Harris blames the sorry condition of current research on scientists taking shortcuts around the once venerated scientific method.

Cutting corners to pursue short-term success is but one consequence of the pressures experienced by today’s scientists. These pressures include the constant need to win research grants as well as to publish research results in high-impact journals. The more spectacular that a paper submitted for publication is, the more likely it is to be accepted, but often at the cost of research quality. It has become more important to be the first to publish or to present sensational findings than to be correct.      

Another consequence of the bind in which scientists find themselves is the ever increasing degree of misunderstanding and misuse of statistics, as detailed in the NAS report. Among other abuses, the report cites spurious correlations in data that researchers claim to be “statistically significant”; the improper use of statistics due to poor understanding of statistical methodology; and the conscious or unconscious biasing of data to fit preconceived ideas.

Ioannidis links irreproducibility to the habit of assigning too much importance to the statistical p-value. The smaller the p-value, the more likely it is that the experimental data can’t be explained by existing theory and that a new hypothesis is needed. Although p-values below 0.05 are commonly regarded as statistically significant, using this condition as a criterion for publication means that one time in twenty, the experimental data could be the result of chance alone. The NAS report recommends defining statistical significance as a p-value less than 0.01 rather than 0.05 – a much more demanding standard.

The report further recommends integration of basic statistics into curricula at high-school and college levels, and rigorous educational programs in those disciplines that rely heavily on statistics. Beyond statistics, other suggested reforms include having researchers make their data available for public inspection, which doesn’t often occur at present, and encouraging government agencies to fund projects designed purely to replicate earlier research, which again is rare today. The NAS believes that measures like these will help to improve reproducibility in scientific studies as well as keeping advocacy and the politicization of science at bay.

Next week: Corruption of Science: Scientific Fraud

Should We Fear Low-Dose Radiation? What Science Says

Modern science is constantly under attack from political forces, often fueled by fear. A big fear is radiation exposure – a fear made only too real by the devastation of the atomic bombs dropped on Japan to end World War II, and the aftereffects of several extensive nuclear accidents around the world in the last few decades. But, while high doses of radiation are known to be harmful to human health or even deadly, the effects of low doses are controversial.

For many years, the prevailing wisdom in the scientific community about radiation protection has been that there is no safe dose of ionizing radiation. This belief is enshrined in the so-called LNT (linear, no-threshold) model used to estimate cancer risks and establish cleanup levels in radioactively contaminated environments. The model dates back to studies of irradiated fruit flies in the 1930s, and subsequent formulation of the LNT dose-response model by American geneticist and Nobel laureate Hermann Muller.

The LNT model assumes that the body’s response to radiation is directly proportional to the radiation dose. So any detrimental health effects – such as cancer or an inheritable genetic mutation – go up and down with dose (and dose rate), but don’t disappear altogether until the dose falls to zero.

A very different concept that is gaining acceptance among radiation workers is the threshold model. Unlike the LNT model, this assumes that exposure to radiation is safe as long as the exposure is below a threshold dose. That is, there are no adverse health effects at all at low radiation doses, but above the threshold there are effects proportional to the dose, as in the no-threshold model.  

A new variation on the threshold model is hormesis, which hypothesizes that below the threshold dose, beneficial health effects actually occur. Hormesis has been championed by Edward Calabrese, an environmental toxicologist at the University of Massachusetts Amherst who has long been critical of the LNT approach to risk assessment, for both radiation and toxic chemicals. In 2015, a petition was submitted to the U.S. NRC (Nuclear Regulatory Commission) to adopt the hormesis model for regulatory purposes.

Which model is the correct picture of how the human body is affected by radiation? The scientific evidence isn’t all that clear.

Even when the LNT model was proposed, only very limited data was available at low doses, a situation that’s unchanged today. This means that the statistical accuracy of individual data points at low doses is poor, and much of the data could equally well fit the LNT, threshold or hormesis models. Two major pieces of evidence that a U.S. NAS (National Academy of Sciences) committee formerly relied on to buttress the LNT model – a study of Japanese atomic bomb survivors and a 15-country study of nuclear workers – are in fact compatible with either the threshold or the LNT model, more recent analysis has shown.   

The threshold model may seem more intuitive, since it’s well known for chemical toxins that any substance is toxic above a certain dose. “The dose makes the poisin,” as medieval Swiss physician Paracelsus observed. But the biological response to radiation isn’t necessarily the same as the response to a toxin.

Evidence in support of the hormesis model, however, includes numerous studies showing that low radiation doses can activate the immune system and thereby protect health. And no increase in the incidence of cancer has been observed among those Japanese bomb survivors exposed to only low doses of the same radiation that, in higher doses, sickened or killed others.

Scientific opinion is divided. The once strong consensus on the validity of the LNT model has evaporated, 70% of scientists at U.S. national laboratories now believing that the threshold model more accurately reflects radiation effects. A similar percentage of scientists in several European countries hold the same view.

Whether or not low doses of radiation are protective, as the hormesis model suggests, no adverse health effects have ever been detected from exposure to low dose, low dose rate radiation. But the public clings to the outmoded scientific consensus of the LNT model that no dose is safe. So society at large is unnecessarily fearful of any exposure to radiation whatsoever, when in reality low doses are most likely benign and could even be beneficial.

Next: Corruption of Science: The Reproducibility Crisis