From the University of Washington:
Carl Bergstrom believes facts stand a fighting chance, especially if science has their back. A professor of biology at the University of Washington, he has used mathematical modeling to investigate the practice of science, and how science could be shaped by the biases and incentives inherent to human institutions.
“Science is a process of revealing facts through experimentation,” said Bergstrom. “But science is also a human endeavor, built on human institutions. Scientists seek status and respond to incentives just like anyone else does. So it is worth asking — with precise, answerable questions — if, when and how these incentives affect the practice of science.”
In an article published Dec. 20 in the journal eLife, Bergstrom and co-authors present a mathematical model that explores whether “publication bias” — the tendency of journals to publish mostly positive experimental results — influences how scientists canonize facts. Their results offer a warning that sharing positive results comes with the risk that a false claim could be canonized as fact. But their findings also offer hope by suggesting that simple changes to publication practices can minimize the risk of false canonization.
These issues have become particularly relevant over the past decade, as prominent articleshave questioned the reproducibility of scientific experiments — a hallmark of validity for discoveries made using the scientific method. But neither Bergstrom nor most of the scientists engaged in these debates are questioning the validity of heavily studied and thoroughly demonstrated scientific truths, such as evolution, anthropogenic climate change or the general safety of vaccination.
For their model, Bergstrom’s team incorporated variables such as the rates of error in experiments, how much evidence is needed to canonize a claim as fact and the frequency with which negative results are published. Their mathematical model showed that the lower the publication rate is for negative results, the higher the risk for false canonization. And according to their model, one possible solution — raising the bar for canonization — didn’t help alleviate this risk.
“It turns out that requiring more evidence before canonizing a claim as fact did not help,” said Bergstrom. “Instead, our model showed that you need to publish more negative results — at least more than we probably are now.”
Direct to Full Text Research Article
Science is facing a “replication crisis” in which many experimental findings cannot be replicated and are likely to be false. Does this imply that many scientific facts are false as well? To find out, we explore the process by which a claim becomes fact. We model the community’s confidence in a claim as a Markov process with successive published results shifting the degree of belief. Publication bias in favor of positive findings influences the distribution of published results. We find that unless a sufficient fraction of negative results are published, false claims frequently can become canonized as fact. Data-dredging, p-hacking, and similar behaviors exacerbate the problem. Should negative results become easier to publish as a claim approaches acceptance as a fact, however, true and false claims would be more readily distinguished. To the degree that the model reflects the real world, there may be serious concerns about the validity of purported facts in some disciplines.