Researchers say more stringent reviews, and AI, should be used to help vet scientific literature -- especially in situations like a pandemic, when studies are being published as fast as possible. Photo by sabinevanerp
Sept. 11 (UPI) -- The rush to conduct and publish research in the midst of the COVID-19 pandemic has led to what scientists are calling an "infodemic" -- an overwhelming flow of scientific literature, including flawed and contradictory findings.
To make sense of all the new information, and to ensure the most accurate and useful information isn't drowned out by less discerning research, the authors of a new paper -- published Friday in the journal Patterns -- argue for both more stringent publishing standards and the use of artificial intelligence.
According to the paper's lead author, Ganesh Mani, an investor, technology entrepreneur and adjunct faculty member in Carnegie Mellon University's Institute for Software Research, accuracy shouldn't be sacrificed for speed.
"Science is a process and should be a more deliberate process," Mani told UPI in an email. "Reviewers should get more career credit and that may help with regards to how they prioritize reviews among their other responsibilities."
"Work needs to be done all around -- but especially in media circles -- to show prominently the more accurate, current, authoritative 'facts' and research over the most popular ones, which is what's often bubbling to the top these days," Mani said.
In the months following the outbreak of the COVID-19 pandemic, scientific journals have published more than 8,000 preprints of scientific papers related to the novel coronavirus. During that time, the average time for peer review has decreased from an average of 117 to 60 days.
As a result, the public, doctors and public health officials have been overwhelmed by an abundance of scientific information.
Mani and co-author Tom Hope, post-doctoral researcher at the Allen Institute for Artificial Intelligence, suggest machine learning and computer algorithms can be used to filter out less reliable information. Additionally, AI could link together and synthesize the results of similar studies, making the outflow of information during the next pandemic -- or similar global challenge -- more reliable and digestible.
Mani suggests the infodemic has revealed a variety of problems related to accuracy and reliability.
"Small cohort studies need particular attention," he said. "Folks should be aware that results might change or be subsumed by larger cohort studies -- or when a more representative population is used. Often there are multiple interpretations of data, especially as it's emerging. Communicating the multiple interpretations can be tricky."
Problems with scientific review and publication aren't new, but Mani and Hope suggest the pandemic has exacerbated the situation.
Artificial intelligence can help survey and organize information, but according to Mani, humans must work together to balance the need for speed and accuracy during an infodemic.
A failure to do so, Mani argues, will result in short- and long-term consequences.
"Poor quotidian decisions at the individual level with regards to health, wealth and wisdom," he said. "Institutions may make longer-term investment and policy decisions that are suboptimal and counterproductive."