Starting with “fake news” and continuing through a harrowing pandemic spanning more than 2 years, there has been an unprecedented politicization of science. Terms such as disinformation, credible source, sheep and alarmist have become ubiquitous.

Now, the first study to examine why Americans are downloading open-access scientific articles revealed a strong demand among ordinary people for the highest quality information. It also reinforced the need for open access journals for all.

Researchers at the Georgia Tech School of Public Policy focused on downloads of National Academies of Sciences, Engineering, and Medicine (NASEM) consensus reports, widely considered some of the most credible scientific literature. For their study, the team looked at 1.6 million comments left on 6.6 million downloads of NASEM consensus reports since 2011, the first year the academies offered the content for free. The comments were left in response to a prompt asking users how they plan to use the reports.

According to a machine learning (ML) algorithm applied to comments, while 48% of reports were downloaded for academic purposes, even more were viewed by people outside strictly educational settings. In fact, BERT – the ML algorithm – noted the use of the word “edification” 3,700 times in the dataset, which the researchers say “indicates a strong desire for lifelong learning. life among users”.

The 52% of non-academic uploaders included radio amateurs, amateur astronomers, lifelong learning providers and retirees wishing to keep up. About 150,000 downloads were categorized as having to do with “personal use,” including topics such as cannabis, death, genetically modified crops, evolution versus creationism, and gun violence reduction. Over 25,000 doctors and nurses have downloaded reports with plans to use the details to improve their clinical work.

Interestingly, the analysis revealed that thousands of veterans had downloaded NASEM reports to use in their disability claims with the United States Veterans Administration.

“This research will hopefully raise awareness of the positive returns that accrue to society from investments in institutions that democratize public access to high-quality research,” said co-author Ameet Doshi, Ph.D. from the School of Public Policy. student and director of the Donald E. Stokes Library at Princeton University.

The use of BERT, the ML algorithm, comes from the lab of Omar Asensio, study co-author and assistant professor at Georgia Tech. Asensio’s Data Science and Policy Lab has used deep learning techniques in recent years to advance knowledge in energy efficiency, sustainable plastics and electric vehicle charging infrastructure. He says using BERT to rank behavioral evidence regarding public interest in scientific information expands and advances the use of machine learning in the social sciences.

“When you get data at this scale, especially when you have unstructured data that grows in real time, there are practical limitations as to why this kind of behavioral insights weren’t known before” , said Asensio. “We show in a number of research areas that experimental approaches to curating human-labeled training data can improve the performance of popular supervised ML algorithms, to a level that can match or even exceed human performance. .”

According to the study, the algorithm was able to identify the correct meaning of comments about 84% of the time.

“This study shows a strong demand among ordinary Americans for the highest quality information to help improve the work they do, to help their loved ones, neighbors and communities, and in some cases simply to learn to learn. “said co-writer Diana. Hicks, a professor at the Georgia Tech School of Public Policy. “We never hear these stories because everyone’s focused on all the misinformation circulating on social media.”