The accuracy of press releases plays a critical role in the public’s understanding of science and public trust in scientific research. However, several types of exaggeration have been found in press releases, such as making causal claims from correlational findings and inference to humans from animal studies. These exaggerations cause misinformation to be spread to the public through both social media and the mainstream media. This project furthers our understanding of how scholarly research findings may be miscommunicated to policymakers and the public by using state-of-the-art techniques in machine learning and artificial intelligence. This project provides an in-depth understanding at both micro- and macro-levels. On the micro-level, the project examines precisely when and where exaggeration occurs. On the macro-level, the project will examine the prevalence of exaggeration over time and across domains and institutions. These findings will provide useful insights for improving the quality of press releases as an important channel for science communication.
This research will provide a large-scale, nuanced understanding of exaggerations in press releases issued by research institutions. The project develops a NLP and deep learning tool called PreCheck, which will automatically checking press release exaggeration, Specifically, PreCheck will link the press releases in EurekAlert!, the major academic press release portal, to the original research articles in PubMed, and then use machine learning and NLP techniques to examine each press release regarding two major types of exaggeration: causal claims of correlational findings and inference to humans from animal studies. Once exaggerations can be identified at the individual press release level, aggregated analyses will be done to answer research questions regarding the status and trend of press release exaggeration and its relationship with research areas and research institutions. The study results will be shared with press officers from various research institutions to discuss the research findings and their implications for communicating scientific research through press releases. In addition, PreCheck will be made publicly available so that science communicators can perform their own communication and science consumers can have a way to measure potential exaggeration in research.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
