Who believes in science? A computational tool for identifying language invoking or disputing scientific knowledge

From disputes over COVID-19 to contestation over climate science, questions of the value and legitimacy of mainstream scientific knowledge have become matters of high-stakes political struggle. Here, we introduce a novel computational tool to identify and track the emergence, proliferation, and historical variations of discourses that either seek to invoke the authority of scientific expertise or to criticize scientific claims, institutions, and experts. We describe the tool's development, demonstrate its predictive and convergent validity, and illustrate its potential across three case studies shedding light on how elite rhetoric may drive political polarization around science in the United States. Among other findings, we find that political statements invoking scientific expertise have historically been more likely to receive coverage in mainstream American newspapers than statements that do not invoke expertise. However, this apparent advantage disappears over time, suggesting the discursive authority associated with the invocation of scientific methods, credentials, and institutions may be diminishing.