This study evaluates how training on eight criteria for recognizing fake news impacts participants’ ability to classify news accurately and how their epistemological beliefs affect this process. It compares participants’ classifications with those of the GPT-based “News Verifier” tool. The objectives are to assess classification accuracy, examine the influence of epistemological beliefs (absolutist, multiplists, evaluativist), and determine if training improves alignment with the “News Verifier”. This approach tests the training’s effectiveness and explores how personal cognitive predispositions influence information processing in technology-mediated learning contexts.

The study included 70 voluntary participants enrolled in a professional development course for high school teaching certification, some already teachers. The initial training phase included a webinar explaining the eight criteria for recognizing fake news and introducing epistemological thinking concepts, accompanied by educational materials detailing these criteria and concepts. Participants completed a questionnaire to classify them into one of three epistemological positions: absolutist, multiplists, or evaluativist. They were then presented with four news items and asked to classify them into categories such as Misinformation, Unproven, Sensationalist, Completely False, Misleading, or True. The GPT “News Verifier”, trained on the eight verification criteria, was used to classify the same four news items.

Data collection involved comparing participants’ classifications with those of the GPT “News Verifier”, analyzing the differences and similarities between the two sets of classifications, and examining the influence of participants’ epistemological beliefs on their classification accuracy using statistical analyses. The results suggested that for “News 1” and “News 2”, there was no significant difference between the participants’ classifications and those of GPT “News Verifier”. However, for “News 3” and “News 4”, the differences were statistically significant.

The results revealed that the dominant epistemological profile among participants was multiplists, suggesting a more pluralistic and inclusive approach to teaching and evaluation, while the evaluativist profile indicated a more critically evaluative approach and the absolutist profile suggested a more rigid and traditional approach. The distribution of classifications showed that most news was categorized as “Unproven” and “Sensationalist”, followed by “Misinformation”, with the least common categories being “Completely False”, “Misleading”, and “True.” Profile-based classification revealed that the absolutist profile had the fewest classifications, while the multiplists had the majority, and the evaluativist profile had an intermediate number of classifications with a tendency towards categorizing news as “Unproven” and “Sensationalist.”

The study provides insights into how training on the eight criteria affects the ability to identify fake news and highlights the significant influence of participants’ epistemological beliefs on their classification accuracy. The findings suggest that training programs for recognizing fake news can be improved by considering participants’ epistemological beliefs to optimize educational strategies, offering valuable guidance for developing effective educational interventions to enhance news literacy in the digital age.

Keywords: Fake News, Epistemological Beliefs, Teacher Professional Development, ChatGPT, News Classification.