Scientific Accountability: Detecting Salient Features of Retracted Articles

The dissemination of misinformation can have detrimental effects on society by promoting false narratives.Similarly, retracted articles can result in the propagation of inaccuracies in scientific information.Despite the formal retraction of these articles, they continue to be cited in scientific literature and online, thereby distorting public understanding and undermining trust in science.Currently, there is no reliable method for identifying articles that require further re-evaluation or even retraction.Moreover, the existing methods for assessing the impact of retracted articles on subsequent research rely primarily on explicit dependencies stated in citation text.However, these dependencies are often implicitly defined, complicating the assessment of their true impact on subsequent research.Given these limitations, our objective is to address two challenges: (i) to identify salient characteristics of retracted articles that can distinguish them from non-retracted articles, based on textual analysis of sections such as the Abstract, Methodology, Results, and Conclusion, as well as their measurable readability and certainty characteristics, which reflect the degree of confidence with which authors report their findings; and (ii) to identify implicit dependencies of retracted articles on subsequent studies by examining experimental protocols used in both citing and cited (retracted) articles.This study aims to provide a proactive approach for flagging potential retractions and their impact on subsequent studies, thereby enhancing scientific integrity. CCS Concepts Information systems Content analysis and feature selection.