Identifying "Hype" in Biomedical Research: A Linguistics and AI Approach

The scientific landscape is increasingly competitive, leading researchers to use "hype"—promotional language that can exaggerate findings and influence perceptions. This phenomenon can undermine objective evaluation of research, potentially misdirecting funding and eroding trust in science.

Our project, combining linguistics and artificial intelligence, aims to address this by:

• Identifying Hype Trends: We observe a significant increase in promotional language in abstracts of successful NIH grant applications and subsequent PubMed publications (1985-2020). This includes a shift towards more confident and assertive language, with terms expressing weak possibility declining and terms conveying certainty and objective verifiability rising.
• Analyzing Influences: The use of hype in grant applications is strongly correlated with its appearance in NIH funding opportunity announcements, suggesting that institutional guidelines may inadvertently encourage this trend. This creates a "path dependency" where initial language choices for funding can become entrenched in later research reporting.
• Developing Detection Tools: We are formalizing annotation guidelines and training machine learning models to automatically identify hype in scientific texts, building on prior work that defined hype as "hyperbolic and/or subjective language that authors use to glamorize, promote, embellish and/or exaggerate aspects of their research".

By understanding and detecting hype, our project seeks to foster greater transparency and objectivity in scientific communication, contributing to a more robust and trustworthy research ecosystem.