Diverse Languages of Innovation: Gender Differences in Patent Value Expressions

Abstract

This study examines whether and how the gender composition of inventor teams is associated with differences in the articulation of innovation value in patents. Moving beyond established evidence on gender disparities in patenting volume, we focus on value expressions (VEs)—statements that frame inventions in terms of societal (public value expressions, PVE), commercial (private value expressions, PRIVE), or both. We use a dataset of 170,864 USPTO patents in artificial intelligence and nanotechnology filed between 2005 and 2023, containing approximately 7.1 million sentences from patent abstracts, backgrounds, and summaries. A two-stage classifier based on GPT-4o and RoBERTa-Large is applied to categorize VEs at scale. We then compare patterns between all-male and female-inclusive inventor teams. Our findings reveal that patents involving female inventors are more likely to include public value expressions and less likely to rely exclusively on private value framing, both in terms of presence and textual density. These differences are moderated by institutional and disciplinary contexts: the gender gap in public value framing is more pronounced in university settings and in nanotechnology, while it is attenuated in non-university settings and AI. Over time, gender differences in value framing have narrowed, suggesting increasing convergence in discursive expectations around innovation. By conceptualizing patent texts as sites of value articulation rather than just technical description, this study contributes to research on gender and innovation, institutional logics, and the evaluation of technological impact.