Why it Matters: Science and technology policy has become increasingly concerned with broader impacts, coupled with a proliferation of ethical language for scientists to demonstrate their commitment to societal benefits. This manifests in various ways: through broader impact statements in grant applications, value-laden language in papers, Ethical, Legal, and Social Implications (ELSI) studies within technology initiatives, and ethical guidelines to govern disruptive technologies. However, a critical question has arisen: are these claims of societal benefit genuine or merely rhetorical? This has significant implications for how we evaluate and develop technologies – are we steering innovation toward social good, or are we just talking about it?
Our Approach: We developed a new method combining large language models with patent analysis to investigate this question. Using both generative (GPT-4o) and discriminative (RoBERTa Large) models, we analyzed 175,730 USPTO patents in the fields of AI and nanotechnology to identify and classify value expressions – statements about societal or commercial implications (N=~7.1MM). We then correlated these expressions with the patents’ technological orientation, determined through their mapping from technology classes to the UN Sustainable Development Goals. Our findings reveal that patents containing societal benefit promises are more likely to be oriented toward addressing social challenges based on their objective technological features. This suggests that such promises go beyond mere rhetoric, reflecting an underlying capacity to address societal issues.
Publications
Sergio Paleaz, Diana Hicks (2024) Are Societal Promises in Science and Technology Substantiated? The Case of Value Expressions in Patents, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4941820.
Pelaez, S., Verma, G., Ribeiro, B., & Shapira, P. (2023). Large-Scale Text Analysis Using Generative Language Models: A Case Study in Discovering Public Value Expressions in AI Patents. Preprint: arXiv:2305.10383; Peer reviewed: Quantitative Science Studies, 2024, https://doi.org/10.1162/qss_a_00285 (February 1, 2024). [Open Access]
A literature review on Artificial Intelligence and Public Values is also available at https://osf.io/preprints/socarxiv/yq57a/
Research Team
- Philip Shapira
- Barbara Ribeiro, SKEMA Business School, France
- Sergio Pelaez (GT School of Public Policy, PhD Program)
- Gaurav Verma (GT School of Computational Science and Engineering)
Alumni (GT Undergraduate Research Assistants, 2022-2023.
- Christine Webster
- Divali Legore