CWTS Seminar
Do societal promises influence patent value? An analysis of inventions in artificial intelligence
- Sergio Pelaez
- Date
- Friday 7 June 2024
- Time
- Location
- Online
Download the recording of this seminar
Abstract
This paper examines the relationship between public value expressions (PVEs) - defined as narrative statements that describe societal problems or solutions to these problems as offered by the invention being patented - and established indicators of patent value in the context of US artificial intelligence (AI) patents. Using a two-step process that relies on a large language model (LLM) to create a labeled training set and BERT-type models for training and prediction purposes, we identify and categorize PVEs across a dataset of over 154,000 AI patents obtained from Innovation Q+ and the USPTO's PatentsView. Topic modeling reveals the key areas of societal impact emphasized in PVEs, such as equity, justice, transparency, accountability, healthcare, data security and privacy. Count data regression analysis demonstrates a significant positive correlation between the occurrence and density of PVEs and patent value indicators, including forward citations, patent family size, number of claims, and number of technology classes, after controlling for confounding factors. We theorize that PVEs, as part of a strategic patent drafting practice we term 'societal-value embedding', serve a dual function. Normatively, they respond to increasing institutional pressures for societal accountability in science, technology, and innovation. Instrumentally, they act as signals of an invention's relevance and broaden its potential scope of application. Our findings suggest that private and public dimensions of value are fundamentally intertwined in AI patents.
Presenter
Sergio Pelaez, doctoral student in Public Policy at Georgia Tech
LinkedIn: https://www.linkedin.com/in/sergio-pelaez-148690147