On September 10, 2020, Iowa’s Innovation, Business & Law Center hosted Professor W. Keith Robinson for a presentation on the accessibility of the patent system. Professor Robinson is an Associate Professor of Law, Altshuler Distinguished Teaching Professor, and Co-Director of the Tsai Center for Law, Science and Innovation at the Southern Methodist University Dedman School of Law. His presentation was the first talk in IBL's Fall 2020 Speaker Series: Examining Institutional Structures: Race, Business, and the Law.

The patent system is not accessible 

Professor Robinson began by asserting that the current US patent system is not accessible. He supported his assertion with a 2016 study exploring the participation of women and minorities in the patent system. This study showed that less than 8% of US-born innovators are minorities. The number of Black innovators participating in the patent system was alarmingly low at only 0.4%. The study also reported that only 12% of US innovators are women. Approximately 60% of these innovators worked in a company, so they were able to use company resources to innovate. Professor Robinson stated that this study, along with similar studies, show that the U.S. patent system is leaving behind women, minorities, solo inventors, and small or solo enterprises.

The USPTO’s interest in AI threatens to make the patent system less accessible

Professor Robinson worries that the USPTO’s recent interest in AI technology threatens to make the patent system even less accessible for underrepresented innovators. Specifically, the USPTO has expressed interest in using AI models to assist during patent examination. And, the USPTO recently issued a request for comments regarding how it should treat inventions created entirely by or with the help of AI.

A major issue with AI technology is AI bias. Because humans create AI models, certain biases can intentionally or unintentionally be passed onto the AI models through biased data, thus leading to biased results and decisions. For example, a 2016 study showed the gender bias of an AI software model, which was more likely to associate a computer programmer with a man, and a homemaker with a woman. The current patent examination system (without AI) already has some bias—patents with women named as inventors are more likely to be rejected and have narrower issued claim scope than patents with male inventors. According to Professor Robinson, the use of AI during patent examination threatens to make the process even more biased. Additionally, if the USPTO were to recognize AI models as inventors, then the system would further crowd-out human innovators and underrepresented innovators.

Best Practices for AI-Assisted Patent Examination

In light of this great potential for AI bias leading to the decreased accessibility of the patent system, Professor Robinson proposed several best practices for AI-assisted patent examination to ensure that the system does not further alienate underrepresented innovators.

1. We should acknowledge that AI models are created by humans. Thus, AI creators should be governed by professional principles, similar to a professional oath, to make sure they apply the proper standards when creating AI models.

2. We need to take into account and avoid potential patent proxies. Specifically, we must consider whether the information provided during patent prosecution is going to be used as a proxy to represent something else. For example, the number of claims could potentially be used as a proxy for the financial resources of the applicant and whether the applicant will be able to commercialize the patented technology. The USPTO should ensure that the provided information is not used as a proxy in this way and that it is only used to determine whether a patent application meets the statutory requirements of patentability.

3. We should obtain applicant consent for the use of AI during examination. This encourages transparency and provides notice, allowing repeat applicants to tailor their strategy to reach successful outcomes.

4. AI systems should be subject to administrative and judicial review. Patent examination is not a black-box system, and AI patent examination should not be either. The creators of these models should be held accountable when they make mistakes, and these mistakes should be subject to review.

5. We should implement human-in-the-loop systems where a human works alongside an AI model. Under this system, humans can identify boundary cases and intervene where needed to prevent harmful results.

6. The AI examination model needs to provide clear outcomes. Specifically, the AI model should define success and produce actionable conclusions.

AI should also be used to help underrepresented inventors

Even with these potentially negative consequences of introducing AI to the patent system, Professor Robinson concluded that AI may also be a useful tool for solving the accessibility problem. AI can augment the USPTO’s current inventor services, including the patent pro bono assistance program and the Patent and Trademark Resource Center, and improve the delivery of legal services. Professor Robinson stated that if the USPTO is going to use AI, it must also provide AI tools to the public and provide underrepresented innovators the necessary tools and resources.

To watch Professor Robinson’s full presentation, please visit IBL’s YouTube channel at https://www.youtube.com/watch?v=qsYU4Fs4hgQ. For a deeper discussion of Professor Robinson’s research and analysis of this accessibility problem, see his full article at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3580850.

 

Sources:

[1]Adams Nager et al., The Demographics of Innovation in the U.S., Information Technology & Innovation Foundation (Feb. 24, 2016), https://itif.org/publications/2016/02/24/demographics-innovation-united-....

[2] Tolga Bolukbasi et al., Man Is to Computer Programmer as Woman Is to Homemaker? Debiasing Word Embeddings (2016), https://arxiv.org/pdf/1607.06520.pdf.

[3]Kyle Jensen et al., Gender Differences in Obtaining and Maintaining Patent Rights, 36 Nature Biotechnology 307 (2018), https://www.researchgate.net/publication/324250172_Gender_differences_in....

 

- Madison Murhammer Colon