5.3.3 Peer-to-Patent

Peer-to-patent is an open call that helps patent examiners find prior art; it shows that open calls can be used for more than prediction problems.

Patent examiners have a hard job. They receive terse, lawyerly descriptions of new inventions, and then must decide if the stated invention is “novel.” That is, the examiner must decide if there is “prior art”—a previously described version of this invention—that would render the proposed patent invalid. To understand how this process works, let’s consider a patent examiner named Albert, in honor of Albert Einstein who got his start in the Swiss Patent Office. Albert could receive an application like U.S. Patent 20070118658 filed by Hewlett Packard for a “User-selectable management alert format” and described extensively in Beth Noveck’s book Wiki Government (2009). Here’s the first claim from the application:

“A computer system, comprising: a processor; a basic input/output system (BIOS) including logic instructions which, when executed by the processor, configure the processor to: initiate power on self test (POST) processing in the basic input/output system of a computing device; present one or more management alert formats in a user interface; receive a selection signal from the user interface identifying one of the management alert formats presented in the user interface; and configure a device coupled to the computing system with the identified management alert format.”

Should Albert award 20 year monopoly rights to this patent or has there been prior art? The stakes in many patent decisions are high, but unfortunately, Albert will have to make this decision without much of the information that he might need. Because of the huge backlog of patents, Albert is working under intense time pressure and must make his decision based on only 20 hours of work. Further, because of the need to keep the proposed invention secret, Albert is not allowed to consult with outside experts (Noveck 2006).

This situation struck law professor Beth Noveck as completely broken. In July 2005, inspired in part by Wikipedia, she created a blog post titled “Peer-to-Patent: A Modest Proposal” that called for an open peer-review system for patents. After collaboration with the US Patent and Trademark Office and leading technology companies such as IBM, Peer-to-Patent was launched in June of 2007. A nearly 200 year old governmental bureaucracy and a group of lawyers seems like an unlikely place to look for innovation, but Peer-to-Patent does a lovely job of balancing everyone’s interest.

Figure 5.9: Peer-to-patent workflow. Source: Bestor and Hamp (2010).

Here’s how it works (Figure 5.9). After an inventor agrees to have her application go through community review (more on why she might do that in a moment), the application is posted to a website. Next, the application is discussed by community reviewers (again, more on why they might participate in a moment) and examples of possible prior art are located, annotated, and uploaded to a website. This process of discussion, research, and uploading continues, until, ultimately, the community of reviewers votes to select the top 10 pieces of suspected prior art that are then sent to the patent examiner for review. The patent examiner then conducts her own research and in combination with the input from Peer-to-Patent renders a judgment.

Returning again to U.S. Patent 20070118658 for a “User-selectable management alert format.” This patent was uploaded to peertopatent.org in June 2007 where it was read by Steve Pearson, a senior software engineer for IBM. Pearson was familiar with this area of research and identified a piece of prior art: a manual from Intel titled “Active Management Technology: Quick Reference Guide” that was published two years earlier. Armed with this document, as well as other prior art and the discussion from the Peer-to-Patent community, a patent examiner began a thorough review of the case, and ultimately threw out the patent application, in part because of the Intel manual that was located by Pearson (Noveck 2009). Of the 66 cases that have completed Peer-to-Patent, nearly 30% were rejected primarily based on prior art found through Peer-to-Patent (Bestor and Hamp 2010).

What makes the design of Peer-to-Patent especially elegant is the way that it gets people with many conflicting interests to all dance together. Inventors have incentive to participate in this open patent review because the patent office reviewed the Peer-to-Patent applications more quickly than patents going through the traditional, secret review process. Reviewers have incentive to participate in order to prevent the granting of “junk patents,” and many seemed to find the process enjoyable. Finally, the patent office and patent examiners have incentive to participate because it can only make their results better. That is, if the community review process finds 10 bad pieces of prior art, this work can be ignored by the patent examiner. In other words, Peer-to-Patent and a patent examiner working together should be as good or better than a patent examiner working in isolation. Thus, open calls do not always replace experts; sometimes they help experts do their work better.

Although Peer-to-Patent may seem different than the Netflix Prize and Foldit, it has a similar structure in that solutions are easier to check than generate. Once someone produces the manual “Active Management Technology: Quick Reference Guide” it is relatively easy—for a patent examiner, at least—to verify that this document is prior art. However, finding that manual is quite difficult. Peer-to-Patent also shows that open call projects are possible even for tasks that are not obviously amenable to quantification.