Math You Can't Use: Patents, Copyright, and Software

Chapter 5. Profiting from Overbroad Patents

What’s wrong with the system of software patents that the courts have put in place? One set of problems stems from weaknesses in the patenting process that could be remedied by revised policies or more resources at the U.S. Patent and Trademark Office (USPTO), while other fundamental issues are rooted in the nature of software and software patents. First, software patents are too broad, and unavoidably so. Since software patents typically apply to the broad description of an idea—the interface instead of the implementation—they can block competitors from developing and marketing hundreds of alternative implementations. Second, as explained in chapter 3, all software is a compound invention, so patents are much more likely to hold up innovation in software than in fields where inventors are less dependent on the work of others. These problems bedevil all industries to some degree, but software is always hit hardest. Moreover, the rapid pace of software development magnifies the negative impact. Do Software Patents Promote Innovation? These problems are of particular concern because there is little evidence that software patents promote innovation—the usual justification for Profitingfrom Overbroad Patents CHAPTER FIVE 73 05-4942-2 CH5 10/27/05 3:30 PM Page 73 granting patents at all. Few scholars have addressed this question, primarily because of the difficulty of coming up with an appropriate measure of innovation. Most studies of how external economic factors affect innovation use a count of patents as the proxy for innovativeness, but some other measure is clearly required in this case.1 Using investment in research and development as a proxy for innovation , James Bessen of Research on Innovation and Robert M. Hunt of the Federal Reserve Bank of Philadelphia found that companies heavy in software patents do less research than expected.2 A lively debate followed the release of these results, with two scholars at the American Enterprise Institute–Brookings Joint Center for Regulatory Studies, Robert Hahn and Scott Wallsten, arguing that the numbers actually showed the expected positive correlation.3 Bessen and Hunt have posted a rebuttal to the rebuttal online.4 But even if Hahn and Wallsten are correct, and patents do shift research funds toward those fields of applied mathematics that are patentable, such gains need to be weighed against the means by which patents can also stifle technological progress. A small gain in the rate of innovation, visible if one looks at the data one way but not another, may not be enough to offset the negative effects patents have on the implementation of new software. Trouble at the USPTO Software is the canary in the USPTO’s coal mine. Although problems with the overall system have touched every industry, they plague software more than other fields because of its unique aspects. Novelty Because claims are innocent until proven guilty and so many inventions are obvious to people who have already seen the patent, patent examiners are encouraged not to use their ex post intuition about whether a patent is obvious. Instead, they are expected to go through a patent claim by claim to find proof that each should be invalidated or 74 PROFITING FROM OVERBROAD PATENTS 1. Trajtenberg (1990). 2. Bessen and Hunt (2004a). 3. Hahn and Wallsten (2003). This paper was partly funded by Microsoft. 4. Bessen and Hunt (2004b). 05-4942-2 CH5 10/27/05 3:30 PM Page 74 restricted. Being inundated by applications, however, USPTO examiners must forgo a full search to invalidate all claims and must let them pass with minimal examination. Examiners have between fifteen and thirty hours to process a patent, which may contain up to hundreds of claims; those claims for which specific existing literature (prior art) cannot be located in that time frame must be left to stand as given.5 The Manual of Patent Examining Procedure (MPEP) includes a single paragraph on the subject of searching prior art for computer-related literature .6 Here is the key sentence: “Generally, a thorough search involves reviewing both U.S. and foreign patents and non-patent literature.”7 In practice, this often becomes a search heavily biased toward U.S. patents. As an aid to the search, the applicant is obligated to list all relevant prior publications or patents in the application. According to one industry observer, 58 percent of software patents granted in 2003 referred only to prior patents and made no mention of prior art in...