In the preceding post, I highlighted two recent critiques of Professor Colleen Chien's recent article on predicting patent litigation. Below, Professor Chien responds to these critiques. – Jason
By Colleen V. Chien, Assistant Professor of Law, University of Santa Clara School of Law
In my article Predicting Patent Litigation, I describe the relationship between a number of intrinsic and previously unexplored acquired patent variables and the likelihood of the patent being litigated. In the article, I state that “from the starting point presented here, there are a number of directions that follow-up research could take to improve the resolution of the ranking approach described here that, while promising, does not provide a ‘commercial grade’ solution to outstanding patent-clearance problems.” Along this vein, Petherbridge and Kesan et al provide thoughtful suggestions and questions about how the analysis could be verified, refined, and extended. Their input is timely as efforts to do so are just getting underway, as part of adapting this exploratory project to commercial settings. Looking at more, and more recent patents, replicating the analysis, and adding additional variables, where it makes sense, will necessarily be part of this effort.
Since my paper was published I have been approached by a dozen or so parties interested in testing and applying the paper’s insights. The diversity of these interests addresses the two questions raised by the commentators. First, how precise must the model be? Though Petherbridge’s answer is “much more so,” the answer necessarily depends on the context of the application. On one hand, no level of precision in the algorithm can best a patent lawyer’s expert analysis of the claims and in very few cases would it make sense to proceed solely based on a mere calculation. On the other hand, if an insurance company is trying to use this as one of a variety of variables for calculating defense insurance rates, or for related applications where this filter is applied with others, this kind of algorithm can be useful. Likewise, the question, what variables should be considered? cannot be answered in a vacuum. Kesan et al suggest variables such as whether the particular patent is embodied in a product and the historical revenue of the patent owner. But even assuming these are better than other variables, the difficulty of getting them may make them impractical in many cases.
My article flags the difficulties with tracing what happens to a patent after issue as a potential problem but was not intended to definitively recommend that recordation rules be changed. Indeed, as I have explained in my recent comment “The Who Owns What Problem in Patent Law,” there are many reasons, the majority of which are economic and inadvertent rather than strategic, that it’s hard to determine who owns what patents. As part of the Kappos administration’s helpful efforts to improve access to patent data (examples here and here), the PTO has recently solicited input on this very question. While greater applicant disclosure is worth considering, I believe that the PTO could do much more with the information it already has. Currently, you cannot search among only in-force patents, or easily tell who the owner of record of patent is, how many times it has been cited, how much longer it may be in force, whether it has been reexamined, who requested the reexamination, or whether or not the current owner is a large or small entity. Yet making this information, which already exists, more accessible and supportive of business decision-making, could go a long way to enhancing the public’s ability to assess patent risks and opportunities.