A Network Theory of Patentability

Patent law is built upon a fundamental premise: only significant inventions receive patent protection while minor improvements remain in the public domain. This premise is indispensable for maintaining an optimal balance between incentivizing new innovation and providing public access to existing innovation. Despite its importance, the doctrine that performs this gate keeping role—non-obviousness—has long remained indeterminate and vague. Judicial opinions have struggled to articulate both what makes an invention significant (or non-obvious) and how to measure non-obviousness in specific cases. These difficulties are due in large part to the existence of two clashing theoretical frameworks, cognitive and economic, that have vied for prominence in justifying non-obviousness. Neither framework, however, has generated doctrinal tests that can be easily and consistently applied. This Article draws on a novel approach—network theory—to answer both the conceptual question (what is a non-obvious invention?) and the measurement question (how do we determine non-obviousness in specific cases?). First, it shows that what is missing in current conceptual definitions of non-obviousness is an underlying theory of innovation. It then supplies this missing piece. Building upon insights from network science, we model innovation as a process of search and recombination of existing knowledge. Distant searches that combine disparate or weakly-connected portions of social and information networks tend to produce high-impact new ideas that open novel innovation trajectories. Distant searches also tend to be costly and risky. In contrast, local searches tend to result in incremental innovation that is more routine, less costly and less risky. From a network theory perspective, then, the goal of non-obviousness should be to reward, and therefore to incentivize, those risky distant searches and recombinations that produce the most socially significant innovations. By emphasizing factors specific to the structure of innovation—namely the risks and costs of the search and recombination process—a network approach complements and deepens current economic understandings of non-obviousness. Second, based on our network theory of innovation, we develop an empirical, algorithmic measure of patentability—what we term a patent’s “network non-obviousness score (NNOS).” We harness data from U.S. patent records to calculate the distance between the technical knowledge areas recombined in any given invention (or patent), allowing us to assign each patent a specific NNOS. We propose a doctrinal framework that incorporates an invention’s NNOS to non-obviousness determinations both at the examination phase and during patent litigation. Our use of network science to develop a legal algorithm is a methodological innovation in law, with implications for broader debates about computational law. We illustrate how differences in algorithm design can lead to different non-obviousness outcomes, and discuss how to mitigate the negative impact of black box algorithms