Strategic Obfuscation? Research Indicates University Patents Are Becoming Harder to Read
Are professors hiding the ball?
The patent system is built on a foundational bargain: in exchange for a temporary monopoly, an inventor must disclose the invention to the public in a manner that is clear enough for others in the field to understand and build upon. This disclosure function is critical for "standing on the shoulders of giants," allowing knowledge to spill over and fuel cumulative innovation (p. 2). Recent research provides quantitative evidence that this bargain with the public may be fraying, particularly in an academic setting. Are college professors hiding the ball?
The paper, from Xizhao Wang of Northwestern University, examines whether university inventor teams (and their patent drafters), influenced by institutional pressures, are strategically making their patents harder to understand.
The core problem addressed is the inherent tension between an inventor's desire to protect their competitive advantage and the public's need for clear, useful disclosure. The paper proposes a novel solution for measuring this behavior by analyzing the readability of patent text itself, using both traditional linguistic metrics and a new AI-based model. It finds that following major policy shifts that encouraged commercialization, "university inventors strategically limit information on how to make and use the invention" (p. 1).
The Tension Between Disclosure and Appropriability
Inventors walk a fine line. On one hand, clear disclosure can attract investors, partners, and licensees. On the other, it can "expose inventions to imitation and intensifies competition, which can diminish their commercialization potential" (p. 10). This creates a powerful incentive to disclose just enough to secure a patent, but not so much that a competitor can easily replicate or design around the invention. While legal standards exist to ensure clarity, their effectiveness has been a subject of debate.
The paper highlights that this is not a new concern, pointing to a body of legal scholarship that has long questioned the adequacy of patent disclosures.
Certain legal scholars point out that, despite the requirements of 35 U.S.C. § 112(a), patent documents often fail to effectively convey the intricacies of advanced technologies (Devlin 2009). They suggest that patents can be difficult to comprehend and are often "unreadable" (Seymore 2009), with patent writers sometimes using vague or unclear language to safeguard their proprietary knowledge (Roin 2005, Fromer 2008). (p. 3)
This perspective is credible because it acknowledges the real-world economic pressures faced by inventors and their institutions. The decision is not simply whether to patent or keep a trade secret, but extends to how much to reveal within the patent document itself. This suggests that "the extent of disclosure in patents may be a strategic decision by inventors, shaped by their institutional environment" (p. 3). For IP professionals and follow-on inventors, this implies that a granted patent may not be the open book it is intended to be.
Measuring Obfuscation with Readability Scores
To move from anecdotal evidence to empirical analysis, the research introduces a systematic way to measure patent clarity. This is done through two distinct methods.
Traditional Readability Scores
The first approach applies well-established linguistic formulas to the patent text. These are not new inventions but are repurposed for the unique context of technical and legal documents.
This paper applies four readability measures to assess patent readability. These four readability scores are developed either to measure the readability of technical manuals or assess a manuscript's complexity during the editing process. (p. 14)
In practice, this means using metrics like the Flesch-Kincaid Grade Level and the Automated Readability Index, which analyze factors like sentence length and word complexity (syllable count) to produce a score (p. 14). A higher grade level or index score suggests a text is harder to comprehend. While simple, these tools provide a standardized, objective baseline for comparing textual complexity across thousands of documents.
A Novel AI-Based Approach
The paper’s more innovative contribution is a custom readability score trained on the decisions of the ultimate arbiters of patent clarity: USPTO examiners.
By combining text embeddings generated by SciBERT with patent examiners' rejection decisions, I train supervised learning models to predict the probability of rejection under the 112(a) rule. These predicted probabilities serve as Al-based readability scores, offering a novel measure of text readability as interpreted by patent examiners. (p. 4)
This AI-based score is trained specifically on rejections issued under 35 U.S.C. § 112(a), which requires the patent's specification to contain a clear written description of how to make and use the invention (p. 15). By using a language model (SciBERT) fine-tuned for scientific text, the model learns the complex linguistic patterns that examiners associate with a lack of clarity. A higher predicted probability of a 112(a) rejection serves as a proxy for lower readability, providing a nuanced, domain-specific metric that goes beyond simple word and sentence counts.
The Bayh-Dole Act as a Natural Experiment
To test the hypothesis that commercial incentives drive obfuscation, the paper cleverly uses two major shifts in the U.S. innovation landscape as "natural experiments" (p. 4). The primary analysis focuses on the 1980 Bayh-Dole Act, which gave universities the right to own and commercialize inventions developed with federal funding. A second analysis examines the staggered establishment of university Technology Transfer Offices (TTOs). The study compares university-affiliated patents (the "treatment group") with corporate patents (the "control group") before and after these events.
The results are striking. The study found a clear and statistically significant decline in the readability of the "detailed description" section of university patents after these institutional shocks.
The results show a decline in the readability of university patents' detailed descriptions following the 1980 Bayh-Dole Act, relative to corporate patents. A similar reduction is observed after the establishment of Technology Transfer Offices (TTOs). (p. 5)
This effect appears to be isolated to detailed descriptions. The research performed a falsification test and found no corresponding decrease in the readability of the patent's "background and summary" sections (p. 22). This detail is important because, as the paper notes, "the detailed description is meant to teach how to make and use the invention, whereas the background and summary provide context" (p. 22).
This selective obfuscation strongly suggests a strategic, rather than accidental, change in disclosure practices. The implication is that when the financial stakes were raised, university inventors began to hold back the "how-to" details core to their inventions.
Why This Matters
This research provides compelling, large-scale evidence for a phenomenon that many patent practitioners have long suspected: that patent documents are not always written to be maximally clear. Whether this is on purpose or a byproduct of the research-paper-to-patent process remains a question.
The paper’s key contribution is to quantify this behavior and link it directly to commercialization incentives. By using the Bayh-Dole Act and the rise of TTOs as a lens, it demonstrates that policies designed to spur technology transfer may carry an unseen cost, for whatever reason.
The potential impact of these findings is a reduction in the "knowledge spillovers" that the patent system is supposed to foster. The study finds an association between reduced readability and a decline in external citations, suggesting "diminished disclosure clarity may have contributed to limited knowledge diffusion" (p. 5). That indicates that many of these patent applications might not be found by fellow inventors or examiners, e.g., due to the obscure language used. It appears that patents with sections that are hard to read are likely also harder to find.
Closing Thoughts
While policies encouraging university patenting have been successful by volume, this research raises a pragmatic and important question about the quality and utility of the resulting disclosures.
As the paper concludes, clarity of the written description is an important consideration for innovation policy. "The findings suggest that universities may be less open in sharing scientific discoveries with the general public after the implementation of university innovation policies, which is an important consideration for policymakers" (p. 27).
On the other hand, section 112 only requires that the patent specification provide enough detail to enable "any person skilled in the art" to make and use the invention without undue experimentation. For a large number of patents and publications, a person skilled in the art may not be bothered by the low readability of a detailed description. This specialized audience is accustomed to dense, jargon-laden text, and what general linguistic models perceive as low readability may simply be the precise and efficient language of a particular technical domain.
The observed decline in readability scores could therefore be a byproduct of increasing technological specialization rather than a conscious effort to obscure the invention's core principles. Consequently, while the paper's data points to a change in writing style, it remains an open question whether this change truly constitutes a failure of legal enablement for the patent system's intended reader—a PHOSITA.
Today, AI can likely fix this apparent disconnect with the public.
Clarity of writing should be a noble goal in itself. Writing a patent specification in a way to teach and explain to various levels of expertise can be valuable to fellow artisans, practitioners, examiners, courts, and the public. Playing “hide the ball” with technical details and inventive concepts can undercut support for legal requirements under sections 101, 102, 103, and 112. These so-called obscure patent disclosures may have missed a key opportunity to maximize their value.
Continued research in this area is essential to ensure that the patent system's dual goals of private incentive and public disclosure remain in balance.
Citation: Wang, Xizhao. (2025). "Patenting and Information Disclosure." Available at SSRN: https://ssrn.com/abstract=5387619.
Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.