Welcome to Doctrine & Data, a blog that uses data to test intellectual property law and looks at how lawyers can use AI reliably.
I’m Thomas Reichert, a registered patent attorney and Assistant Professor of Law at Southern Illinois University Simmons Law School. My work tests legal doctrine against data and asks a simple question: Does the law work the way we think it does?
What This Blog Is About
Legal doctrine is full of multifactor tests, balancing frameworks, and totality-of-the-circumstances analyses. Courts and practitioners treat these frameworks as if every factor matters, as if the careful weighing of numerous considerations is what produces just outcomes.
But what if that’s not actually how decisions get made?
My recent research on trademark confusion doctrine used large language models to code roughly 4,000 decisions of the Trademark Trial and Appeal Board (TTAB). It found that the traditional thirteen-factor DuPont test effectively collapses to two factors: a rule using only mark similarity and goods relatedness matched the Board’s outcome in 99.55% of 4,651 mark comparisons drawn from those decisions (one decision can compare more than one pair of marks). I checked the model’s coding against my own review of 1,002 factor-level findings; they agreed 97.11% of the time. The paper (opens in new tab) appears at 36 Fordham Intell. Prop. Media & Ent. L.J. 678 (2026), drew coverage from MLex, World Trademark Review, and The Fashion Law, and was cited in the cert petition in World Champ Tech v. Peloton (cert. denied Feb. 2026). When the Supreme Court took up RiseandShine v. PepsiCo, I wrote an amicus brief in support of neither party.
I also built The Fair Use Database (opens in new tab), which codes every substantive federal fair use opinion. Its research assistant, Folsom, answers only from the coded cases and cites its evidence, and its coding of case outcomes matched human coding in 97.4% of 453 hand-coded validation analyses.
Posts here cover:
- AI in legal work: Structured methods, where models fail, and how to check their output. Until those posts are up, the method is on the AI Advisory page.
- Trademark doctrine: What the data show about how confusion is actually decided.
- Empirical legal studies: Methodology, findings, and implications for practice.
- Patent practice: Lessons from prosecution, examination, and litigation.
Who This Is For
I write for practitioners who want to know what the numbers mean for their cases, academics interested in empirical approaches to doctrine, and anyone curious about how the law actually works.
The tone here will be rigorous but (I hope) readable.
Following Along
Subscribe via RSS, follow me on LinkedIn (opens in new tab), or see my research, publications, and CV on the Academic page.
Thanks for reading.
TR