Michael Smith

@msmith750.bsky.social

Associate professor of law at University of Oklahoma College of Law. Researching constitutional law, criminal law, and legal oddities. https://law.ou.edu/node/716

Tired: AI is bad at predicting Supreme Court case outcomes because it's led astray by coverage emphasizing partisan splits. Wired: Virtually all academic Supreme Court case prediction discourse is a waste of time, and moving beyond it would make everyone more productive.

SSRN@ssrn.bsky.social · 7h ago

This paper tests #GPT-5, Gemini 2.5 Pro, Claude Sonnet 4.5 & Grok 4 on every SCOTUS argued case from OT 2025. Accuracy looked good until you realise most of it just reflected the Court's usual ideological alignment, not actual legal analysis. spkl.io/633247zQpk #AI #LegalTech

Predicting the Court: Evaluating Large Language Models as Forecasters of Supreme Court Decisions

While I've now added a quote to this moment from "In the Loop" to this piece, I've used ellipses to cut out choice language. A model of restraint. (Also, this exchange is a load-bearing piece of dialogue for me. I'm reminded of it multiple times each week.) www.youtube.com/watch?v=1HFp...

You're A Legend

YouTube video by aznhustla88

youtube.com

Michael Smith@msmith750.bsky.social · 3d ago

Just nixed an extended "Andor" reference from this last symposium essay, in case you thought I was becoming entirely undisciplined in my work.

As another PSA, The University of Pennsylvania Law Review is requiring AI disclosures too. They don't have a Google form and are asking authors to include a note in their cover letters. Anyway, RIP to my submissions as I found this out too late, but did I really have a chance?

NYU Law Review has a nice AI disclosure requirement that lets you include an optional explanation of your philosophy regarding AI use in drafting legal scholarship at the end of the document. Always happy to share my thoughts.

I don't think you should necessarily go to hell for listening to your phone on full volume in a public place. BUT you should be forced to sit for a long time (days, maybe years?) in a crowded, pre-Heaven waiting room where almost everyone else is listening to stupid shit on their phones.

As the submission season opens, a list of journals on my "Cool List" based on what they've risked publishing: Quinnipiac (Originalism in the Year Three Thousand - papers.ssrn.com/sol3/papers....) Toledo (Resisting Generative AI with Fake Scholarship (forthcoming) papers.ssrn.com/sol3/papers....)

Originalism in the Year Three Thousand

<span>While procrastinating from grading at the end of the Spring 2024 semester, a series of mishaps led to my inadvertent teleportation to the year three thous

papers.ssrn.com

Just got through the first round of edits on this piece. Kudos to the students at Toledo for an excellent job of putting up with my nonsense.

Michael Smith@msmith750.bsky.social · 4mo ago

My article, "Resisting Generative AI with Fake Scholarship" is forthcoming in the University of Toledo Law Review! papers.ssrn.com/sol3/papers.... (The updated draft incorporates some recent fake scholarship developments, plus color-coded citations to identify information that isn't made up).

Resisting Generative AI With Fake Scholarship
58 University of Toledo Law Review (forthcoming 2027)

32 Pages Posted: 9 Mar 2026 Last revised: 13 Apr 2026
Michael L. Smith
University of Oklahoma - College of Law

Date Written: March 07, 2026

Abstract
Generative AI companies and enthusiasts claim that it's only a matter of time before lawyers, judges, law students, and law professors embrace this new technology and cede most, if not all, of their work over to skillful machines. Legal scholarship is one such domain that AI will purportedly dominate. This Article takes issue with such a dystopian prospect and proposes that authors and law journals resist by writing and publishing fake scholarship. Doing so may corrupt the body of scholarship upon which language models train, leading to increasingly inaccurate outputs that render generative AI useless for legal scholarship and other forms of legal writing. After describing and defending this proposal against likely objections, I provide an example of fake scholarship as a template for those interested in joining the cause.

Keywords: generative AI, artificial intelligence, legal writing, legal scholarship, language models, credibility, fun, fake scholarship

Hoping to start on my last symposium piece today. Can't wait to resolve not to commit to so many symposia in the upcoming year and then break this promise to myself by September or October.