Nontraditional Approaches to Judicial Opinions

Recently, there have been a few stories in the news about nontraditional approaches to expressing judicial opinions.  For example, Judge Lawrence VanDyke of the Ninth Circuit Court of Appeals issued a video dissent via YouTube in the case of Duncan et al. v. Bonta, No. 23-55805.[i]  The Arizona Supreme Court created a pair of AI-generated avatars “to keep the public informed about court-related news, information, and activities.”[ii]  And judges in both the United Kingdom and the United States have openly discussed the potential uses of generative AI in both case analysis and opinion drafting.[iii]  While innovation is generally a good thing (e.g., it’s hard to imagine using actual books for legal research or typewriters for legal drafting anymore), should we embrace these new approaches to judicial opinions?

To answer that question, we must first identify the components and value of a traditional judicial opinion.  A traditional judicial opinion is a written text drafted by one or more human authors employed by the judiciary upon review of written (and sometimes oral) arguments presented by the parties and any amici.  It typically identifies the factual context of the case, the claims presented for review, and the applicable standards of review before analyzing the claims under the governing law.  It is then memorialized in published reporters and/or court records for future reference.

Taking the nontraditional approaches out of order, using generative AI to further assist the drafting process seems, more or less, like simply adding a third person into the mix (akin to an additional law clerk or administrative assistant), and, in my opinion, seems like a good use of the technology to improve efficiency.  But using it in the analysis process feels less comfortable because an AI platform was neither elected nor appointed (nor hired by someone who was) to analyze and evaluate claims presented to the judicial branch of government.  And, while we know that its responses to prompts are based primarily on probability-based algorithms, we don’t fully understand its “decision-making process.”  Therefore, unlike with humans, we cannot discern all factors affecting its “decision,” including baked-in bias and lack of human experience.  On the other hand, there is a vetting process for all human members of the judicial branch, whether it be through voting, nomination and consent for appointment, or the interviewing and hiring process.  And, in that process, we can learn about those people’s backgrounds to discern how they might approach legal analysis.  But AI does not undergo that kind of rigor.  And there is currently no method by which it could.  In that sense, using AI for analysis is akin to a judge asking a random stranger for input on how to decide a case.

But what about using AI avatars to help the public understand a decision?  Judicial transparency is generally desirable to ensure public faith and trust, and the Arizona Supreme Court’s purpose in using their new avatars is “to help[] the public understand Arizona’s judiciary and the administration of justice.”[iv]  But is that what the AI avatars accomplish? 

According to the Arizona Supreme Court, the avatars, Victoria and Daniel, appear as newscasters “specializ[ing] in delivering clear, accessible explanations of case decisions and opinions.”[v]  In Victoria’s introduction video, she notes that her purpose is to “bridge the gap between complex legal proceedings and the public’s understanding of them.”[vi] While providing written case summaries and press releases is not new for courts, presenting them as live or recorded newscasts is.  One reason the court gave for the new format was to “meet the public where they are,” noting that, “[i]n today’s fast-paced digital world, people turn to short videos for news and updates.”[vii]

Doing so, however, discourages the public from actually reading the court’s opinion and engaging in critical thinking in favor of passive consumption of filtered summaries.[viii]  While written summaries also allow the public to avoid reading the entire decision, they still require the public to read, and that matters.  “The collective research shows that digital media have common features and user practices that can constrain learning. These include diminished concentration, an entertainment mindset, a propensity to multitask, lack of a fixed physical reference point, reduced use of annotation and less frequent reviewing of what has been read, heard or viewed.”[ix]  Reading content, on the other hand, results in greater recall and understanding than listening to identical material.[x]  And, for judicial opinions specifically, “[t]he meaning of a judgment often depends on its accompanying opinion (e.g., ‘The case is remanded for proceedings consistent with this opinion’), and a precedential rule—the proverbial ‘holding’ of a court—derives much, if not all, of its content from its surrounding justif[i]cation.”[xi]  There is a reason we don’t cite to summaries or the syllabus of a court opinion.

And about that video dissent on YouTube . . .

If you’re unfamiliar with Judge VanDyke’s dissent in Duncan et al. v. Bonta, No. 23-55805, he incorporated into his written dissent an 18-minute video he created and posted on YouTube, explaining how certain firearms work to hammer home his point that his colleagues on the bench erred in their majority decision based on a fundamental misunderstanding of firearms operation.[xii]  Setting aside the legal and ethical implications of this particular dissent,[xiii] is there any value in issuing opinions by video rather than written text?  Is there any harm?

In Judge Marsha Berzon’s concurrence in Duncan, she argued that video-based decisions have no place in our legal system because “we ground our jurisprudence in written precedent.”[xiv] And, she noted, “we do so for good reason: Written opinions promote uniformity, predictability, accountability, and care.”[xv]   While recognizing that technological advances make the preservation of oral opinions (as were given long ago) easier to preserve and distribute, she argued that “written opinions are more clear, useful, and accessible, and there are many potential challenges with video dispositions,” such as retention for later access, access for those without internet or digital devices, and the role—if any—transcripts of videos would play in subsequent analysis.[xvi]

On the flip side, as Judge Berzon noted, American judicial decisions were not required to be written until the late eighteenth century; before that, they were orally provided and preserved in only the notes of diligent counsel.[xvii]  Even today, justices in the United States Supreme Court occasionally deliver their dissents orally by reading the opinions aloud for added emphasis.[xviii]  But the written documents themselves are important and valuable.  They have proven their worth through their ability to be retained, accessed, and distributed, ensuring our precedential system of law endures.

In short, while technological advances allow for changes to the creation and dissemination of the traditional judicial opinion, the written word rooted in human analysis remains the best medium for the genre if we wish to preserve a culture that values critical thinking, predictability, and human experience.

[i] Kerry Breen, Judge releases video of himself disassembling guns in chambers in dissent against court ruling, CBS News, available at: https://www.cbsnews.com/news/judge-lawrence-vandyke-california-guns-video/ (last accessed April 7, 2025).

[ii] Arizona Supreme Court News Release, Arizona Supreme Court Introduces AI-Generated Court News Reporters to Enhance Public Engagement (Mar. 11, 2025), available at: https://www.azcourts.gov/Portals/201/News%20Release%20-%20Arizona%20Supreme%20Court%20Introduces%20AI-Generated%20Court%20News%20Reporters.pdf (last accessed April 7, 2025).

[iii] Jane Dalton, Judge admits using ‘jolly useful’ ChatGPT to write court ruling, The Independent (Sept. 15, 2023), available at: https://www.independent.co.uk/news/uk/home-news/chatgpt-ai-judge-chatbot-ruling-b2412378.htmlOn Using ChatGPT for Statutory Interpretation, Appellate Advocacy Blog (June 11, 2024), available at: https://lawprofessors.typepad.com/appellate_advocacy/2024/06/on-using-chatgpt-for-statutory-interpretation.html (last accessed April 7, 2025).

[iv] See News Release, supra, note ii.

[v] Id.

[vi]AZCourts, Victoria’s Introduction Video, Arizona Supreme Court AI Reporter, YouTube (Mar. 11, 2025),https://www.youtube.com/watch?v=jSSo4ScFzzU&list=PL5tiXCOtd9v7WfA0aJxo65dnlyEwJ7w6t&index=2

[vii] See New Release, supra, note ii.

[viii] Tonya Mosley, How social media algorithms ‘flatten’ our culture by making decisions for us, NPR Interview with Kyle Chayka (Jan. 17, 2024), available at: https://www.npr.org/2024/01/17/1224955473/social-media-algorithm-filterworld (last accessed April 7, 2025).

[ix] Naomi S. Baron, Why we remember more by reading – especially print – than from audio or video, The Conversation (May 3, 2021), available at: https://theconversation.com/why-we-remember-more-by-reading-especially-print-than-from-audio-or-video-159522 (last accessed April 7, 2025).

[x] Id.  See also Will Thalheimer, Debunk This: People Remember 10 Percent of What They Read, ATD Blog (Mar. 12, 2015), available at: https://www.td.org/content/atd-blog/debunk-this-people-remember-10-percent-of-what-they-read (last accessed April 7, 2025) (noting that the famed Cone of Experience from Edgar Dale suggesting that we retain 10% of what we read, 20% of what we hear, 30% of what we see, etc., is itself misinformation).

[xi] Richard M. Re, Artificial Authorship and Judicial Opinions, 92 Geo. Wash. L. Rev. 1558, 1562 (Dec. 2024), available at:  https://www.gwlr.org/wp-content/uploads/2024/12/92-Geo.-Wash.-L.-Rev.-1558.pdf (last accessed April 7, 2025).

[xii] Breen, supra, note i. A link to the video was embedded within a written dissent.  Duncan v. Bonta, No. 23-55805, 2025 WL 867583, at *49 (9th Cir. Mar. 20, 2025) (VanDyke, J., dissenting).

[xiii] In a concurring opinion, Judge Marsha Berzon criticized Judge VanDyke’s video for both relying on facts outside the record and making himself “an expert witness in th[e] case, providing a factual presentation with the express aim of convincing the readers of his view of the facts without complying with any of the procedural safeguards that usually apply to experts and their testimony, while simultaneously serving on the panel deciding the case.”  Id. at *23 (Berzon, J., concurring).

[xiv] Id. at *24.

[xv] Id.

[xvi] Id.

[xvii] Idsee also Erwin C. Surrency, Law Reports in the United States, 25 Am. J. Legal Hist. 48, 55 (1981).

[xviii] Stephen Wermiel, SCOTUS for law students (sponsored by Bloomberg Law): Dissenting from the bench, SCOTUSblog (Jul. 2, 2013, 10:34 AM), https://www.scotusblog.com/2013/07/scotus-for-law-students-sponsored-by-bloomberg-law-dissenting-from-the-bench/ (last accessed April 7, 2025).