This is the property of the Daily Journal Corporation and fully protected by copyright. It is made available only to Daily Journal subscribers for personal or collaborative purposes and may not be distributed, reproduced, modified, stored or transferred without written permission. Please click "Reprint" to order presentation-ready copies to distribute to clients or use in commercial marketing materials or for permission to post on a website. and copyright (showing year of publication) at the bottom.

Judges and Judiciary

Apr. 24, 2023

Why we write

Why AI programs will never come close to producing an opinion as elegant and enduring as Cardozo's Hynes v. N.Y. Cent R. Co; and why they should never be asked to try.

Michael Antonovich Antelope Valley Courthouse

Michael C. Kelley


Yale Law School

In the wake of reports that an AI program had passed the bar exam, and on an impulse I can trace to a thoughtful column in the March 6 Daily Journal by Justice Arthur Gilbert about AI-generated judicial opinions ("The Horror! The Horror!"), I recently asked one of the current versions of these natural language AI technology tools to take a crack at composing a passage describing the doctrine of proximate cause for one of my rulings - and to do so in the style of Justice Cardozo.

As modest insurance that you will read the rest of this column, which I concede lacks the impulsivity that launched it, I have reserved revelation of the results of my AI dabbling for the final paragraph. Also, I must state for the record that I did not incorporate any of the output in my order as it was finally drafted and issued. "Faux Cardozo" stayed in chambers; the parties received what they were entitled to - authentic Kelley.

Still, the episode got me thinking a bit about why we write opinions (or in the trial court "statements of decision" and other forms of written rulings). The simplest answer is that we write because the rules require it. For example, California Constitution, Art VI 14, requires a writing with "reasons stated" when the courts of appeal "determine causes"; California Code of Civil Procedure (CCP) § 632, provides for a "statement of decision" when a trial court tries an issue of fact; and CCP § 437c(g) imposes requirements for the content of orders denying or granting motions for summary judgment. Although it is true that we write because we are expected to, as often happens after my grandchildren leave a visit, the question "but why?" still hangs in the air.

In the course of these ruminations, I learned that a former colleague, Mark Haddad, teaches a seminar at the USC Gould School of Law called "Judicial Opinion Writing." So, I reached out to Professor Haddad to find out if the legal academy had insights on the whole point of these reason-writing exercises.

You may be surprised to learn that the ivory tower in University Park provided some keen insights - but I was not. That is because Professor Haddad is first and foremost a highly accomplished appellate lawyer. He was also a law clerk for Justice Brennan and before that was editor-in-chief of the Yale Law Journal. So, while he certainly has the academic credentials to teach an upper-level theory course at an elite law school like USC, he has also had great success in, and insights about, the real world in which most of us function. His reading recommendations did not disappoint. Several of the recommended sources were published in a special issue of the University of Chicago Law Review in 1995. This issue was devoted to what one of the contributing authors described as the "question... whether it matters how judicial opinions are written and if so, why." See James B. White, What's an Opinion For? 62 U.Chi.L.Rev. 1363 (1995).

Opinions as precedent

White, a professor of law and English at the University of Michigan, started his article by suggesting a thought exercise: "to imagine a legal world without the judicial opinion at all." In this world "the law would be what (the judges) did, not what they said" and they would simply "issue orders reflecting their judgment." Even on appeal, in this imagined Kafkaesque process, the appellate court would simply affirm or reverse "never issuing an opinion explaining or justifying its decision." [Id. at 1364]

Professor White initially posited that in such a world there would be no such thing as "precedent," but as law professors often do, he followed this intellectual head fake with an expression of skepticism about it and suggested that, even without explication of the reasoning in particular decisions, we would still know the outcome of cases, and over time we could gain some expertise in predicting the outcome of other cases. Lawyers "who observed a set of such cases... would get some notion of which facts were relevant, and which ones were not; of which arguments appealed, and which did not; of the methods of statutory and constitutional interpretation that were congenial or uncongenial... ." [Id. at 1365]

In 1995, the idea that judicial outcomes without any revealed reasoning might support anything approaching a system of "precedent" was worth musing over in a "thought experiment," but in 2023 the prospect is less a matter of musing and more a matter of imminent reckoning. The legal profession must confront the implications of large-language models, self-supervised learning algorithms and other current and yet to be developed data-driven tools (which go far beyond being quite good at predicting how likely you may be to enjoy a new song from Adele) for predicting outcomes in real cases. Once the actual results of decades of prior cases have been coded along with their facts (or simply digested whole by a bank of supercomputers), will a highly reliable system for predicting legal results emerge? I don't know for sure, but I suspect the prospect that it may looms larger than we might imagine.

Fortunately, Professor White did not dwell long on the prospect that a legal system without opinions might yet generate something resembling a functioning model for predicting outcomes. That is because our concept of precedent is broader than merely a system that enables a certain degree of outcome predictability. As White explained, the concept of precedent also entails a system that imbues the prior outcomes with the authority to constrain or guide future results. As he stated, "It is here, in the creation of legal authority, rather than in the facilitation of prediction, that the opinion performs its peculiar and most important task. For the law as we know it is not merely prediction of what those with power will do, nor even of what arguments will move them in one direction or another; it is also the invocation of the authority of prior texts to shape and constrain what may be done in the present." [Id. at 1367]

But what is it about a judicial opinion that creates this authority to influence outcomes? Professor White identified two dimensions in which this authoritative feature of legal opinions operates: (1) the prior texts and judgments that the author of the opinion decides to invoke and (2) the methods that are employed to discuss and apply them. This second dimension includes the manner in which the facts are established, the reading of the prior texts and the construction of legal meaning that the opinion creates). [Id. at 1368] According to White, the judicial opinion not only "explains or justifies the result," it also "connects the case with earlier cases" and "translates the experience of the parties, and the languages in which they naturally speak of it, into the language of the law." [Id. at 1367] Finally, it "translates the texts of the law - the statutes, opinions and constitutional provisions - into the terms defined by the facts in the present case." Crafting an opinion thus entails conducting what is, in essence, a law-creating "conversation" between the specifics of the case at hand and the cited authoritative texts. Doing this well can show the parties "that their case has been treated with intelligence and respect." It also "has large consequences for the future [because] it deeply affects and shapes the way we think and argue and, in so doing, constitute ourselves through the law." [Id. at 1368]

At this point, White synthesized why he considered a collection of law review articles on opinion writing to be timely and important: "[T]he great question of the day is whether law will move in the direction of trivializing human experience, and itself, or in the direction of dignifying itself and that experience. This is in large measure a function of the way in which the minds that work in this field manifest themselves. The deepest sources of meaning and dignity in human life are activities of love and art; properly understood, the law cannot only enable them, it can be one of them, an activity fully worthy of the human mind and spirit." [Id. at 1369]

These insights, and in particular the passage just quoted, provide the first plank for a foundation I ultimately will stand upon to pronounce - with some measure of conviction - that AI will never author judicial opinions. It is the very "humanity" of the authors of such opinions (the "We" in the title of this essay) that animates the legitimate power of the words that end up expressing the outcomes in our cases.

Opinions as sources of credibility and consistency in the law

With my mind and spirit thus energized with the prospect that my daily task of writing out reasons for my decisions might sit at the table with "love and art," I turned to a piece in the same journal by a highly experienced judge, Patricia M. Wald, of the U.S. Court of Appeals for the District of Columbia Circuit (chief judge from 1986-1991). See Patricia M. Wald, The Rhetoric of Results and the Results of Rhetoric: Judicial Writings 62 U.Chi.L.Rev. 1371 (1995). Her article provides some thoughtful and real-world observations that relate to the process or craft of writing appellate opinions, but she started more broadly by identifying two reasons for writing opinions, both of which she linked to the legitimacy of judicial power. The first is "to reinforce our oft - challenged and arguably shaky authority to tell others - including our duly elected political leaders - what to do" noting that "[o]ne of the few ways we have to justify our power to decide matters important to our fellow citizens is to explain why we decide as we do." The second reason to write is "to demonstrate our recognition that under a government of laws, ordinary people have a right to expect that the law will apply to all citizens alike" which means that judges "should interpret the same laws the same way," a goal of consistency that requires those interested in the legal system to "know why a particular judge came to a particular decision in a particular set of circumstances." [Id. at 1372]

Given the goals Wald described for opinions - lending "credibility" and "consistency" to the justice system - it is not surprising that she bemoaned the impact of burgeoning caseloads, which, even as early as 1995, resulted in fewer and fewer appeals being resolved with fully written opinions, and even fewer being resolved with published ones. She highlighted the problem with unpublished opinions in particular. "[The] process [of writing an opinion to be published] puts the writer on the line, reminds her with each tap of the key that she will be held responsible for the logic and persuasiveness of the reasoning and its implications for the larger body of ... law." In contrast, "writing to explain a preordained result with no concern for its precedential effect under a self - imposed time constraint of hours is something else entirely, inviting no backward looks or self-doubt. Rhetoric will always be tied to import and permanence, and its absence in unpublished decisions signifies that they are the product of a different and much abbreviated decision - making process." [Id. at 1375]

Wald referred to the more robust, reflective process entailed in crafting a publishable opinion as "judicial rhetoric" and allowed that "[i]t may indeed be possible to have individual justice without judicial rhetoric" but she insisted that without "any permanent record as to why cases come out one way or the other" it was impossible to have "reasonably consistent justice administered by hundreds of judges for millions of people." [Id. at 1376] Ultimately, she recognized that, in an appellate court, there are constraints on judicial rhetoric, including the necessity of achieving consensus among colleagues, not only on the outcome but on how the reasons for that outcome should be expressed. But, even within those constraints, "judges still use rhetoric to maneuver. The way they present the facts, the way they describe the rules and standards of review, the way they 'handle' precedent, their decisions to write separately or stay with the pack, all provide wide avenues in which to drive the law forward. A judge's individual skill at working these levers of power, and doing so in a way that does not overly antagonize colleagues, continues to have a powerful influence on decision-making. That is why, in the end, judges - as well as their words - matter so much." [Id. at 1419]

In sum, Wald's view was that opinions are written to persuade the community that cares about the law that a particular outcome is the product of a reasoned process and to expose that process to scrutiny as a kind of quality control to help ensure consistency in the application of the law.

For its part, an AI chat tool does not reveal the process that lies behind the sequence of words that it guesses (smartly based on the body of language data it has absorbed) to be an appropriate response to a query, and it frankly doesn't care if, upon scrutiny, its process or output is found wanting. Thus, the process of judging that includes writing reasons that meet Judge Wald's goals of credibility and consistency also seems to me to be one that is inherently human.

Opinions as expressions of sympathetic human judgment

In the same symposium issue, I next found: Martha C. Nussbaum Poets as Judges: Judicial Rhetoric and the Literary Imagination 62 U. Chi. L. Rev. 1477.

Professor Nussbaum, who holds a joint appointment to the law school and the department of philosophy at the University of Chicago, took as her point of departure the Walt Whitman poem By Blue Ontario's Shore, a piece which, in the immediate wake of the Civil War, sprang from the poet's imagined encounter with a "Phantom gigantic superb" offering these observations about the state of the Nation:

"Of these States the poet is the equable man, ...

He bestows on every object or quality its fit proportion,

Neither more nor less,

He is the arbiter of the diverse, he is the key,

He is the equalizer of his age and land,...

He judges not as the judge judges but as the sun falling

round a helpless thing...

He sees eternity in men and women, he does not see men and women as dreams or dots."

These passages led Nussbaum to observe that "[h]ere is the contrast between an abstract pseudo-mathematical vision of human beings and a richly human and concrete vision that does justice to the complexity of human lives." [Id. at 1479] She commented that the "complex image" of the "sun falling round a helpless thing" as a mode of the poet's judging, suggests "enormous detail and particularity" because "[w]hen the sun falls around a thing, it illuminates every curve, every nook; nothing remains hidden, noting unperceived... . So too the poet's judgment falls around all the complexities of a concrete case, perceiving all that is there and disclosing it to our view." This examination of the thing to be judged is "stern and rather pitiless: by comparing judgment to sunlight rather than gentle shade, Whitman indicates that the poet's commitment to fairness does not yield to bias or favor; his confrontation with the particular, while intimate, is unswerving. There is an ideal of judicial neutrality here, but it is a neutrality linked not with quasi-scientific abstractness, but with rich historical concreteness." [Id. at 1479-80]

As Nussbaum summed up her insight for judicial opinion writing, she wrote "[i]n order to be fully rational a judge must be capable of literary imagining and sympathy. She must educate not only her technical capacities but also her capacity for humanity. This means, I think, that literary art is an essential part of the formation of the judge - and, more generally, of the formation of citizenship and public life." [Id. at 1519]

My takeaway here? A rationality that is capable of "literary imagining and sympathy" does not spring from an algorithm. It must come from the intellect and lived-experience of a human judge.

Opinions as literature of the law

I finished my research by returning to the source that sparked my quest - Justice Cardozo himself. I meant to focus on his best known work, The Nature of the Judicial Process, a dusty copy of which still sits on my bookshelf, but was soon drawn to an earlier work I had not encountered before, a 1925 essay called Law and Literature. See Benjamin N. Cardozo Law and Literature, 14 Yale Review 699 (1925), in which Cardozo wrote, in part, to address the "amused or cynical indifference" he perceived as prevalent among lawyers of the era to the idea that the form of opinions, as opposed to their substance, had importance. His thesis was that "[f]orm is not something added to substance as a mere protuberant adornment. The two are fused into a unity." It was Cardozo's "faith" that "[f]orm alone takes and holds and preserves substance, saves it from the welter of helpless verbiage that we swim in as a sea of tasteless tepid pudding." [Id. at 700] (emphasis in original)

A future essay might delve into Cardozo's attempt in Law and Literature to articulate a classification scheme of judicial opinion forms, which he thought range from the "magisterial or imperative" (embodied in much of the work of Justice John Marshall, whose "organ tones... were meant to fill cathedrals or most exalted of tribunals"); to the "laconic or sententious" and the "conversational or homely" (of which Cardozo opined the "English judges have been masters," citing such examples as Lord Bowen's dictum "The state of a man's mind is as much a fact as the state of his digestion"); to the "refined or artificial"; to the "demonstrative or persuasive"; and finally, the "tonsorial or agglutinative." For present purposes, it suffices to note (and endorse) Cardozo's major point that literary style, and what he called the "architectonics" of opinions, matter when considering their role in a system of justice.

Cardozo's essay is approaching its centennial. I concede that his prose can land on modern sensibilities as a bit heavy - as if the time and care taken to craft the words into a literary form somehow have been impounded into a density belied by its relatively modest volume (his essay is but 18 sparse pages - less than many modern unpublished opinions). However, linguistic styles apart, the questions still arise: Do Cardozo's insights on form and substance in opinions still matter? Is there anything from Cardozo's own approach to writing opinions worth emulating or adapting nearly a hundred years later?

I think the answer to both questions is "yes. I am not alone. See Richard A. Posner, Cardozo: A Study in Reputation 51 (University of Chicago Press 1990) (Cardozo writes with "striking freshness, clarity, and vividness," with "[t]he power to compress a tradition of legal thought into a sentence." (pp. 45, 56). David A. Logan, The Man in the Mirror: Cardozo : A Study in Reputation (book review) 90 Mich. L. Rev. 1739, 1769 (1992) (more than a half century after Cardozo's death his "personal qualities of humility, compassion, and moderation are stuff of legends" and his reputation is secure, due in part to "his pragmatic jurisprudence, the breadth and depth of his writing [and] the sheen of his rhetoric... ." ); Richard H. Weisberg Cardozo's "Law and Literature": A Guide to His Judicial Writing Style, 334 Touro L. Rev. 349, 350 (2018) (Cardozo's essay "remains topical" and its view of "literary relation to law ... has very little to do with ornament and everything to do with the everyday use of language by lawyers").

Weisberg in particular makes the case for the relevance of Cardozo in modern law. He describes Cardozo's opinion in Hynes v. New York Cent. R. Co. 231 N.Y.229 (1921) as "the greatest appellate opinion of all time." Hynes was the case where a "lad of 16...poised for his dive" into the Harlem River was killed by a negligently maintained railway wire. The boy had been diving from a plank that was anchored under a rock on the railway's property, but extended out over the public waterway, which is where he was standing "when the wires ...flung him from the shattered board and plunged him to his death below." [Id. at 232] The case arose at the intersection of competing doctrines (landowners did not owe a duty to trespassers but did owe a duty to those on adjoining public ways).

Cardozo's concluding paragraph, which sums up how he resolved the doctrinal conflict, states: "In one sense, and that a highly technical and artificial one, the diver at the end of the springboard is an intruder on the adjoining lands. In another sense, and one that realists will accept more readily, he is still on public waters in the exercise of public rights. The law must say whether it will subject him to the rule of the one field or of the other, of this sphere or of that. We think that considerations of analogy, of convenience, of policy, and of justice, exclude him from the field of the defendant's immunity and exemption, and place him in the field of liability and duty." [Id. at 236]

Weisberg elaborates on his view of the brilliance of this opinion: "Justice, as Hynes tells us, needs every ounce of what literature has to offer law. Justice is both real and achievable, but you get there using the lawyer's basic tool: language, architectonics, style. In hard cases, the judge must especially rely on what Cardozo calls her 'informed intuition' and where that leads her must then be articulated through as perfect a fit as possible between substance and form. As Cardozo said in The Growth of the Law, '[j]ustice in this sense is a concept by far more subtle and indefinite than any that is yielded by mere obedience to a rule.' And in Paradoxes of Legal Science: 'many are the times when there are no legislative pronouncements to give direction to a judge's reading of the book of life and manners .... Objective tests may fail him or may be so confused as to bewilder. He must then look within himself.' This jurisprudence may not be what a social scientist would desire, but it is quite attractive to the lawyer-humanist in Cardozo. He wanted more of this in the profession, and this is why he wrote Law and Literature." [Id. at 357 (emphasis added)].

Cardozo's own words warrant emphasis as the coda of this essay: "But clearness, though the sovereign quality (of an opinion) is not the only one to be pursued, and even if it were, may be gained through many avenues of approach. The opinion will need persuasive force, or the impressive virtue of sincerity and fire, or the mnemonic power of alliteration and antithesis, or the terseness and tang of the proverb and maxim. Neglect the help of these allies, and it may never win its way." [Law and Literature, supra, at 702]

Words to write by. They also seal the case in my mind for why AI programs will never come close to producing an opinion as elegant and enduring as Hynes; and why they should never be asked to try.


I promised at the outset to share the results of my request that an AI program provide a description of proximate cause in the style of Cardozo. Here is what it delivered:

"To approach the legal doctrine of proximate cause is to grapple with the quintessential principle in the law of torts. Its essence lies in determining whether a defendant's actions are causally linked to a plaintiff's harm. Proximate cause signifies the legal notion that a defendant can be held accountable solely for the harm that was directly caused by their actions and that was reasonably foreseeable."

Maybe it's just my human petulance at work, but really? "To grapple?" "quintessential?" "signifies?"

Seems like "tepid pudding" or perhaps pudding with some "protuberant adornment."

One thing is clear. It's not me; and it's definitely not Cardozo. It is, for now... nothing in the law.


Submit your own column for publication to Diana Bosetti

For reprint rights or to order a copy of your photo:

Email for prices.
Direct dial: 949-702-5390

Send a letter to the editor: