Fear and trembling with apologies to Kierkegaard

The unease continues. In my last column "Frankenstein's Monster*," (Daily Journal (DJ), May 1, 2023), I responded to Judge Michael Kelley's excellent DJ article, "Why we write" (April 24, 2023). He marginally eased my concerns about artificial intelligence (AI) writing judicial opinions. This prompted an asterisk after the title. Yes, AI merely spits out words culled from innumerable sources in the past in milliseconds. And, yes, I agree that does not produce the eloquence and perfect turn of phrase of a Justice Cardozo. So? I am hard-pressed to find anyone who does. But even Cardozo doesn't always get it right, at least in my opinion. Numerous scholars agree. Take for example the famous case Palsgraf v. Long Island Railroad Co., 248 N.Y. 339 (1928). However eloquent the presentation of facts, the result is open to serious question. And if Justice Cardozo can get it wrong, even once, imagine the damage AI can do. What follows is a quick review of the decision, and its positive and negative characteristics.

The Palsgraf decision is a landmark case in American tort law. It was decided by the New York Court of Appeals in 1928. The case involved a woman named Helen Palsgraf who was waiting for a train at a Long Island Railroad station. While she was standing on the platform, two men attempted to board a moving train. One of the men was carrying a package that contained fireworks. The package fell onto the tracks and exploded, causing scales at the other end of the platform to fall on Palsgraf and injure her. The court held that the railroad company was not liable for Palsgraf's injuries because there was no way they could have foreseen that the package contained fireworks and that it would cause scales to fall on her.

Cardozo's Palsgraf decision established the principle of foreseeability as a criterion for determining whether a defendant owes a duty of care to a plaintiff. The decision has been praised for its clarity and elegance, but also criticized for its narrow and formalistic approach to the concept of negligence. Some possible criticisms of Cardozo's decision are: It ignores the causal connection between the defendant's conduct and the plaintiff's injury, and focuses only on the foreseeability of the harm at the time of the act. It creates an arbitrary distinction between direct and indirect harms, and fails to account for the complexity and unpredictability of human affairs. It imposes an unrealistic burden on plaintiffs to prove that their injuries were within the scope of the risk created by the defendant's negligence. It disregards the moral and social values that underlie tort law, and reduces it to a mechanical application of rules. It contradicts the principle of corrective justice, which requires that those who wrongfully cause harm to others should compensate them for their losses.

There you have it. Oh, by the way, I did not write the preceding two paragraphs. Bing's recent AI program wrote them. I left out the quotes to fool you. Did I? Not my style, but I agree with the premise, the writing of which could pass for much of what is current faire. Nothing wrong with it, though different writers might further develop the causal connection of injury to Mrs. Palsgraf with the acts of the guards to help the late passengers board the moving train.

I resisted AI's desire to take over the balance of this column. If you believe it, what follows is my own modest effort. Let's look at Judge Andrews' dissent. His statement of the facts, though not as colorful as Cardozo's, is concise and to the point. The same may be said of his statement of the issue. Please trust me on this, the quotes that follow are from Andrews' dissent:

"Assisting a passenger to board a train, the defendant's servant negligently knocked a package from his arms. It fell between the platform and the cars. Of its contents the servant knew and could know nothing. A violent explosion followed. The concussion broke some scales standing a considerable distance away. In falling they injured the plaintiff, an intending passenger.

"Upon these facts may she recover the damages she has suffered in an action brought against the master? The result we shall reach depends upon our theory as to the nature of negligence. Is it a relative concept - the breach of some duty owing to a particular person or to particular persons? Or where there is an act which unreasonably threatens the safety of others, is the doer liable for all its proximate consequences, even where they result in injury to one who would generally be thought to be outside the radius of danger? This is not a mere dispute as to words. We might not believe that to the average mind the dropping of the bundle would seem to involve the probability of harm to the plaintiff standing many feet away whatever might be the case as to the owner or to one so near as to be likely to be struck by its fall. If, however, we adopt the second hypothesis we have to inquire only as to the relation between cause and effect. We deal in terms of proximate cause, not of negligence."

I urge you to read the entire opinion as I did years after graduation from law school. Compare Andrews' discussion of the legal issues based on proximate cause with the so-called Cardozo's "foreseeability" discussion. And for eloquent writing, read Andrews' comparison of a river to illustrate the limits of proximate cause.

I acknowledge that the proximate cause theory has its limitations. Andrews' dissent acknowledges it stops somewhere. But certainly the railroad has a duty to Mrs. Palsgraf, a paying customer who was injured due to the negligence of the railroad. This view puts me in good company with a host of commentators and scholars, many of whom puzzle over the rationale of the majority and the dissent. No need to mention them all here. I do however wish to mention my torts professor, Dean Prosser, praised the concise, elegant writing style of both Cardozo and Andrews. But he wondered, as I do, how Cardozo could ignore that Mrs. Palsgraf was a passenger and paying customer. And by the way, the word "foreseeability" does not appear in the Palsgraf opinion. The words "unforeseen" and "unforeseeable" appear in the dissent.

So what is the point of all this? The asterisk is relevant in my previous column, Frankenstein's Monster*. The examples of AI illustrated there and here may be simplistic, not deeply analytical, but they can point one in the right direction. I have heard that some courts are using AI as a kind of research attorney or editor. The alleged purpose is not to write a statement of decision, an opinion or memo, but to set the writer along the right path. I can see how that could be a legitimate purpose. Nevertheless, I am wary. No need to cite the numerous articles in leading journals warning of the dangers of unchecked AI.

Even if the results AI achieves are helpful, its use may foster the atrophy of the human creative brain. If this happens, our ability to think creatively will separate us from Cardozo by light years. And in other contexts, AI can spread misinformation that leads to chaos. I have read that AI is being used as psychological therapists and even as virtual boyfriends. One article quoted some women as reporting that fake men are better than the real thing. Hope that wasn't a quote from my wife.

AI can be of immense benefit and immense harm. And it is not fully understood by its creators who can only conjecture about its potential. AI was developed by humans. It can accomplish some tasks in milliseconds that humans would take days, weeks, or years to accomplish. A segment on 60 minutes involved conjectures about when or if AI becomes conscious of itself. Yikes! I try to show a positive outlook while whistling in the dark. Hence the title of this column "Fear and Trembling" and no asterisk. Frankenstein's monster threatens to kill himself, but in Mary Shelley's novel Frankenstein; or, The Modern Prometheus, we do not know for sure that he does.