AI & Law: Legal Doctrines And AI

Image for post
Image for post
Ascertaining how AI will “learn” legal doctrines

by Dr. Lance B. Eliot

For a free podcast of this article, visit this link or find our AI & Law podcast series on Spotify, iTunes, iHeartRadio, plus on other audio services. For the latest trends about AI & Law, visit our website

Key briefing points about this article:

  • The law consists of numerous and vital legal doctrines that are invented rules-of-law
  • One well-known instance was first derived via a court case of the wagon and the donkey
  • AI is anticipated to gradually and inexorably become adept at AI-based legal reasoning
  • An unresolved question entails how such AI systems will be imbued with law doctrines
  • By exploring the doctrine rising the wagon and donkey we can gauge Machine Learning (ML)


There is a quite legendary court case that entails a wagon and a donkey, dating back to 1842 and become a foundation for a legal doctrine that was eventually transferred from English law into American law. This law-expanding case involved a horse-drawn wagon that rammed into and sadly overran a donkey.

Can you name the case?

Can you also name the area of law that was forever changed due to the case?

Take a moment and search your mental legal corpus to try and ferret out what the answer is. A bit of a puzzler, perhaps, so let’s proceed into the details of the case and then provide a grand reveal, though you might otherwise have sooner guessed the crux of the matter once you have been given a deeper set of clues.

Here’s the saga.

A stagecoach style wagon was being driven by a man that was employed by a wagon owner. The wagon was rumbling along on a dirt road that was relatively well-marked and had delineated boundaries marking either side of the roadway. The driver was seated in the wagon and sat behind a team of three horses that were pulling the wagon.

So far, those facts seem to be clear cut and lacking in dispute, thus take them as a given.

A donkey had been tethered to a post that was nearby the road. The man that owned and tethered the donkey stepped away and was not immediately present to continually watch over the donkey. Since a donkey is a donkey, and they oft wander to-and-fro, the donkey opted to meander as far as it might, subject to the limits of the tethered rope.

Those facts also seem to be without any discord in this legal case.

Unfortunately, things take a turn for the worse.

The donkey ends up standing in the roadway, still tethered, but having wandered nonetheless into an active thoroughfare. The wagon comes along. Alas, the driver apparently does not see the donkey and therefore does not take any action to try and stop the wagon, nor steer away from the donkey. The team of charging horses’ rams into the donkey, knocking the animal to the ground. The donkey gets trampled and dies soon thereafter.

The owner of the donkey, the plaintiff in the subsequent legal case, sues the owner of the wagon, the defendant, and contends that the wagon was being operated inappropriately and that the actions of the owner/operator-led indisputably to the death of his donkey.

Okay, you’ve now been provided with the foundational facts, though I am admittedly somewhat cheekily holding out a few finer details and those will certainly tip you to the root of the matter.

Here is a recap of the declaration as filed for the case:

“The declaration stated, that the plaintiff theretofore, and at the time of the committing of the grievance thereinafter mentioned, to wit, was lawfully possessed of a certain donkey, which said donkey of the plaintiff was then lawfully in a certain highway, and the defendant was then possessed of a certain wagon and certain horses drawing the same, which said wagon and horses of the defendant were then under the care, government, and direction of a certain then servant of the defendant, in and along the said highway; nevertheless the defendant, by his said servant, so carelessly, negligently, unskillfully, and improperly governed and directed his said wagon and horses, that by and through the carelessness, negligence, unskillfulness, and improper conduct of the defendant, by his said servant, the said wagon and horses of the defendant then ran and struck with great violence against the said donkey of the plaintiff, and thereby then wounded, crushed, and killed the same.”

Even just a quick glance at this legal predicament suggests that this case revolves around a question of negligence, which is ostensibly abundantly obvious as we have someone that presumably has been negligent in their actions and ergo led to the death of the donkey.

You might be eyeing the defendant as being negligent because, well, the operator of the wagon undeniably ran over and killed the donkey. On the other hand, you might be looking sternly at the plaintiff, since he tethered his donkey in a locale that was fraught with danger and did not seem to exercise due care for his beast of burden.

Another viewpoint is that perhaps they are both equally at fault, is a type of regrettable status quo condition, whereby the donkey owner should not have tethered the donkey such that it could wander onto the path of the wagon or that he should have stood to watch and been there to prevent his donkey from getting into an ill-advised quandary. In that light, you might also find equal fault with the wagon driver for not having stopped or averted hitting the donkey, and thus cancel out the two separate but intertwined wrongs as though they were unlucky counterbalancing forces of nature.

Time To Reveal The Case

You have assuredly summarized that this case has to do with tort law and the role of comparative negligence versus contributory negligence.

In particular, what new doctrine was borne out of this case?

If you are wavering back-and-forth about the potential actions of the wagon driver and the actions of the owner of the donkey, here’s an added facet of the case that might sway you and be construed as a pivotal element. I’ll also name the case, it was Davies v. Mann (152 Eng Rep 588, 1842).

During the trial, it was indicated by a witness that the wagon was coming along at a “smartish pace” and this seemed to measure heavily in the minds of the court, and eventually likewise pivotal for the appellate court that heard the case after it was appealed. At trial, the defendant was found to be negligent, based on the logic that if there had been a semblance of ordinary or reasonable care exercised then the donkey would not have been killed (i.e., the alleged smartish pace was akin to an overburdening leaden weight pushing voraciously down upon the scales of justice).

It was theorized that the wagon driver had a last clear chance to avoid the ramming of the donkey yet had failed to undertake that chance. Of course, today we know this as the famous “last clear chance” doctrine, sometimes also referred to as the last opportunity rule.

For some final details about the famous or perhaps infamous case of the wagon and the donkey, there was upon appeal the question raised as to whether the donkey was illegally in the roadway and thus this made the stance for the defendant even stronger and surely weakened the position of the plaintiff. But, surprisingly (or scandalously), this assertion had not been apparently raised as a notable point of contention at trial by the defense.

The appellate ruling proffered that whether the donkey might or might not have been legally or illegally placed was essentially immaterial, and concluded therein: “All that is perfectly correct; for, although the [donkey] may have been wrongfully there, still the defendant was bound to go along the road at such a pace as would be likely to prevent mischief. Were this not so, a man might justify the driving over goods left on a public highway, or even over a man lying asleep there, or the purposely running against a carriage going on the wrong side of the road.”

The notion of the last clear chance doctrine came squarely into U.S. law during a case in North Carolina of Gunter v. Wicker (85 N.C. 310, 1881). Was it another instance of a wagon and a donkey? Nope.

This case in 1881 involved a man that was working in a sawmill and had been employed to do several tasks including the oiling of the equipment. At one point, a saw was purposely stopped to allow for the oiling activity, the man entered into the flywheel to provide an oil coating and horrifically was injured when the owner/operator at the sawmill turned on the steam and for which caused serious injuries to the oiler.

At trial, the oiler as the plaintiff contended that the sawmill owner/operator was negligent in having turned on the steam while the oiler was inside the machinery. The owner/operator as the defendant argued that the oiler had put himself into danger by maneuvering into the flywheel and that such posturing was unnecessary to do the oiling task.

The ruling made was that despite acknowledging that the oiler had put himself into harm’s way, possibly even needlessly so, nonetheless the steam would only become injurious due to the actions of the defendant by having turned on the steam at the wrong time.

In short, the key takeaway was that the sawmill owner/operator had the last clear chance to avoid the injurious incident but failed to exercise ordinary or reasonable care in doing so. A probing legal analysis of the last clear chance doctrine was conducted in 1926 by Matthew Myers and published in the North Carolina Law Review (see volume 5, number 1). Though the doctrine was still relatively nascent at the time, it had already been used in a multitude of court cases regarding railroad accidents. According to Myers: “The North Carolina Court holds a defendant liable not only where he has the last clear chance and knows of it, but where he has the last clear chance and doesn’t know of it, but would have known of it if he had not been negligent.”

Today’s modern world certainly has plentiful examples of legal cases that dovetail into the last clear chance doctrine. Rather than a wagon and a donkey, envision a driver of a car that strikes a pedestrian crossing the street. If that isn’t modern enough for you, consider the instance of a so-called self-driving car that struck and killed a pedestrian in Phoenix, Arizona just a few years ago, and ask yourself whether the last clear chance doctrine will come to play in a world filled with AI-based autonomous vehicles.

AI In The Law And The Role Of Doctrines

Speaking of AI, let’s take a moment and reconsider the last clear chance doctrine in a completely different light.

In a prior article, I posited that we will eventually see that AI will become infused into the law via advancements in autonomous levels of legal reasoning. AI systems will be able to aid lawyers and attorneys via serving as an over-the-shoulder legal oriented advisory tool.

Inevitably, this capability will be further improved such that the question will arise as to whether the AI-powered legal reasoning can effectively practice law, as it were, doing so without any human lawyer being involved. This is an open-ended matter and we have yet to witness AI of that caliber, though this does not necessarily imply that we aren’t headed down that path.

For details about the latest trends entailing AI and the Law, see my textbook entitled “AI and Legal Reasoning Essentials” at this link here:

Ponder the next question as it will be the basis for a macroscopic perspective on how AI will become ingrained in the law and possibly become an active legal reasoner.

How would an AI system become versed in the last clear chance doctrine?

Seems like a relatively straightforward question. There is no denying that we would expect any fully enabled AI legal reasoning system to be able to grasp the nature of the last clear chance doctrine. It is assuredly a topic covered for budding human lawyers and typically tested as part of any bar exam. Any attorney worth their salt should have some familiarity with the doctrine, regardless of whether it is avidly utilized in their chosen specialty of law.

Some might suggest that the AI could use Machine Learning (ML) as a means of coming up-to-speed about this particular doctrine. For clarification, today’s notion of Machine Learning is really a mathematical approach that uses computational pattern matching to try and find discernible patterns in data. Think of a statistical tool such as multiple regression and that’s generally the same idea. I point this out because some have ascribed mystical qualities to Machine Learning, as though it is akin to the way that humans learn, but this is a misleading and astronomically overstated exaggeration of what this type of Machine Learning can do.

Okay, so we need lots of data to feed into the pattern matching, and out of which will hopefully arise identified patterns.

Without going into a mode of reductio ad absurdum, pretend that we were able to collect hundreds or possibly thousands of legal cases encompassing a wagon and a donkey. The data consisted of the raw facts of the case and the resulting decision. Assume that the last clear chance doctrine is not specifically called out. Your only recourse to discerning the doctrine would be by the implied pattern of the facts of the case and the outcomes of the cases.

Does it seem reasonable to expect that the pattern matching would eventually land onto a doctrine, devised by its own pattern detection, which otherwise resembled the last clear chance rule?

That’s a hard one to answer with any certainty.

It is possible that the raw data could be analyzed mathematically to the degree that a mathematical result would approximate the last clear chance doctrine. On the other hand, there might other patterns discerned, some that might be contrary to the last clear chance or be somewhat byzantine and defy any everyday logical explanation.

Think of this as though a computer has analyzed zillions of chess games and found computationally complex patterns that might not readily be translatable into an overarching strategy of how to play chess. Thus, you end-up with an inscrutably good, automated chess player that frustratingly, for us humans, cannot explain why this is so.

This then brings us to the crux of why I’ve brought up the entire discussion about the last clear chance doctrine. It wasn’t conveyed as simply an engaging rehash underlying the history of the doctrine, nor was it intended as any kind of revelation about this now-classic longstanding legal rule. The point here is more so about how we are going to approach the application of AI into the law, along with the nature of how this anticipated AI will be suitably able to inject legal reasoning and the law.

Machine Learning As A Doctrine Seer

One approach being heralded are the latest techniques underlying Machine Learning.

Though this bodes for an exciting promise, there are inherent limitations, some say dangers, whereby the AI so crafted would not find the patterns that we humans have already invented, such as the last clear chance doctrine. Furthermore, the AI might seemingly detect such patterns and then use those to act upon the law, performing the services of an attorney or a judge, and yet we might not have any viable means of logically ascertaining how the AI is performing those legal tasks.

The AI could conceivably come up with a combobulation doctrine, something that either is an oddball rule that defies rational explanation or might ironically reveal a new and ostensibly useful doctrine that humans had not heretofore devised.

There is an ongoing debate among legal scholars and legal engineers about using an explicit approach toward arming AI with the law and legal reasoning versus doing so via an implicit avenue. In the explicit technique, you expressly program the AI with the legal rules, and those are considered ingrained or the DNA as it will of the AI system (be cautious though in overstepping such an analogy, we ought to avoid anthropomorphizing the AI). Others contend that this is a grueling path, taking too much time, and would involve a lowest-common-denominator level of granularity that we might find intractable and impossible to ultimately attain for any proficiency across-the-board about the law. Thus, they would propose the Machine Learning angle be leveraged instead, allowing the AI to somewhat train itself, as it were, rather than being fed inches at a time by a human hand.

Some refer to this in the AI field all told as the ongoing and acrimonious debate or battle between the symbolics and the sub-symbolics.

The symbolics consists of those that tend toward being explicit and programming the AI to do its desired tasks (recall the earlier days of AI and the use of expert systems and knowledge-based systems), while the sub-symbolics insist that the AI has to train itself (via the use of Artificial Neural Networks and the likes of Machine Learning and Deep Learning). Per any field or specialty, some sit at the extremes of those two poles, while some prefer to try and blend the two positions.


In the future, I suppose that if any of us ever perchance leave our donkey tied-up nearby to a road, and a wagon comes along, we would hope that the AI adjudication system would be versed in the last clear chance doctrine. Since that is a futuristic scenario, substitute a self-driving car for the wagon, but I guess we can retain the donkey in that setting since we are going to hopefully still have donkeys, as cantankerous as they might be.

Of course, per the legendary proverb, one should never stand before a judge and nor behind a donkey.

For the latest trends about AI & Law, visit our website

Additional writings by Dr. Lance Eliot:

And to follow Dr. Lance Eliot (@LanceEliot) on Twitter use:

Copyright © 2020 Dr. Lance Eliot. All Rights Reserved.

Dr. Lance B. Eliot is a renowned global expert on AI, Stanford Fellow at Stanford University, was a professor at USC, headed an AI Lab, top exec at a major VC.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store