top of page

Striving Toward a Legal Trifecta of AI-Based Robo-Lawyers, Robo-Judges, and Robo-Jurors

By Lance Eliot.

You are likely familiar with the word trifecta.

Let’s briefly explore the meaning of a trifecta and then see how a special type of trifecta is being aimed squarely at the legal profession and the practice of law. As a hint, the trifecta I’m referring to consists of infusing AI into the act of performing legal reasoning. More on this in a moment.

We ordinarily use the catchword trifecta for just about anything that seems as going against quite staggering odds. In a specific sense, the initial usage began when trifecta was coined to depict horse racing bets that managed to miraculously choose the first-place winner and the second-place winner, and even the third-place winner, doing so in each respective winning position and altogether in their precise order in a given race.

It is a very hard bet to select and win.

Of course, you can be somewhat argumentative and make the case that sometimes the top three are rather clear-cut choices. A particular race or competition might already have contenders that can somewhat readily be ranked as the top three selections. By and large, you could say that the odds of accurately predicting a trifecta might be more manageable if there are a lot of prior known conditions.

Nonetheless, there is always the opportunity for the vagaries of chance to spoil things, assuming that an overt fix isn’t already in the cards. A trifecta could be rigged if each of the jockeys is secretly paid off or otherwise induced to ensure that the three designated horses arrive in the one-two-three finishing order. In contrast, when a situation is fairly being run and all else is considered equal, the likelihood of picking a trifecta can be extremely difficult and randomness alone might spoil the best of prognostications.

We take it as a base assumption that a trifecta is amazingly triumphant since the bet was made in a milieu of random chance. Out of the myriad of permutations, a trifecta betting winner assuredly got wildly propitious and can try to brazenly claim it was skill alone, though we are apt to believe that their luck happened to strike gold on that particular day and in that exact instance.

Shifting gears, let’s consider another kind of trifecta, one that some believe could eventually occur in the practice of law. The aim of many AI researchers and developers consists of making strident advances in AI-based legal reasoning (AILR) toward trying to replicate into computer systems the legal prowess of those immersed in the legal realm.

You’ve undoubtedly heard or read about the ongoing attempts to devise so-called robo-lawyers, consisting of AILR that can imbue computers with the legal acumen of human attorneys. There is also a lesser-known set of efforts to craft robo-judges. Further down on the list are the attempts to formulate AI-based jurors which are usually referred to as robo-jurors.

The AI-based legal reasoning trifecta that is being pursued comes out this way:

1) AI-based robo-lawyers

2) AI-based robo-judges

3) AI-based robo-jurors

Let’s take a brief look at each of the trifecta choices and also consider why the anticipated sequence consists of robo-lawyers, followed by robo-judges, and at the tail-end, we would have robo-jurors.

Examining The AI Legal Trifecta

The most discussed and likeliest first winner in the horse race toward infusing AI into the law would be the likes of AI-based robo-lawyers. As a side warning, those within the field of crafting such AI are somewhat leery of and find loathsome the catchphrase of robo-lawyer. I use it here with caution and acknowledgment that it is a potential misnomer (the same goes for any of the robo-related nicknames).

So far, trying to create AI that is on par with human lawyering has proven to be a quite challenging problem. The odds are that such challenges will continue to only slowly be overcome. You do not need to lay awake at night over the worry that autonomous robo-lawyers will tomorrow be taking your job. That being said, keep in mind that there are already today lots of opportunities to leverage semi-autonomous AI-based lawyering tools. These can amplify what human lawyers do. Law firms need to be paying attention to the human attorney efficiencies and effectiveness that today’s AI capabilities can achieve when combined with LegalTech and placed into the hands of willing lawyers.

Returning to the aspirations of autonomous robo-lawyers, the early days of such endeavors focused on if-then style constructs. It was thought that perhaps legal knowledge could be represented in a vast series of interconnected logical conditions, centered around the idea of “if something is true then something else is also true” predilections. By exhaustively churning through laws and dissecting them into the appropriate if-then concoctions the hope was that legal beagle smartness would ensue. This was undertaken in the 1980s and 1990s during an AI era that heralded the use of knowledge-based systems and expert systems. Some refer to that time period as the AI symbolics era whereby human knowledge was going to be computerized via symbolic logic alone.

Any human lawyer can pat themselves on the back for how hard it is to mimic or simulate the practice of law via the use of a computerized set of if-then constructs. Getting AI to do the legal wrangling of attorneys is an extremely difficult endeavor and we do not ostensibly know when or if there will be devised AI that can be anywhere on par with human lawyers. Latest efforts have shifted toward using Machine Learning and Deep Learning techniques. The notion is that you might be able to replicate legal reasoning by doing large-scale computational pattern matching of our laws and legal cases. Some refer to this approach as the AI sub-symbolics era since it relies on somewhat arcane sub-symbolic oriented mathematical computations.

Meanwhile, some contend that we should not let ourselves fall into the classic trap of acting as though there is only one way to solve this problem. Continual barbs are being tossed from the AI sub-symbolics camp at the AI symbolics contenders, and likewise vice versa. This kind of AI infighting would not seem especially conducive to figuring out how to best produce AILR. The latest wave of a hybrid nature seeks to combine both the AI sub-symbolics and AI symbolics techniques and technologies and refers to this as the neuro-symbolics AI approach.

What else is there besides the vaunted AI-based robo-lawyer?

Coming in second place is the rise of the AI-based robo-judge.

Yes, the idea is that we would have judges that aren’t only humanoids. These AI-invoked judges might at first be relegated to only lower court levels or perhaps be used in extraordinarily minor legal matters.

In any case, wherever they fit in, there is at least an erstwhile quest to create them.

The third-place position might be quite surprising to you. I say this because there is little talk about this tail-end position. You see, bringing up the final member of the trifecta is the AI-based robo-juror. A bit of a brouhaha surrounds the AI-based robo-juror notion. It is one thing to have a robo-lawyer representing a human client that perchance is AI-based. We can stretch our minds to also somewhat tolerate the use of AI as a robo-judge. But taking that cliff-hanging step into the legal abyss and anointing AI to serve as jurors is beyond the pale for many.

We hold dear the heartfelt and justice-wielding concept of a jury of our peers. Humans sitting as jurors in trials involving humans that are facing adjudication seems ostensibly the only proper and judiciously sound way to do things. Some would fervently argue that it is entirely nonsensical to have AI as a juror, whereby a human defendant is dependent upon a trier of facts as a so-called “peer” that is not even a human being.

Anyway, we’ll have to wait and see how that pans out. Perhaps we will be willing to accept some form of AI augmentation for human jurors. The idea is that jurors could make use of an AI jury-augmentation system during the course of a trial. Jurors would primarily use such an AI system while in the jury room and undertaking their deliberations. Whether this opens Pandora’s box of legal troubles and outcries of jury tampering or similar concerns is an open question.

We can take the robo-juror a step further and suggest that maybe an AI system would sit on a jury and become the veritable thirteenth juror (a common phrasing due to juries often having a dozen jurors). Some earnestly argue that a suitable AI system might be devised as an intellectual peer of human jurors, despite lacking the other attributes of being human such as common-sense reasoning and akin attributes (that’s one of those potential legal loopholes that could be contested).

Now that we’ve discussed the anticipated trifecta, let’s take a moment to further reflect on the matter.

You are for example welcome to debate the sequence or order of the claimed winners. For example, perhaps we are able to arrive at AI-based robo-judges sooner than we can attain AI-based robo-lawyers. In that case, the top two contenders in the aforementioned sequence would need to swap positions.

Another consideration is the desire to summarily reject utterly the AI-based robo-juror construct. Take it out of the list, you might exhort. Indeed, skeptics and critics find the idea so ridiculous and impossible that it should not be ranked as a contender in this legal horse race. That means that the trifecta is busted since we are presumably down to just two horses (robo-lawyers and robo-judges).


Though there is perhaps a smattering of outsized speculation underlying this trifecta deliberation, you can find some salient hidden points that are worthy of mindfully mulling.

Consider these mindbending twists:

  • How will society react to the anticipated AI infusion into our judicial activities and processes?

  • Will this AI be subject to the same rigors of legal practice and ethics as humans are?

  • Would the AI used for one of the three be repurposed into the other two?

  • What means would be used to ensure that the AI stays within proper legal bounds?

  • Etc.

A final thought is that whatever you might argue is the sequence for the trifecta, including those that reject the trifecta as ever feasible at all, anyone putting down some dough as a bet on this is facing some mighty tall odds. But that’s the way humans are, always aiming at the stars.


About The Author

Dr. Lance Eliot is a Stanford Fellow at Stanford University as affiliated with the Center for Legal Informatics, a co-joining of the Stanford Law School and the Stanford Computer Science Department.

He is also Chief AI Scientist at Techbrium Inc., has been a top exec at a major Venture Capital firm, served as a global tech executive at several large firms, previously was a professor at the University of Southern California (USC), and headed an AI lab there.

His columns have amassed over 4.5+ million views including for Forbes, AI Trends, The Daily Journal, and other notable publications. His several books on AI & Law are globally recognized and highly praised.

bottom of page