Imagining a Role for Artificially Intelligent Juries

Screen Shot 2017-07-07 at 11.22.49 AMby Judge Kenneth C. Mennemeier, Jr. and Kelly A. Mennemeier

Kelly A. Mennemeier is a 2016 graduate of Northwestern University Law School. She currently serves as a law clerk for Justice Peter J. Maassen of the Alaska Supreme Court. She can be contacted at

Kenneth C. Mennemeier, Jr. is a judge on the Sacramento County Superior Court. He can be contacted at

Could artificial intelligence (“AI”) be applied to the fact-finding responsibility traditionally given to juries? The possibility invites reflection on what would be lost, and what could be gained, if our justice system moved away from using juries comprised of human beings and instead used computerized “juries” applying artificial intelligence.  This essay shares some reflections.

Our jury system is subject to a number of criticisms, which may give rise to temptation – and perhaps even demands – to use this new technology to achieve better results and greater efficiency. Indeed, critics ascribe several flaws to jurors, jury performance, and the jury trial system that AI may be able to correct. For instance, some critics argue that jurors bring implicit biases into their service and that those biases infect their analysis. Bias in favor of law enforcement might make it difficult for jurors to fairly judge officers accused of brutality and excessive force. Bias against a person of color or a person with uncertain immigration status might impair a juror’s capacity to view evidence impartially.

Our jury system is subject to a number of criticisms, which may give rise to temptation – and perhaps even demands – to use this new technology to achieve better results and greater efficiency

Other critics insist that juries, comprised of people selected from voter and/or DMV rolls, lack the capacity to understand complex matters, such as complicated financial instruments (e.g., interest-rate swap futures contracts) or sophisticated technological issues (e.g., patent law disputes). Others complain that, even in straightforward matters within the comprehension of all, the jury-specific parts of the process (such as voir dire, jury selection, jury instructions, and jury deliberations) add untold hours to a trial and, with it, substantial and perhaps avoidable expense. Similarly, some observe that lengthy trials may result in fatigue, leading to lapses in jurors’ attention. Still others complain that jurors are vulnerable to having their passions enflamed by the skillful telling of a sympathetic story, thus yielding runaway verdicts and unjust awards. And then, for better or worse, there is the possibility of jury nullification – i.e., a jury refusing to adhere to a law with which the jurors disagree.

It is not hard to imagine computer programs obviating most, if not all, of these concerns. Use of a computerized “jury” programmed with AI, including a compilation of all statutes and case law, would surely shorten the trial process by eliminating the time spent in voir dire, jury instructions, and jury deliberations. This could generate huge efficiencies and corresponding economic savings. Using a programmed AI “jury” would also address the defense lawyer’s concern about a jury becoming impassioned or enraged by the effective telling of a plaintiff’s story, and an AI “jury” would never fail to absorb and assess any evidence due to fatigue or boredom. Using an AI “jury” would similarly address the concern about implicit human biases, as a computerized fact-finder could be programmed to be color-blind and otherwise oblivious to the varied facts and circumstances that give rise to bias. And, a computerized fact-finder would always remain faithful to the law, eliminating the risk of (or, for some criminal defendants, any hope for) jury nullification.

Whatever AI’s promise may be for other aspects of the legal profession, AI programming will never completely obviate the desire for a jury comprised of real people chosen from the community.

At first blush, then, just as the prospect of driverless vehicles holds out a promise of safer roadways, so too might the prospect of artificially-intelligent computerized “juries” hold out a promise of more consistent, fair, and just verdicts. But before jumping to conclusions, it is important to pause to consider the history and core values underlying our jury system.

In criminal law, the importance of the constitutional right to a jury trial is highlighted by, among other things, the simple fact that it is the only right that appears twice in the United States Constitution (in both Article III and the Sixth Amendment). In civil law, the right to trial by jury is enshrined in the Seventh Amendment. Moreover, the constitution of every state protects the right to a jury trial, at least in criminal cases.1 The California Constitution recognizes trial by jury in both civil and criminal cases as an “inviolate right.”2

It is also useful to reflect on the core values served by our jury system. For one, a jury of 12 (or six in some civil cases in federal court3) is meant to represent the “community.”4 Comprised as they are from people randomly selected from DMV and voter registration rolls, juries are drawn from a cross-section of the community and, as a body, become the “voice” of the community.5

Perhaps more important, the jury is a critical component in a system of government built on a separation-of-powers premise. In a system with three branches of government, juries stand apart from each branch. Each jury serves for a limited period of time. Each serves a role separate from the legislative bodies that make the law, the executive (and prosecutorial) authorities who enforce the law, and the judicial officers who interpret the law and oversee the courtroom. The jury represents an authority detached from government. That detachment honors the distrust the nation’s founders had for a system in which a centralized governmental authority serves as both prosecutor and judge.6

The virtues of a human jury, then, suggest a limit on the potential for AI “juries.”  Computerized, programmed, and artificially intelligent finders-of-fact might someday address and obviate some or all of the criticisms cast upon the jury process, but it is difficult to envision how they could serve the core values juries presently serve. Specifically, it is hard to imagine a computerized fact-finder ever being perceived as a jury “of one’s peers.” It is even harder to imagine how an AI jury, programmed by hidden, anonymous forces, might ever be viewed as separate from centralized governmental authority. All of this suggests that, whatever AI’s promise may be for other aspects of the legal profession, AI programming will never completely obviate the desire for a jury comprised of real people chosen from the community.

This is not to say there is no future for AI juries. The use of computerized, programmed finders-of-fact could shorten trials, delivering value to those who demand a trial but seek a more time-efficient and economical process. And, if AI “juries” reduce or eliminate hidden biases and the risk of impassioned jurors, and/or demonstrate the ability to navigate complex fields of fact, technology, and law, they could prove an appealing dispute resolution option in certain types of cases.

We envision an important role for AI fact-finders in the future. But, because artificially intelligent finders-of-fact can never serve the core values that human juries serve, we predict that artificial intelligence will never fully supplant the need and desire for human juries.

1 See Duncan v. Louisiana, 391 U.S. 145, 153 (1968).

2 Cal. Const., art. I, § 16.

3 Fed. R. Civ. P. 48.

4 Smith v. Texas, 311 U.S. 128, 165 (1940) (“It is part of the established tradition in the use of juries as instruments of public justice that the jury be a body truly representative of the community.”).

5 BMW of North America, Inc. v. Gore, 517 U.S. 559, 600 (1996) (describing the jury as “the voice of the community”); Wainwright v. Witt, 469 U.S. 412, 458 (1985) (same).

6 See Duncan v. Louisiana, 391 U.S. at 155-56.