The Future: An Efficient Justice System Run By Artificial Intelligence
The Future: An Efficient Justice System Run By Artificial Intelligence
I should preface this article by saying that I don’t believe in the current justice system or how it is designed. For starters, I don’t believe we should be judged by a jury of our peers, but rather a council that knows and has studied the law. I will talk about how artificial intelligence will have a dominant role to play in the future of the justice system, but first I want to talk about how I would change the justice system today if I could. I’m hoping this will give you a better perspective on my thinking and where my belief is coming from.
The average person in the United States is not knowledgeable and unbiased enough to fairly deliberate a judgement. In addition to this, it is fact that a single strong minded juror can sway the opinion of multiple other jurors in a short period of time, despite how illogical or wrong their reasons might be for voting one way. If I was the person in charge of setting up a trial system today, I would take the following three steps. One, end trial by your peers and replace them with a council of judges to have better representation from those who understand the law. Two, remove the defendant’s and the victim’s name and ethnicities from all documents pertaining to the case in order to deter racial or economic bias. Three, remove everyone from the courtroom except the lawyers and judges with testimony to be filmed and transcribed at a different location. Defendants, victims, families and witnesses for both sides would sit in another room and watch the lawyers and judges deliberate over the case. In my opinion these steps would help the vast inequality in the justice system today caused by bias. While this really isn’t realistic and will most likely never happen, there is a possibility that might happen in the future, an efficient justice system run by artificial intelligence.
As the technology of AI continues to get better and better, there will most likely be a turning point in the next 100 years where it will become sufficient enough to analyze a case and testimony far better than any human being could do. Computers, even artificial intelligence systems that are programmed to follow a strict set of rules, by their nature can’t be biased. All they would see would be the data put before them to analyze. Data would rule king and hard evidence would be weighted heavily in a deliberation. There would be a learned algorithm by the artificial judge that could look at both arguments and make a decision. Their judgement would be void of emotions that often can skew the thinking of judges and jurors.
Some may say emotions are a large part of trials because you can read humans by looking at them and have a hunch that they are guilty. However, the reality is that if we look at many fields where hunches and feelings are what people claim brings them ahead, in many of those cases computers end up out performing them. For example, in the finance industry where hedge fund managers and mutual funds hold much of the population’s retirement funds, we are starting to see competing computer algorithms that are doing as well or better than the competition to the magnitude of tens of percentages and this is only the beginning. As artificial intelligence systems become more and more advanced we are going to see many industries that are disrupted primarily because computers will be able to do the job much more efficiently and just plain better than a human being.
Some of the less obvious advantages of having artificial intelligent computer systems running the justice system comes just in pure practicality. For example, computers don’t need to take breaks and can process cases twenty-four hours a day, seven days a week without stopping. With a justice system that is currently awaiting tens of thousands of trials and with jails housing thousands of prisoners for months, just awaiting a trial, AI could massively cut down on the backlog. Artificial intelligence also can’t be bribed by advancement or other perks should one of the parties win over the other. With recent scandals in the news with judges taking bribes, or getting throwbacks for sending people to jail, this would alleviate this problem entirely. Lastly artificial intelligence wouldn’t be able to be blamed for the decisions they make because they are making them by data analysis.
All these reasons add up to why I think that not only will we be seeing artificial intelligence running the justice system in the future, but I believe it will be running it far better than how it is currently run. In delicate situations like the justice system, where someone’s life could literally be ruined in an instant, there is no room to leave human error and bias to chance. We must make a change that is more efficient and better performing in order to increase the welfare of our citizens. To achieve these outcomes, in math and science we can trust.
-Calaber24p
Awesome.
I've for a long time hated the idea of jury deliberation. As you say, a strong willed juror can influence others. If they can all be influenced by each other, you dampen the point of having multiple decision outputs. It's closer to just having one person. Or possibly worse since someone who is good at talking people into things might be bad at judging accurately.
I think an easy improvement that's available to us already is that each juror/judge decides on their own, with no communication at all with the others, and then they're rewarded for deciding in the majority. (Since you have no idea what the others will do, judging what seems to be true is your only strategy.)
Moving forward, you're probably right that AI can do a better job.
You could have prediction markets on Augur that trade shares on the outcome, and any time the final prediction on Augur is different than what the AI decides, you could consider that cause for a retrial, perhaps with a different AI system.
Or maybe there should be an oracle of humans (using blockchain stuff obv) who decide on the case first, and then AI systems go after that, since they aren't biased by what the humans already decided.
Another aspect of this that I'm looking forward to (when we move beyond using punishment to resolve crimes) is decentralization of how we deal with the criminals. If there's ambiguity over whether they did something, all that's really important is that everybody has access to all the info. You'd know that the oracle decided this, Augur predicted that, the AI said this. (You can watch the trial if you want.) And you can take it all for whatever it's worth to you. And others will do the same, and some type of consensus will emerge. We get there without a black and white decision issued by a central party.
This sets up sort of a kensyian beauty contest, don't you think?
I don't think so, at least it's less than what exists now.
Take an extreme example of overwhelming evidence, like it's caught on video and it's clearly the person or whatever. If they use fancy language or they're really pretty or something, you'll still know the other jurors won't be fooled by it and will realize they're guilty.
Maybe at the margins there's always the aspect of being biased towards certain people. But that's true with deliberation too.
What's important is that your motivation is always to determine what actually happened. There's really no way for you to guess when the other jurors will be biased, because that's such a nuanced and subtle thing. If you could determine that, then you're some sort of highly intelligent AI type of thing (and then so are the other jurors, and then there's no more bias because you're all so smart that you don't fall for that stuff, so you're still back to just judging what happened).
It's probably important to have some sort of mechanism to select people who are good at being jurors. But this is true anyways, with juror deliberation.
I'm not sure you understood what I was talking about. A Keynesian beauty contest means that the jurors would be judging the accused according to what they believe the other jurors believe (instead according to what they themselves believed) https://en.wikipedia.org/wiki/Keynesian_beauty_contest
So for example, let's say you are on a jury. You personally believe the defendant is innocent. But you also think that the other 11 jurors think he is guilty. The system you're talking about (where jurors are rewarded for being on the winning side) would incentivize you to vote guilty in order to be on the winning side.
I realize youre talking about them having no contact with the other jurors, but they could still speculate. For example, if you're in a really racist state, and the defendant is black, you might not be a racist yourself, but you can assume many on the jury are. You would be incentivized to vote guilty in order to get the reward for being in a majority
That's what I thought you meant, I maybe didn't explain my answer too well.
I'm saying that judging how the other jurors might mis-judge the case is kind of like an intellectual high wire act that you can't accurately do in practice.
I think you'd be more likely to stumble over your own feet and go outside of the majority if you start speculating on how others will mis-judge certain info. (It's hard to really guess what anonymous persons' biases are.)
The extreme case where the other jurors ars likely to be racist... they'd have been racist anyways if they were allowed to deliberate. The problem is bad jurors, not the mechanism of trying to vote in the majority.
Maybe there could be some kind of appeal process, where if one person dissents (or if enough people dissent) it gets reviewed by some higher level oracle, and if they decide the one person was right, it's a very high reward for that person. So then there's some incentive to hold your ground even when you know the other jurors are biased.
I think you'd ideally have some way of determining who the best jurors are. You could use the best ones for the biggest cases, and more important as you weed out the worst ones, there's less and less people acting racistly, and less fuel for the beauty contest to happen.
At the end of the day I just cringe so hard at the idea that random people are summoned, against their will with no incentive to do it correctly. Lol. There has to be SOME glaring improvement to that. Intuitively I'm pretty confident that my method works but I'm not sure how well I made the case. Happy to field more questions.
Well said, I really like your idea on separating jurors from eachother "I think an easy improvement that's available to us already is that each juror/judge decides on their own, with no communication at all with the others" The jury should cast votes like the supreme court would and the majority wins.
I believe AI will do a better job in the long run, but I still do think we need to make many changes that would help people today. Sadly the government doesn't take fixing age old systems well.
Hehe ya, the people who profit from it now would be out of jobs if we flipped the pancake to something that actually works :p
AI justice already exists.
Just think about penalty for speed.
Camera makes a photo and send you a notification.
A huge part of justify duty is just an algorithm.
Precedence law -- human find a solution, justice system multiplies.
@calaber24p
The future will most likely not have any primitive concepts such as judges, priests and politicians. All laws and crime exist because of inefficiencies in current technologies.
Very true. It's pretty strange to think that the majority of horror and injustices in the world can literally be fixed through applying the right technology.
@kyriacos
Youre probably right and like flan said its a strange reality to even comprehend, but I think that would be far past the time of AI technology. Your talking about type II or Type I civilizations here.
Hi @calaber
I like your idea of the judges being in a different location than the witnesses to prevent biased judgement based on race etc.
I also like the idea of AI doing the judging. I just think it would be difficult to put the evidence into a data base for the AI to judge, but I am sure the IT guys will find a practical way. The problem I see is that to judge the facts, certain algorithms will have to be written that give specific weights to certain criteria to serve as mitigating or aggravating circumstances. If people know these algorithms they will also be able to influence the judgement
The magic of true AI is the data wouldnt even need to be put into a data base, the system would be able to just view the data and process it itself. The algorithms would be basic but would learn and evolve over time theoretically. Another idea is that they would be so basic and open that both sides would actively try to influence the judgement. For example it weighs positive dna analysis more than it would weigh a witness testimony, so both lawyers would essentially try to bring hard evidence to the table rather than relying on witnesses who much of the time are flat out wrong or confused anyway.
I think a jury of our peers, as it was intended originally, where our peers are people who actually know us and can provide in depth perception into the matter of our character, was a great idea.
I also believe in jury nullification where not only is the defendant on trial but the very law itself. If the jury believes the law to be out of line, they can vote not guilty even if the defendant is clearly guilty of breaking that law, simply because they do not believe in the law.
I love your thoughts, and that you have this opportunity to share them, but I've got to disagree with you on this one.
What we need is a restoration of an informed populace that can apply the original judicial system intentions, which included a jury of our real peers and the option of jury nullification.
I would rather have my actual peers determine if the law is valid, and if so, if I'm guilty of it, rather than an artificial intelligence.
I disagree with you on this one as well :) but I see where you are coming from. However in the small amount of cases where juries will not convict because of their beliefs, there are many more that do choose to convict because of human error or bias. I think the idea of AIs running the justice system would take a very long time and a very tough adjustment to make , but I truly believe that at the level of AI I am talking about, the systems would be near human in emotion, compassion and ideas, just with a better analytical skill set.
I certainly understand where you are coming from, and I get it. It's just that personally I don't trust it, since man has to make the AI in the first place. Programs can be tweaked and modified based on the preference of the one creating or supervising it, which I think can be dangerous. If AI judges could be proven to the best method, and everyone agreed, then it would be easy for those in control of them to modify them to work the favor of their personal preferences and beliefs. Once they were allowed and the human judges were taken out of the picture, there really would be no turning back.
Ultimately, "justice" seems to be too relative down here anyway, and it will always be skewed, which is no reason to give up, but most likely will always remain true nonetheless.
At least we can agree to disagree on this one. Thanks for the post and the reply @calaber24p!
I have participated in the MOOC AI from Stanford. At the moment I would not trust AI to be able to make the final judgement in court but it could be a potential aid. However AI heavily depends on the state space that it draws it conclusions from, what is not in the state space it cannot see. This will limit their judgement and they cannot deal with totally unforeseen items. @steempowerwhale 🐳
🌞 upvoting your lifetime dreams!
The Jury system is very specific to the US 'tradition' (It reminds me the great movie "12 angry men"), we don't have a similar system in France (where it is one or several judges depending on the case who make the final decision).
Beyond technology and AI, one key issue for me is that the amount of money one can put on the table to defend oneself and hire good lawyers will have a real impact on the outcome of any judgement. What is true, what is right should not depend on the money.
Hi @jako
Well said!!
You've got some really good ideas here. My Master's is in Criminal Justice, although I focused on domestic security issues.
I especially like the idea about videotaping witnesses outside the courtroom; we have done this to protect child witnesses in the past, and I think it works. Just make sure there is a judge and the defendant's lawyer in the process as well as an advocate for the witness. We allow the defendant to confront the witness, but it should be understood that a lawyer is the proper agent for this process.
You have obviously put a great deal of thought into your analysis, but here are some things to consider as well.
Studies have shown that juries and judges (human) have the same statistical rates of error in assigning verdicts, I don't have the sources to verify that, but that was something I remember as having been surprised about in my study.
One reason that we use judges instead of juries is that they make value decisions, not data decisions...should we be merciful? or does a guilty defendant deserve to to be punished to the full extant of the law. While some of these factors can be coded into a judging database, I'm not sure that all can.
Moving on, using analytic databases to make decisions can be problematic. Stock market runs are often triggered by computers making sale decisions set on preset standards that are not valid for the circumstances at the time of the run, leading to market crashes.
However, you are dead on on noting that politics, human bias, and human ignorance negatively affects the criminal justice system.
Perhaps an AI or analytic system could be of use in confirm a human judge or jury's decision. Just be aware that regardless of whether a judicial decision is accurate or not, politics and bias can still affect the public's acceptance of the decision, and I suspect that there is a bias against having a "computer" making human decisions.
In any case, good work, and I look forward to hearing more from you!
This is very interesting to read, im really surprised as well that judges would make the same amount of mistakes as jurors, possibly because the mistakes are caused by perception bias ( for example perceiving someone as "looking like a rapist" or "looking like a theif") that exists in both judges and juries. I agree with most of the other points you make in regards to today's AI, but AI in the next 100-200 years will be unrecognizable from anything we could imagine. It wouldnt a computer taking orders, it would be more like an incredibly intelligent and efficient life being. I guess only time will tell, and most likely more time than you or I have left on this earth, but just thinking about it makes me wonder about what else can be possibly changed in the future. I appreciate the response and if ever write anything similarly about the justice system or criminal justice, let me know!
I'll do that. I have been lazy in my writing so far lol.
I think that you're right in that juries and judges make the same mistakes due to the same basic biases. As humans, we are more emotional and/or heuristic in our decision-making than we like to believe. Huemer goes into this in detail here:
http://www.owl232.net/irrationality.htm
and one point in Huemer's argument is that people that are more intelligent and educated often make LESS rational decisions than they believe they do. (cue Homer Simpson burning down his own house and singing I AM SO SMART, I AM SO SMART)
The good thing about the future is that we can always work towards making it better for ourselves via technology and self-improvement!
You realize that juries don't interpret the law, right. Judges do. Juries are deciders of fact. The only thing a jury decides is "did he really do it?" whether or not "it" is legal is a question for the judge (who answers it by giving jury instructions."
I think this is an idea that has some merit. But there are also a lot of problems that you really never get into. To be honest, i think your case would be far stronger for getting rid of judges and replacing them with AI.
For example, can an AI understand human emotions? The emotions and feelings of a defendant (his state of mind if you will) are very important to deciding innocence or guilt in many cases where something like self-defense or alleged consent is involved.
Also, the human touch (call it empathy, emotion, instinct whatever) is very important in weighing the credibility of testimony. What happens if a police officer says one thing, but the accused (or a witness) says another. Is the computer program (presumably coded by civil servants) going to take the word of a police officer as more credible than the accused by default?
With a computer as the judge of fact, youre not getting rid of bias, youre just kicking the bias can down the road. Because some government employed programmer has to decide how much weight the computer gives to what evidence.
Not sure i'm 100% sold but i like the conversation. Criminal justice is a joke right now.