Algorithms, it turns out, can already predict when and where a crime will happen. That way, a few steps further down the road, mass killings could be prevented and countless lives saved. Humanity could be forced to bid farewell to that age-old dark side of its own—crime.
And what about the justice system? In China, some verdicts are already officially decided by artificial intelligence, and its rulings are challenged only when there is serious evidence. Does this mean the end of judicial errors, legal tricks, the imperfections we have never managed to overcome in all of human history whenever the talk turns to justice?
Will we put John Grisham out of work? Or will we end up in the terrible final scene of a famous novel—a man and a woman are talking when a steely voice behind them says, “You are the dead”?
If every crime can be predicted or justly punished, will our lives become incredibly peaceful? Or will we find ourselves in an over-controlled global concentration camp?
…and the Judge Is a Robot!
Of course, everything said so far is an extreme. Absolute justice is a utopia, just as the little word “justice” itself is a utopia. And yet, whenever we talk about crime, the human factor has always come first. From the one who commits it, to the one who is supposed to prevent it, to the last one, who must then avenge it—this is what our civilization has rested on.
And it is precisely from the human factor, in each of these three positions, that the problems come. The difficulties, the imperfections, the mistakes that turn the pursuit of justice into a lost cause. Can technology help in this process? Not only can it, it is already happening. And the great difficulty is one of balance—what we can afford to “hand over” to it as a responsibility, and what must never fall into its “hands.”
Law is, with every justification, one of the most conservative areas of human society. People there still wear robes, still speak Latin to one another… and this is not a game for its own sake, not merely a symbol of conservatism and tradition. It is much more—a sign that this field can hardly be changed, because it stands on an enormous body of experience accumulated over millennia.
Well, hardly many Bulgarians would agree with that premise. In this country there has always been both distrust of the system and a lack of confidence that things could ever work better across the whole cycle of the fight against crime. Isn’t that exactly why we, of all people, ought to look more boldly in this direction? Yes, I know it sounds utopian, but…
Sit Crooked, Judge Straight
“Whoever controls the courts controls the state.” Was that Geshev—Ivan Geshev, Bulgaria’s prosecutor general at the time? No, Aristotle.
Technology certainly has something to offer already, even before a crime has taken place. Because it turns out that most offenses can be predicted, and that gives us even more food for thought about the nature of the phenomenon.
In July, scientists from the University of Chicago presented a project, developed jointly with the Pentagon, that can predict… 90 percent of crimes before they happen!
The algorithm was tested in Chicago, Atlanta, Los Angeles and Philadelphia. Each city was divided into zones of 92 square meters, and then the neural network was fed detailed data on the crimes committed over the past decades. It turned out that the algorithm predicts remarkably well when a crime will occur in a given zone. And since we are talking about an area the size of an apartment, it is more than easy to prevent the crime on that basis.
And then what? Once the police surround the perimeter on all sides, won’t the evildoers simply go elsewhere? Won’t criminals very quickly adapt to the weaknesses of the AI and even exploit them? The idea would then turn into just another round in the eternal game of cops and robbers…
You Have the Right to Remain Silent
A decade ago, the Chicago police developed a model that estimated how likely a person was to commit a crime or to become a victim of one. Its assessments proved remarkably accurate, and the main reason the venture was shut down was that the algorithm produced rather racist predictions—it was considerably more inclined to see people of color as potential criminals and white people as victims.
Where does the border of utopia lie? If two such algorithms can tell us when a crime will happen and who the most likely perpetrator is… will we reach the plot in which the police arrest a criminal for something they might someday commit?
This really is a science fiction scenario, but when we talk about technology it is important always to look a few steps ahead. And since we’ve touched on bias—this is where the big question comes in. Could introducing algorithms make the administration of justice more objective, or the opposite: could it bring in biases even more irreversible than those typical of humans?
Words, Not Cases
Given the American practice of juries, the latter hardly seems possible. Through complex algorithms, true, but the system is nonetheless designed to place the verdicts in the most important cases in the hands of people chosen, supposedly, as randomly as possible. People without legal training, meant to be as representative a sample of society as possible.
The idea is wonderful, but does it really deliver the justice we seek? Wouldn’t it be even fairer if, in place of the jurors, we simply had an artificial intelligence—one broad-minded enough, with prejudices more thoroughly suppressed than any human being could manage? Or, even better, a combination of artificial intelligences differing in design and training?
If punishing a crime in advance seems a frightening option, justice in the courtroom has always been a dream that perhaps can be achieved… And perhaps it already has!
“Estonia has developed, for $8,000, an AI that decides court cases when the amount in dispute is small.” You can find this news item from recent years in quite a few online sources.
Not in My Name
And yet it isn’t true—even the fact-checkers will tell you so, however many questions this line of development leaves open. The buzz started with a piece in Wired, which was then repeated onward.
Yes, but gullibility can catch up with even the dozens of media outlets that shared the story. Do you really think that an algorithm awarding damages can be built today for less than the monthly salary of a reasonably good specialist in the field? And then be released to hand down verdicts in a European Union country—and the one, at that, with the reputation of being the most digitized of them all?
Estonia’s Ministry of Justice refutes the story here. But with the clarification that artificial intelligence is indeed being used in more and more areas of the justice system. One direction, for example, is the “retirement” of court clerks, now that systems for transcribing Estonian text are already more precise than humans, with 92 percent accuracy.
The arrival of technology in law is a fact in Western countries, too. Not in the role of judge, of course, but machine-learning software has long been part of the arsenal of the law firms that can afford it. And there are more and more of them, because here, too, the algorithm can process enormous volumes of information much faster and better than humans.
In the Name of the Robot
Perhaps you, too, have been surprised by reports in our own press that the judge in some case had familiarized themselves with several hundred volumes of submitted documents. A large part of lawyers’ work today involves processing huge volumes of information. Yes, this is surely not the most entertaining part, which is why it often falls to interns, junior associates, people whose labor is cheaper. AI, however, can go further when it comes to this kind of work—not to mention its financial terms.
That is why there is now software that performs preliminary legal analysis, analyzing cases similar to the one in which a given law firm is a party. It gathers and prepares the data so that it is convenient for the lawyers. Here is one indisputable application in which artificial intelligence can be, and already is, useful—you can quickly find startups like this online.
As the futurologist Mariana Todorova wrote back in 2019 here, “there are several companies offering such a product—Kira Systems, Leverton, eBrevia, Thought River. They develop a variety of services—more precise due diligence before contracts are signed; reading and processing documents in 20 languages, with more than 50 processed in about a minute; checking around 150 contract attributes across 12,000 credit agreements in just a few minutes, the equivalent of 36,000 hours of manual legal work; finding and concealing sensitive data in potential contracts.”
All of this could have both a social and an environmental effect—think of the enormous volumes of paper involved in any more tangled case, even here in Bulgaria.
Forecasting how cases will unfold, analyzing documents and the attitudes of jurors, processing large volumes of paperwork—these are all areas in which algorithms can undoubtedly be useful and save tons of labor.
An Instrument in God’s Hands
And yet, is there really no way to include them in decision-making too? To get rid of that wretched subjectivity we traditionally complain about in these parts of the world?
After all, we have yet to see anyone in the upper echelons convicted of corruption in Bulgaria—are we left to rely solely on the notorious Kövesi, Laura Codruța Kövesi, the European Chief Prosecutor? We don’t even have spectacular trials here—show programs like the serial starring Johnny Depp and Amber Heard—so we have nothing to lose.
How much worse could it get? Even historically, if we recall that the Patriarch of Bulgarian literature, Ivan Vazov, while working as a judge… sentenced a dog. And now that algorithms even write this well…
There is probably hope, but only if we turn our gaze not west but toward the Celestial Empire. Yes, everything listed so far may suggest that technology’s place in the fight against crime today is more of a supporting one—until we get to China. That is the country developing the least known, yet apparently working, systems based on artificial intelligence.
No Pardon for a Stolen Hen
This is entirely understandable. We are talking about a country that needs an abrupt change in its administration of justice even more than we do. And it even admits as much officially. In recent decades, with the huge growth of the middle class, the need for more and more judges has grown too, because of the rising number of cases and the shrinking number of people choosing a legal career. Cases drag on for decades here in Bulgaria sometimes, too, but there it is a far more common practice, and it is getting worse.
In 2000, the caseload across all courts was 5.35 million, against 28 million in 2018—a growth of 423 percent!
That is why technology is rapidly entering every level. In 2019, 97.8 percent of Chinese courts supported filing documents online. Almost like in Bulgaria, right?
And then come the technologies in the field of artificial intelligence which, contrary to our typical stereotypes, are booming today far beyond the Western world.
Here the caveat is obligatory that, as always when the Celestial Empire comes up, there is a shortage of information, especially information that can be verified. As with topics like the notorious social credit system, so with the entry of artificial intelligence into the justice system: you can always be misled.
Uncle Robot Takes the Bench
And yet, by all indications, in some parts of the country algorithms that make decisions really are already at work. For the time being, of course, they are supervised by human judges. If the judges agree with the algorithm’s conclusions, the case is settled; if they disagree, they must write long and detailed explanations, which are then fed into the AI systems to improve them. By all indications, the final word still belongs to the human.
The system is by no means new; it has been in development since 2016, when Chief Justice Zhou Qiang called for the technology to be used to improve “fairness, efficiency and credibility.”
In Shanghai, the algorithm takes on the role of prosecutor, automatically preparing criminal charges.
Bonbons and Clyde
According to this report, coming straight from China, the applications are impressive. Artificial intelligence is now involved in every single case heard in the country! This happens through a common system of subsystems called “Smart Court,” to which every judge in the country has access, in the words of Xu Jianfeng, head of the Supreme Court’s information center.
The systems automatically monitor for references, regulations, templates of various documents. And some of them are designed to search for wrong court decisions and other inaccuracies. At first the platform worked simply as a big database for everyone involved in the process; today it plays an ever greater role in decision-making. At every step, it watches for malpractice and corruption… and, you know, in China that is no joke.
The system is estimated to already save a third of lawyers’ working time, and in the two years since its introduction it has “put in” 1.7 billion hours. “It is too early to regard smart court systems as a panacea,” Judge Sun Yubao says nonetheless. “AI really cannot do everything.” Let’s hope the smart system itself doesn’t hear him, or it might cook something up for him…
A Clear Penalty
Alas, the problem is not only that we have no way of knowing how far such systems have really been put into practice in China; even less is known about how exactly they work. For example, how they deal with prejudice—the tendency of algorithms to be “harsher” toward a certain nationality, ethnic group or even gender simply because they were trained on data from an older period. Or the opposite—in an attempt to solve that problem, going to the extreme of acquitting a given ethnic group almost by default…
Where and how does the algorithm draw its information? Is it based only on previous cases—in which case the mistakes made before will be multiplied… Does it also gather some form of outside information about the suspects or the parties to a case? If so, it will have a hard time “reading” and telling apart what applies to every “Ivan Ivanov”—Bulgaria’s John Doe—or, God forbid, every Li.
Does the algorithm try to set aside completely everything associated with emotion in a given decision? Can everything be turned into statistics?
How can we trust a system when we don’t know exactly how it works, while deep neural networks remain, for now, a black box even to scientists?
Pleading Insanity
On the other hand, we could say the same about the human brain—its workings, too, hold quite a few mysteries. Such as how we can advance technologically and still have wars, murders, thefts.
As usual, the optimal option, at least for now, is to combine the best of both. Yes, systems that can work as judges could provide an additional point of view in decision-making. They could show, for example, the need for one more level of appeal when their conclusions clash with those of the human judge.
Or we could separate them: AI could be good at searching for deviations and errors in the decisions of its human competitor, which may turn out to be important.
Whether we will one day have such working systems in the West depends on many factors, among them cost. AI has made a huge revolution in so many fields in recent years, but that always goes through serious investment. Here, by all indications, the only interested parties are states, which, to put it mildly, are not among the great fans of investing in startups.
Beyond artificial intelligence, there are quite a few other fields that can join the search for justice. Vitalik Buterin, for example, believes the solution could be systems that decide cases on the basis of a kind of voting. Blockchain technologies are planned to come into play here, and intriguing ideas are also coming from game theory.
A Robot at the Anti-Corruption Commission
It is believed that, for now, algorithms in China make decisions only in parts of the country, in civil disputes, and only over smaller sums. And that seems reasonable—decisions there don’t carry that much public weight.
I don’t know about you, but I would be curious to see the Chinese AIs in the role of Zhoro Ignatov, the judge on the Bulgarian TV courtroom show Sadeben spor (Court Dispute). If they can cope with the grotesque interpersonal squabbles there, then surely nothing can frighten them. Nor should anything frighten us about the future.
By the way, this isn’t just a joke—in Canada there are projects for AI as a mediator, to bring separated couples together and settle other “safe” disputes. Good thing Johnny and Amber’s case wasn’t there, or society might have been deprived of that lavish spectacle.
Catch 0/1
Will we ever have a world free of crime? If at some point you come to believe it, maybe it’s time to (re)read good old Dostoevsky. Or, if you don’t have the time, the no less good and old Elin Pelin, the classic Bulgarian short story writer, will do the job. Andreshko—the title character of his story about a peasant cart driver who strands a bailiff in the mud to spare a poor debtor—however much we mock him, is surely a figure a little of whom each of us carries hidden somewhere deep inside, behind the wallet.
For better or worse, humans are built in such a way that they are unlikely ever to lose their eternal need to break the rules. And yet technology can save many lives when it is applied wisely. The fight against crime is certainly a place where it has something to offer, thanks to its strengths. We shouldn’t close our eyes to that, nor should it frighten us. What could be more important than the life and fate hidden behind every crime prevented or avenged?
