In October 2019, authorities from the Modernization area of the Spanish Ministry of Justice visited the office of the Public Prosecutor's Office of Buenos Aires. The goal was to get to know Probetea, the system of Artificial Intelligence (IA) that is used in Argentina's capital to solve different cases complexity and matter: issues of negligence, traffic accidents and gender violence, among others.

Does this mean that Spain is ready to admit applying IA to an issue as sensitive as Justice? According to the above authorities, yes. Would Spain be an IA's entrance to European Justice? Well, that's how it looks.

The statements by Sofia Duarte Domínguez, Director of Modernization of Spain, have been read in Argentina's press: "We have studied everything about Promethea, we know that it's a fabulous system and we want to see if we can get him to Spain. Even the Secretary of State - N of R: Dr. Manuel Dolzo - gave us a clear letter to advance with this, and that's certainly the future of Justice."

To be honest, the issue should not be taken by surprise. That same month, a few days before the Spanish Commission's visit to the Spanish agency, Professor of Law and Political Science at the Universidad Oberta de Barcelona (UOC) David Martínez explained, in an article published by La Vanguardia, that IA could well be used in Spain in "easy-to-answer" cases leading to the dismantling of traffic in judicial proceedings.

In that same article, Professor Martínez advanced his analysis and noted how useful it was to put into place a computer system to automate responses but as an example what Estonia had developed in terms of robotization of justice. Also, at that time Martínez made clear that the American model, using COMPAS, a predictive software to compute the recidivism of an accused person for criminal offences, was not an example to be followed.

To start with, applying information and information technologies to judicial processes that are capable of: learning, generating predictions about human conduct, attending to judges while pondering probabilities and displacement of many employees by absorbing bureaucratic and repetitive tasks, that's a very sensitive matter.

On that journey to Buenos Aires, Duarte Domínguez stated that the devapelization of the Spanish justice ministry as a whole was an encouragement with respect to the technological progress that means applying automatic proceedings in judicial proceedings: "The digital file has been law since 2009. To that end, we have come a long way that today allows us to think about the innovation that means applying an IA as Promethea."

Thus, she intended to show that the technological advance undertaken was fertile ground to move from a digital state to an intelligent one, including by coping with the pateling of employees and trade unions from as bureaucratic agencies as Justice.

Meanwhile, Martinez's comments, as well as those of other European colleagues, are consistent with what some Argentinian experts think, committed to the task of performing an intelligent justice.

That was the case, for instance, with the Supreme Court of Justice of Argentina's Mendoza Province, Mario Adaro, who recently participated at the first Ibero-American Summit of Artificial Intelligence at MIT headquarters in Boston.

Adaro explains: "The IA has an information processing capacity in large volumes that cuts the bureaucratic deadlines to an extent that cannot be overlooked, because usually more and few decision makers have more time per case. Using automatic processes, the judge has more analysis capacity. The judge became an analytics reader. But what's fundamental about the matter at issue."

"At Mendoza, for instance, we did a tax pilot test, and in that area, they are serious and high-volume judgements, where decisions are grouped into clear sets and everything's quite mechanical and predictable. That's where Promises cheer and allows the judge to avoid repetitive work with no added value. The human effort in that area was more about monitoring specific data than about arguments, as there was little room for thinking or innovation in decision-making criteria."

To close, the lawyer emphasizes: "By using IA for that type of problems, Promethea makes the amount of errors in load, tipeo, redundancy, and so on, come down sensitibly."

To date, efficiency appears to be inobjectionable, especially with regard to state bureaucracy.

Now good. What exactly does IA's application of Justice mean?

At this time, there are 4 flagship and active cases worldwide applying IA to judicial proceedings. Maybe the most famous and controversial is the COMPAS program (an acronym of Correctional Offensive Management Profiling for Alternative Sanctions), used in at least 10 districts of the United States. Let's see how that works.

COMPAS, whose translation could be Medical Profiles for Alternative Penalty from the U.S. Prison System. The UU is a software that has been used since 1998, mainly to analyse, depending on an accused's criminal background, his chances of re-inciting. The fact that you rely on an algorithm to convict a human in criminal proceedings is the first point of the dispute, given the gravity of what's been judged.

The computer system is applied by applying a questionnaire to each accused. Once the complainant answers all questions, the system calculates the risk of recidivism and so the judge defines, for example, whether or not to give parole while the court proceedings are complete.

According to a report published in El País in 2018, the system analyzes 137 aspects of each accused. But by contrasting the level of success between the predictions of COMPAS and those of lawyers of flesh and blood, we note that the level of success of IA is no higher. Even serious mistakes are clear. Lorena Jaume - Palasi, a European expert in ethics and technology, analyzes:

"Statistical averages say something about common behavior patterns in a collective. They do not describe individual profiles and are unable to capture human individuality. Thus we can understand collective with a somewhat more architectural look but we also have a risk of bringing individuals into standards that they do not match. For example, a homosexual or single mother who does not get social assistance because a software program defining who's capable of applying for help only defines as mothers women married to men)."

One of the main discussions was to consider whether it was fair for a digital system to be a source of justice as such. To that end, we have to understand with what criteria the algorithm operates. Jaume - Palso seems to have been a key issue: actually, law is an algorithm that's applied well before computer information existed:

"They have all put their eye to the computer system and have been outraged by racism, but COMPAS let us know about the biases that judges have, because finally the system created him human, that they have been working and deciding with those biases that later attested to his program. The statistics that COMPAS uses, however, strengthen Florida's structural racism, among other American states."

"Even more. There was a single researcher who put his eye to the judges' decision, and not to the program, and saw how they did what the software recommended and how they did not. And it's very interesting that, in some cases, COMPAS advised an African American to be allowed parole and they denied it. And in others, upside down. Thus, it was found that the court decision isn't so conditioned by the algorithm. That allows us to think that perhaps in some decisions humans hide behind technologies to cover up their prejudices."

Finally, Jaume - Palasi defines: "It's nothing simple to identify when the judge gets influenced by a computer program and when he's very capable of discerning and not letting his conduct. That's what we need statistics for, and that's achieved with data, not without them. And we should not focus on math itself, because these technologies do not appear out of nothing but the prediction of a digital system starts with a training with data that humans burden them."

In this sense, it seems appropriate to resort to those who deeply understand the mathematics of this time, applied to an algorithmic process such as those operating after the Big Data and IA. Pablo Mlynkiewicz holds a Bachelor's degree in Statistics and for 4 years directed the office of Modernization of the Buenos Aires City Government.

"What the algorithms certainly allow is to standardize decisions. That is to say, to standardize criteria so that two different responses to the same problem are not given. But, of course, for that to translate into a real advance in justice, the database should have representation from all communities. Otherwise there will be mistakes."

Mlynkiewicz thus coincides with Jaume - Palasi and with Adaro at the same time, because Argentina's judge also points out, as Corvalan, a strong point of the automation of judicial proceedings: avoid two different responses to a same problem, that is to say, to bring argumental consistency to court decisions.

Indeed, while the most critical, the philosopher born in Majorca makes it clear: "We have known for a while that the judges and the judicial system we know are not very consistent. I'm no longer talking about bias, as stated by the COMPAS scandal (...) The lack of consistency is documented in studies from decades ago, showing that judges are people, nothing more or less. And they decide differently, unintentionally, according to their mood, or many subjective variables. Thus, thanks to the IA, being able to track and statistic court decisions isn't bad at all."

Robot judges in China

If there is a political organisation in the world that deserves to be taken with care of by European democracies, that's the Chinese state. It turns out that last October, an online dispute centre was set up in Beijing as an online dispute centre.

The first to note is that, according to official information, this is a platform where the parties to the dispute carry the information about the problem to be solved, and the IA does the rest: looks for jurisprudence, analyzes the issue, contrasts evidence and pronounces judgment. The key was that there was no human intervention throughout the process.

To get deeper into this model, Dante Avaro, Doctor of Political Philosophy, dedicated himself to studying China's model of governance, focusing his social monitoring system. When consulted about Eastern robot judges, she explains:

"The development of virtual and cyber judges in China has been following the same line as Social Credit System: from bottom up. Both began at the beginning of the new millennium. In the case of IA in justice, it was experienced in cities such as Shandong, later in Hengezhou, Peking and Guangzhou. The aim was to effectively conduct judicial proceedings on issues of electronic commerce, virtual payments, cloud transactions and intellectual property disputes."

"Those are, exactly, the central concerns of the Social Credit System: to provide more security to (re) put in place a system of confidence in a credit and trade system that grows up and is filled with corruption and violations of intellectual property rights. The Social Credit System can be synthesized as a monitoring and population tracing device to give loads and distribute benefits, under the best of the rule at one place and get punished from all sides."

"That system of scores, that promise to become highly sophisticated, does not only constitute, with Yitu Dragonfly Eye - N of R: the face recognition system - a brutal monitoring system but a powerful test archiver that means digital inputs, for the courts. Thus, China has been able to advance in two dimensions simultaneously: on the one hand, the Social Credit System involves a disproportionate system of population traceability and, on the other, that system is necessary to accelerate the incorporation of IA into its system of providing justice."

Before, as you wanted to know how the climate was before you came out of the house, you would sky out of the window, or go outside, and look at the clouds, and hear the horizon and feel the fresh wind. That's how I get information about how to get dressed or what means of transport to use. You understand something about moisture, temperature, sunlight.

Today, we have a phone phone number. That made us no longer understand the atmospheric situation and we don't understand what's going on in the environment. We simply followed the information provided. That simplified us and made our lives easier but eliminated human experience.

The same, with no analytical capacity, happens to decision makers only using the algorithm. Finding patterns that allow them to classify to learn from the past and predict, the algorithm chooses variables. That choice was human and currently debatable. But, if, above that, whoever decides doesn't put back his or her background and his or her experience to decide, then his or her decision will certainly be wrong.

That's how we can't figure out how to operate. It's proprietary software packed, so it's either take it or leave it. But, I have to say, the problem isn't the algorithm, as far as his margin of error is reasonable. The problem is that the judge understands how the system operates, how it gets to predict, and if it analyzes the context, especially something so sensitive.

What the algorithms certainly allow is to standardize decisions. That is to say, to standardize criteria so that two different responses to the same problem are not given. But, of course, for that to translate into a real advance in justice, the database should have representation from all communities. Otherwise there will be mistakes.

An algorithm is a set of logical indications with a mathematical basis. It usually has the capacity to learn, that's to say, that the more it works, the better the percentage of right results it gets. It can be inferred from this that Compas as, as well as any other IA program, can predict because it was trained by programmers that enter the system a number of cases that serve as a final sample.

What Compas does is what scientists of empirical sciences call statistical inference: given a given number of cases, they are interpreted as a representative sample of a certain universe, they observe the relationship between variables and they infer logical relations. That inference then contrasts with reality, to see what percentage of cases the theory is verified and details are adjusted. That's what physics, biology and medicine do, among other sciences. Perfect laws such as those of mathematics are never obtained, but the advances have shown that they have access to true or at least reliable knowledge.