Private virtual Iudex in the era of 'per formulas' 2.0 process
Modernity does not necessarily transport us to the future.
On my blog I have already discussed several problems related to the use of Artificial Intelligence to make legal decisions (here, here and here). I return to the topic, because of insights that were suggested to me by the study of Roman Law.
In the chapter dealing with the evolution of Roman Law, José Rubens de Moares highlights that the main difference between the cognitio extra ordinem instituted by Emperor Augustus and the per formulas process that then existed was procedural in nature.
In the per formulas process there were two distinct procedural phases. The first was conducted exclusively by the magistrate who belonged to the republican administration. When called, the Praetor defined which formula should be applied to resolve the case. Conducting “… the second and final phase was the responsibility of a judicial body, singular or, when necessary, collegiate, which did not belong to the state judiciary. Its function was undoubtedly of exceptional relevance, being responsible for the important instructional phase, that is, the admission, production and evaluation of the evidence, and finally, the culminating act of the process, the delivery of the sentence, the decision of the cause. But nature was eminently particular.” (Sociedade e Verdade: Evolução Histórica da Prova, José Rubens de Moares, edusp, São Paulo, 2015, p. 137)
The judicial body mentioned by the author was the iudex, that is, “… a Roman citizen chosen or drawn by lot by the parties.” (Sociedade e Verdade: Evolução Histórica da Prova, José Rubens de Moares, edusp, São Paulo, 2015, p. 93). This iudex was “… generally a senator of the Republic, who, in the second phase, called in judicio, following the praetor's guidance in the edict, judged the case by sentence or declared himself unclear. It was non liquet, which led the parties to withdraw from the action or seek another judge.” ((História do Direito e do Pensamento Jurídico, Claudio De Cicco, Saraiva, São Paulo, 2017, p. 73)
When pronouncing the sentence, the iudex evaluated the evidence he had collected. However, at that stage of the development of Law, what the Romans understood as evidence was something very different from what would come to be treated as material evidence (documentary, testimonial and expert evidence) that was legitimate and capable of attesting with reasonable certainty the veracity of the thesis of one of the parties to the process.
“… for the formation of an iudex, the source of available knowledge was not included in legal treatises but in fictional works (a genre that illustrated the cultured standard of the time), and, mainly, in works of rhetoric, which allowed us to learn the discursive activity of lawyers working in the forensic field. The discourse about the instruction that the iudex should interpret was, therefore, that of rhetoric.” (Sociedade e Verdade: Evolução Histórica da Prova, José Rubens de Moares, edusp, São Paulo, 2015, p. 95)
This is so true, in the 1st century AD, in one of the few Roman legal treatises that have come down to us, the jurist Fabio Quintiliano suggests that:
“… there are two large groups of proofs: those that are independent of rhetoric and those provided by rhetoric itself, according to Aristotle's traditional division.
The first group comprises the category designated by Quintilian as ‘non-artificial proofs’. The second concerns the evidence that is extracted and arises, therefore, from the rhetorical debate itself, the so-called ‘artificial evidence’.” (Sociedade e Verdade: Evolução Histórica da Prova, José Rubens de Moares, edusp, São Paulo, 2015, p. 100)
In cognitio extra ordinem the entire process was conducted and ultimately decided by a magistrate. The investigation of the case and the decision made therein are therefore no longer private.
“... although before the parties could refuse, eventually the iudex indicated for the judgment of their case, under suspicion of partiality, it now became impossible to refuse the final judge, in this case, the princeps himself or those who acted in delegation of his authority." (Sociedade e Verdade: Evolução Histórica da Prova, José Rubens de Moares, edusp, São Paulo, 2015, p. 100)
The process per formulas and cognitio extra ordinem coexisted for a period, but the latter ended up prevailing. After falling into disuse, the first was definitively abandoned.
According to José Rubens de Moares, the evolution of the process would have corresponded to an evolution of the concept of evidence. The iudex who instructed and judged the per formulas process was a private citizen trained in the art of rhetoric. But the cognitio extra ordinem was conducted from beginning to end by a magistrate in the emperor's service.
“The progressive acquisition of technical-legal knowledge would result in the circumstance that, over time, magistrates would become less influenceable and less receptive to the old techniques of eloquent speech, typical of the orator-lawyer modality.
A slow professionalization of the career of the official judicial judiciary was progressively built, one of the consequences of the introduction of cognitio, which, in the same way, would end up giving rise to changes in the way lawyers themselves act.
It is about signaling the path that would lead, at the time of the maturation of cognition, to the systematic formation of judicial cognition in the trial through evidence of a rational nature.” (Sociedade e Verdade: Evolução Histórica da Prova, José Rubens de Moares, edusp, São Paulo, 2015, p. 143/144)
The judicial process as we know it and practice it today can, without a doubt, be considered a perfected heir to the cognitio extra ordinem introduced by Emperor Augustus. Judicial cases are instructed by magistrates who belong to the public service and must necessarily be judged by them based on the evidence produced by the parties in the case.
Rhetoric can even be used by the parties’ lawyers. But it should never be considered a means of proof equivalent to public or private documents, statements, expert reports, etc... The facts alleged by the parties must be timely demonstrated in the case records through all means considered legitimate. But illegal evidence has no legal value and its use by the judge results in the nullity of the decision given.
At this point we were when Artificial Intelligences became a reality. In Brazil, if it depends on the Federal Supreme Court, the National Council of Justice and the state Courts of Justice, AIs will soon begin to be widely used to render complex judicial decisions. Rereading the book Sociedade e Verdade: Evolução Histórica da Prova (Society and Truth: Historical Evolution of Evidence) convinced me that we are facing a true involution of procedural law.
Even though they are trained based on the Courts' public databases, Artificial Intelligence is created by private companies. Therefore, in a certain way, as a result of their use, modern cognitio extra ordinem will undergo an anachronistic mutation. Judicial cases will now be judged by a private virtual iudex and the magistrate who belongs to the public service will limit himself to feeding the robot with the specific data of the specific case, as occurred in the per formulas Roman process phase.
Rhetoric will once again occupy a central role in the judgment of cases. Artificial Intelligences do not evaluate evidence in the same way as a human being. What they do is calculate the probability of the most plausible answer taking into account the rhetorical elements provided to them.
The result provided by AI is always a text apparently endowed with cohesion and coherence, but no robot is capable of differentiating consistent adequate responses from grotesque hallucinations. This work will be the responsibility of the human judge forced to sign the sentence as if he had handed it down while reflecting on the case. This is being sold to us as an inevitable modernization of the Judiciary. However, we must suspect that the accumulation of work combined with the trivialization of the act of judging with the help of AIs will greatly reduce legal certainty.
Among the Romans, when the iudex refused to judge the formula offered by the Praetor, the pacification of society could occur through a more or less deliberate omission. Non liquet does not exist today, among us every legal case has to be decided by the natural judge of the case (even if the decision is only the termination of the process due to the illegitimacy of a party, for example). This characteristic will be maintained, as Artificial Intelligence always provides answers. They cannot simply refuse to evaluate and judge a formula presented to them by the judge.
At the time of the per formulas process, the Romans could refuse the iudex. Brazilian citizens will not be able to refuse to have their cases judged by private Artificial Intelligence. Ultimately, AIs created by American Big Techs will even be able to judge lawsuits filed against their branches in Brazil. However, it will be impossible for anyone to prove that the robot used by the judge in a specific case was designed to operate according to a bias in favor of business as usual. What can't go wrong?
Two thousand years of legal-procedural evolution will be thrown in the trash because judges want to work less. But they obviously won't give up salaries above the ceiling and trinket salaries below morality. In the first case judged by ChatGPT that became public, Artificial Intelligence provided a hallucination that the judge transformed into a judicial decision. Publicly confronted for the mistake, he blamed AI, but the company obviously will not be held responsible for what happened.
In judging whether an AI can or will be granted a patent, an English Court ruled that under the country's current Patent Law, the term 'inventor' must be a ‘natural person’.
In Brazil, the Civil Procedure Code, the Criminal Procedure Code, the Consumer Protection Code, etc... grant various powers to the judge to instruct, decide incidents, judge cases and enforce their decisions. As a result of the ambition of members of the Justice System to allow judges to use Artificial Intelligence to render interlocutory decisions and sentences on the merits of cases, the main Brazilian court will soon have to decide whether the expression "natural judge of the cause" refers exclusively or not to the "natural person" invested with public office.
Before delivering this decision, the Supreme Court Justices will have to face a paradox. Just as the "legal ChatGPT" cannot be equated with the "natural judge of the case", they cannot state that this expression is equivalent to the "natural person" who passed the exam and was sworn in as a magistrate.
A supposedly plausible solution (and which is already being discussed in texts produced by experts from the National Council of Justice) is to state that the "natural judge of the case" will have the duty to review what is suggested by Artificial Intelligence. But it is clear that no lawyer in Brazil will be in a position to prove that the human review of the AI's decision was not made in a specific case.
The Devil always hides in details and abuses the ambiguities of language. Through a rhetorical trick of the
National Council of Justice, soon the "legal ChatGPT" may or may not become the "natural judge of the case" depending on this or that magistrate who uses it in the specific case.
Before the introduction of cognitio extra ordinem by Augustus in the 1st century BC, Roman lawyers studied rhetoric. After the reintroduction of the process per formulas 2.0 judged by virtual iudex, will Brazilian lawyers have to study Information Technology and learn to use AIs to evaluate judicial decisions and prepare initial petitions and appeals? Despite their professional experience, older lawyers are likely to be discarded by the market. The youngest will be trained with the help of Artificial Intelligence. What seems like a solution can be a source of new problems.
The damage caused by social media is already evident. In the USA, Europe and Brazil, millions of people reject vaccines believing that they cause deaths instead of saving lives. The Flat Earth theory spreads with frightening speed. Misinformation empowered by algorithms guarantees fabulous profits for the same Big Techs that now make powerful Artificial Intelligences available to internet users that can generate texts, produce images and manipulate sounds and videos.
Inconsistent responses and hallucinations are common, but there are many enthusiasts for this technology in the legal world. Used to produce legal texts, ChatGPT has already caused several problems inside and outside Brazil. Many others will result from the pedagogical use of AI.
No one knows for sure whether law students who use ChatGPT, Bard, Poe, etc... will learn better or unlearn what they have learned more quickly. In classrooms, the immense authority gained by AIs (in part because of propaganda disguised as journalism) will compete with the authority of human teachers. What will be the results of this new dispute?
The advancement of the Flat Earth theory has already killed at least one person (the American who built a homemade rocket to prove that the Earth is not round). When the legal flat-earthing practice gains strength in the Justice System will the damage be irreversible?