Digital Justice: Can Artificial Intelligence be applied to procedural structures?

In a world dominated by the digital era, justice must adapt to new technologies to ensure a more efficient protection of the rights of litigants, in order to achieve their full confidence in the system.

To this end, it is necessary to talk about the famous ‘Artificial Intelligence’ (AI), but always taking due care not to fall into the ‘hype’ that exists around it, as well as the fear of novelty or the unknown. For all of the above reasons, it is necessary to inform ourselves about it.

So, what is AI? It is a technology that, through data processing, seeks to imitate human activities. Within AI there is ‘machine learning’, which is a branch that consists of developing algorithms to identify and learn statistical patterns in data sets, which can build models and make predictions without the need for rules or pre-programmed models, based on probability and statistics.

This can be further developed by arriving at so-called ‘deep learning’, i.e. the aggregation of the use of multi-layered artificial neural networks, which seek to mimic the functioning of the human brain, thus learning to identify features from examples, ordering them into levels to form a hierarchical representation. Thus, algorithms can perform classification tasks more accurately.

In short, we can conclude that AI is nothing more than complex computational processes, which are mathematical calculations based on databases, supported by resources such as embeddings or tokenisation, which facilitate obtaining efficient and realistic results.

Now, does this mean that AI has the capacity to solve the existing problems for Justice, achieving the automation of procedural structures in their entirety?

To answer this question, we must first understand that, in order to be able to apply AI, it is necessary to generate what authors such as the Spaniard Julià-Pijoan have called a ‘favourable habitat’ for it to function well, as it works best in closed systems that follow patterns. The more defined and delimited a problem is, the easier it will be to create the AI needed to tackle it.

Taking into account that the world of law is full of open and changing situations and that AI would have to deal with several obstacles -such as the indeterminacy of normative texts, the lack of centralised and uniform databases (in Uruguay, for example, there are no uniform jurisprudential databases to operate with big data or codification in all areas of law) or the indeterminacy of jurisdictional activity- the answer to the question posed above is a resounding no, because, as it was said, AI does nothing more than mathematical calculations. It cannot think like a human, it cannot solve beyond its database, nor can it appreciate special circumstances on which it has not been trained.

Even so, this does not mean that it is impossible to implement AI in any procedural structure, because although it cannot be given the same degree of prominence in all of them, as has just been explained, there are some in which it can play an important role, performing optimally and helping to decongest the justice administration system.

So, to which structures could AI be applied without this implying a loss of certainty and confidence in the system, and under what characteristics does it perform accurately and efficiently?

We should bear in mind what was previously mentioned about the ‘enabling habitat’ for the operation of these technologies. As mentioned above, AI is mostly successful in closed and stable systems, with fixed and clear rules, where it does not face new challenges. This is why it becomes feasible to implement this tool in non-apparent voluntary procedural (or jurisdictional) structures, as will be explained below.

Firstly, it is worth mentioning that Uruguayan procedural doctrine has defined it as that exercised by the judge in all acts in which, by their nature or nature, or by the will of the parties, there is no dispute or litigation.

Now, within this category we can find two types of voluntary structures; the apparent and the non-apparent or pure. The former refers to structures in which the germ of the dispute is latent, while in the latter there is no possibility of dispute or litigation, since the claim that is processed through them cannot affect the legal-patrimonial spheres of third parties.

In Spain, Law 15/2015 calls into question the jurisdictional nature of some of these structures, while Law 13/2009 grants certain powers - previously unrecognised - to judicial officials who are not judges (such as the Judicial Secretary, Notary or Registrar), thus allowing them to be entrusted with the processing of some structures of this nature; therefore, there would be no obstacle for an IA, supervised by them, to be in charge of this.

Having made this clarification, it is possible to recognise certain characteristics in these processes that make up the ‘favourable habitat’ for the AI. For example, it can be seen that these are simple processes; we do not find two or more conflicting claims, but rather there are stakeholders pursuing the same end, making it a predictable procedure.

Furthermore, these processes do not start from the indetermination of a specific right, but from the need to comply with pre-established legal requirements in order to reach a sentence, granting certainty and/or authorising certain acts to be carried out.

It is on this basis, and in view of the social need for speedy justice, that it is possible to answer the question posed above by concluding that the most appropriate procedural structures for the prompt implementation of AI are those categorised as non-apparent or purely voluntary, and that AI can even be given a role as important as that of the delivery of the final judgment, by virtue of the simplicity of these structures, which would be unthinkable in others of a different nature.