Toomas Vaher
Attorney-at-Law, Partner at Ellex Raidla Law Firm, Member of the Board of the Estonian Bar Association, Member of the Council for Administration of Courts

 

The topic of artificial intelligence is hot and maybe even overexploited: it is talked about at conferences, in training sessions and, with a cup of coffee in hand, in small talk. Until last year, it was an exciting toy whose practical use remained a little obscure for most people, but its potential seemed immense.

With the advent of new versions of artificial intelligence based on large language models (ChatGPT, Bard, Microsoft CoPilot, etc.), many people who are not very familiar with IT also came to realise that artificial intelligence is in fact here. It began to diligently write school essays and papers, and teachers had to start acquiring skills on the go in order to distinguish between texts generated by students and robots and implement new possibilities in teaching.

Of course, for those familiar with the field, this was not a new phenomenon but a faster development of artificial intelligence than before, which suddenly crossed the news threshold and reached the masses. It continues at the same rapid pace in 2024.

Artificial intelligence is becoming a part of everyday office life and all kinds of written work. It is too late to discuss whether AI can participate in the world of the administration of justice. The process cannot be stopped; the question is whether and how we can steer it. If we did nothing, ignoring the existence of AI in the courthouse, or even banning its use, it would not be of much use. In any case, new technology secretly secures its place at the table through our daily activities, starting from drafting procedural documents, translations, legal background research, and the analysis of case law, and ending with riskier situations, such as giving incorrect or forbidden advice. At the same time, societal pressure on courts is increasing, with the expectation that justice is administered faster and more cases are resolved with the help of an increasingly smaller number of people. Without the help of artificial intelligence, resolving the situation is questionable in the long run, even if it were possible to temporarily create more positions for judges. The lack of people is not a problem that can only be attributed to the courts, the declining population does not leave any part of the labour market untouched. So, do we see a future judicial system without artificial intelligence? Citing the words of a well-known film character, the judge should state here: “I am not a free man. I have no other choice.”

Mind games with the future

It will soon not be possible to be engaged in almost any area of life without using modern technology, which also raises a question of the other extreme: could artificial intelligence be a judge and replace humans in decision-making in the future? This is currently a utopian idea, but as a mind game, it would help us to better understand the elements that judicial proceedings can be divided into and the potential role of artificial intelligence as a tool.

Without delving deep into the fantasy world, and assuming (perhaps mistakenly) that there will be no artificial common sense, it must be said that artificial intelligence will probably not be able to rise to a level where it could make entirely high-quality substantive judgments. Analysis carried out by artificial intelligence cannot cover those aspects that are very definitively related to being human. This does not mean that so-called surrogate decisions could never have a place in fast decision-making, but in this case, we would have to accept that these decisions would be based on some more primitive standard.

The process of the administration of justice contains many inputs that are objective and subjective. Artificial intelligence can more or less cope with objective inputs, but it cannot imitate the subjective qualities inherent in the nature of humans as biological beings, such as creativity, empathy and social cognition. As AI is only an algorithm in spite of its complexity, there is currently no prospect of developing its human characteristics. It also has difficulty solving moral dilemmas. Familiar examples come from the field of self-driving cars: whether to run over two elderly people or a child, etc.

In addition to professionalism and diligence, judges are expected to be independent, impartial, honest, moral and equal in the treatment of others. In order to resolve cases fairly and justly, it is necessary to be aware of the general principles of law and, for example, to refrain from applying a rule where necessary, declaring it unconstitutional, or to reject a clause of a contract on your own initiative due to its nullity. The judge may play an important role in directing the parties to an agreement.

Within this whole set of human qualities and abilities of judges, the strengths of AI lie, perhaps, in expertise and diligence, and, with certain concessions, independence. Artificial intelligence could know all the laws and court decisions; it has expertise in every relevant non-legislative field, forgets nothing, misses no deadlines and acts extremely quickly. If we leave aside the influence of the writer of the code and the choice of material used to teach AI, it could also be quite independent in its analysis (at least of the parties to the dispute). However, its problem is bias in resolving individual cases, as it prefers options with a higher statistical probability. Artificial intelligence is hopelessly at a loss when it comes to understanding the general principles of law, as it is difficult to put situations into the language of the code if, due to a vague principle, a clearly written norm is suppressed in the decision-making process. Empathy towards the parties and other human qualities cannot be expected from the machine; it does not have them and it does not know how to imitate them (or does so poorly).

By contrast, the administration of justice encompasses a wide range of administrative activities, formal communication with direct or indirect participants in the proceedings and auxiliary functions (translation, recording, taking of minutes, document management, process schedule and administration of court resources, automatic data updating, systematisation, all kinds of statistics, categorisation and data analysis, etc.). Artificial intelligence would match the description of such a position perfectly, and if it could take the shape of a person and appear at a job interview, it would be very gladly hired.

Automatic surrogate decisions

Judicial proceedings under the leadership of a judge are relatively expensive and time-consuming, although they undoubtedly give the best possible result in terms of the quality of the decision (or even the only result that could be defined as a court decision, considering all of its characteristics). However, from an economic point of view, this is not always expedient and in some cases, where speed decides everything, it cannot be used at all. Society probably also needs alternative, simple, cheap and quick solutions to certain disputes (e.g. time-critical civil or commercial disputes). Any competent nutritionist will say that you need to eat regularly, your meals should be balanced, made from high-quality raw materials, and you should not rush when eating. But when rushing to the airport, if you have a choice between taking your flight with an empty stomach or grabbing a hamburger, fast food is a better solution than staying hungry.

If we were to divide the process of making court decisions notionally into two levels, the first could be a legal positivist analysis that ruthlessly follows only written norms, previous case law and facts, and the second level following it would include considerations based on human characteristics as well as the general principles of law and so-called softer values. Artificial intelligence could soon be able to handle the first level – its work would result in a draft decision, to which the judge would add all the missing elements and make the final decision in the second level.

If society agrees that in certain matters it is allowed to loosen the requirements for the decision, i.e. lower the quality bar, there would probably be a fairly high demand for hamburger-type surrogate decisions comparable to fast food in the future. In this case, the parties knowingly forego the judge’s thoroughly considered ‘real decision’ and choose this less valuable substitute (surrogate) – a decision made by the machine. Artificial intelligence would make such a decision cheaply and quickly, the related state fee and other costs of resolving the case could be relatively small. It goes without saying that there is a downside that is not insignificant – the decision may not be right or fair.

If we followed the same path, it would be technically possible to achieve real-time dispute resolution. If, in the course of their legal relationship, the parties made data available to AI in a confined space on an ongoing basis, it would provide a solution based on algorithms. In this case, the defendant should be prepared to allow the infringement of their rights to a certain extent, for instance, give up confidentiality or privacy, in exchange for technological benefits that offer speed, simplicity and affordability.

Would such surrogate decisions, which do not meet all the requirements and prerequisites of a court decision, be part of the ‘menu’ offered by the court or would the court as an institution exclude this possibility and designate such decisions for use in the extra-judicial resolution of disputes? This is a legal policy choice.

What could be done now?

The introduction of AI solutions does not mean that a revolution should be organised and the entire judicial system made to work in a completely different way. Progress can also be made in small steps.

The public sector will probably not be able to develop an AI model that specifically covers the work of courts, or if it attempted to do so, it would probably be outdated even before it is completed because developments in the private sector are simply much faster. It would be wise to use components that already exist, and if something new emerges, the previous version can be replaced. It makes no sense to invent the bicycle yourself, but the transition to a new, trendier model could be logical.

However, a technologically appropriate environment is essential. The electronic working environment of the courts can be developed and adapted so that applications developed elsewhere can be used within it or together with it (interfacing). In view of the advent of artificial intelligence, the form in which data are submitted to the court and the manner in which communication with parties outside the court system takes place needs to be organised now in order to ensure the machine readability and reasonable automatic processability of data, including the conversion of incoming files that are in a non-standard format into readable files.

Data processing problems can be resolved and are also addressed in the European Union’s Artificial Intelligence Act. The use of AI in the legal system involves a high risk; therefore, precautions and mandatory analyses corresponding to that level must be implemented, including ensuring compliance with personal data protection requirements. Some models use publicly available and user added data, which means that if any information is ‘fed’ to AI, it gets out of control, may become public and can no longer be taken back. For example, when reading the terms of use of three major language models, it turns out that in the case of two of them (ChatGPT, Bard), the information entered is used to develop machine learning technologies. Consequently, such a tool is not suitable for the field of the administration of justice. However, there are models that work with existing data in a confined space and generally do not share them outside (e.g. CoPilot).

The use of AI in court proceedings – as in any other process – requires analysis, then optimisation and the automation of certain workflows (where possible and appropriate). Quite practical AI tools are already available for many aspects of the administration of justice, such as:

  1. translation (real-time interpreting, translation of documents);
  2. document management (including the systematisation of evidence submitted by the parties);
  3. data collection, extracting certain information from the submitted dataset (e.g. in the course of the analysis of evidence);
  4. processing large volumes of data;
  5. summarising longer texts and oral presentations;
  6. administrative organisation (scheduling, coordination, premises, resources);
  7. recording (including the preparation of a verbatim report of meetings and the minutes on the basis thereof);
  8. data visualisation;
  9. systematic presentation of topics and data;
  10. verification of the consistency of the content of references;
  11. automatic updating of the data of the parties to the proceedings;
  12. allocation or reallocation of cases, measurement of workload.

Logically, it could be feasible in the coming years for some artificial intelligence models to trawl through the databases of legislation and court decisions in force in Estonia, pick up the legal provisions and practices that are relevant in terms of dispute resolution and present it in a systematic form along with relevant arguments.

Finally, what is most important: implementing AI definitely requires a positive attitude and in-house experts whose help is constantly available. We cannot expect people who are professionals in law to be professionals in information technology and the field of artificial intelligence. The availability of legal technology specialists in the judicial system is indispensable; the division of roles between algorithms and people does not happen by itself. If there are not enough of such experts (this is probably the case), it is necessary to immediately start training them and promoting the respective fields of study at universities as well.

As the field of artificial intelligence evolves rapidly, the moment when you can say, ‘It is done, it is ready’ will never come. This is an ongoing process. New technology offers opportunities to make people available for tasks in which they are stronger than others. Artificial intelligence will not replace human creativity, but it will soon be indispensable as a tool.