What issues may arise fromjudges’ use of artificialintelligence in the decision-makingprocess of local courts?

Since the moment artificial intelligence (hereinafter, AI) appeared, it has been integrated into every area of ​​our lives. During the processes, technological innovations naturally related to AI raise many questions. Since it became a part of our lives, we have been faced with a series of concerns. One of the most interesting questions is: which professions will AI eliminate? The good news for lawyers is that, according to research by DRI – The Leading Defense Litigation Bar, the legal profession is not expected to be completely replaced by artificial intelligence. However, we cannot make the same statement for local practice, because practical application in this area seems to be still limited. However, the use of AI by judges, especially in the decision-making process, may mean facing various dilemmas. For instance: What problems can this create within the framework of the principle of fair trial when judges use AI when making decisions?

            The reason we need to think about such a question is the machine learning issue that arises during the AI ​​training process. Machine learning allows AI to learn information from databases as a system. To be more specific, during machine learning, AI can answer questions addressed to it based on the information entered into the AI. For example, if AI is trained to learn what a “dog” is, it is fed written or image-based information so that the AI ​​can understand how a “dog” is formed [5]. This information is presented to the AI ​​model during training, which allows it to recognize dogs even in images it has never seen before. This process requires special sensitivity, because if only the image and characteristics of running dogs are used during AI training, the AI ​​may not recognize sitting or lying dogs in the future [5]. Therefore, if the information used for AI training is biased, subjective, discriminatory, or contrary to legal principles, this can become a potential source of error for the judge who relies on it.

            Article 127, Part 2 of the Constitution of the Republic of Azerbaijan (hereinafter, the Constitution) states that “judges shall consider cases impartially and fairly, on the basis of equality of the parties, on the basis of facts, and in accordance with the law” [2]. In Part 4 of the same article, it is stated that “justice is administered on the basis of equality of citizens before the law and the court” [2]. An article that reflects almost the same requirement is also found in Articles 7 and 8 of the Law of the Republic of Azerbaijan “About Courts and Judges”[4]. In a broad sense, justice means the restoration of justice that has been violated in society, and in our country, the authority to implement it belongs only to the courts. As in other areas of society, no individual or social group can use superior rights or restrict their rights in the implementation of justice. When considering a specific case in court, a citizen’s race, nationality, religion, property status, service position, political party affiliation, or any other circumstance cannot influence the judicial process or the final decision. The principle of the implementation of justice on the basis of equality of citizens before the law and the court implies that all citizens are subject to a single judicial system governed by the same procedural rules, legislation and legal guarantees [3]. Articles 4 and 8 of the “Code of Ethical Conduct for Judges”, approved by the decision of the Judicial Legal Counsel dated June 22, 2007, state that a judge must be fair, impartial and not discriminate on any grounds while performing his or her duties [1].

            As can be seen, the Constitution specifically addresses the concepts of impartiality, the rule of law, justice, and equality. Let’s think for a moment that a local judge uses AI to make a decision. From a slightly pessimistic perspective, it can be said that many of our judges do not have much knowledge of how AI works technically. If the AI used by that judge were trained without ignoring the most fundamental legal principles, the result would naturally contradict the essence of the legal system. In other words, let’s imagine a situation where only white people were shown when training the AI ​​on the concept of “human”. In this case, the question or conclusion reached on the issue regarding a black person fed to the AI ​​would not reflect reality or would be a clear discriminatory error. Such a mistake would be a very extreme situation for those involved in machine learning, but in more possible situations, those involved in this activity could intentionally or unintentionally input incorrect data during the training process. Particularly, in cases where the legislation has fine details and exceptions, wrongly entered information can result in not giving a fair trial decision.

            From another aspect, we may be faced with a new sensitive dilemma. Thus, a fair trial per se involves the application of certain legal guarantees. One of such guarantees is the confidentiality of judges’ deliberations. This means that judges must make their decisions completely independently and without any outside interference. In fact, it is strictly forbidden for outsiders to be in the deliberation room during the discussion and issuance of the final court decision [3]. In such a situation, how right is it to use AI during the decision-making process? It has already been mentioned above that when training an AI system, databases are entered by humans as part of the machine learning process. That is, people who enter data may have entered their biased, subjective opinions into the AI ​​as information. Even if they do not do this intentionally, it is probably clear to everyone that the environment, social, cultural, and economic conditions in which people grow up influence their opinions even if they do not want to. For example, “equality” may mean a different concept to a 22-year-old Azerbaijani woman than to a 70-year-old Indian man. If the presence of outsiders in the room during deliberations is prohibited, does the use of AI, which is trained by entering data by humans, in the decision-making process call into question justice?

            From my point of view, the problems do not end there. In addition to the above, one more issue can be highlighted. Hence, the AI system is structured in such a way that the questions addressed to it, along with the answers it provides, are stored in the system as a history. It is clear that there is a sufficient number of cases in court proceedings that contain confidential, personal information. It means, judges who use AI assistance in such cases, in some way, unintentionally transfer that information to the AI system. In other words, the principle of “non-disclosure of judgments” covered by justice may also be at risk.

            The purpose of this short article was to show what questions will arise in the case of judges using AI during the decision-making process in court proceedings in accordance with the requirements of justice contained in the Constitution. The goal is not to present AI in a negative light or to emphasize that its use is wrong. Because even if we wanted to, it is almost impossible to escape today’s realities or turn away from a tool that is so much part of our lives. We must try to keep up with progress, and we must do this not only with ourselves, but also with our legal system and laws so that our development is sustainable. For example, some ideas can be put forward to solve the problems listed in the previous paragraphs. If judges systematically use AI in the decision-making process or will do so, first of all, the AI ​​itself should be managed and controlled from a certain center. The use of publicly available tools such as ChatGPT and Gemini by judges can be the cause of a number of violations, primarily confidentiality. In my opinion, it is possible to create a certified and thoroughly tested (from legal principles, impartiality, compliance with legislation, engineering, technological and security perspectives) AI system for use by judges under the supervision of the Judicial Law Council. After judges’ start using it, that AI system can be periodically checked after a certain period of time and its usability can be confirmed. If such a system is developed and put into use, in my opinion, questions about the confidentiality of consultations and the confidentiality of data will almost disappear. Because the system will be dependent on the Judicial Law Council, and the confidentiality of data will thus be directly protected by the state. That is, the system will be used only on court computers, without going beyond their server. Since the Judicial Legal Council is interested in the independence, impartiality, and fairness of judges, judges will be able to make decisions independently using AI without being exposed to any outside influence.