The use of artificial intelligence techniques for automating decisions in the judicial system
- Watson, system developed by IBM using artificial intelligence used in the Montgomery County Juvenile Court, USA.
- PROMETEA, applied in the Ministerio Público Fiscal, Autonomous City of Buenos Aires, Argentina
- Ensuring consistency in judicial decisions; thus, fairness in applying justice
- Improving efficiency and reducing time for making judicial sentences
- Releasing time of prosecutors and judges and allowing them to make a better use of their time dedicating their efforts to complicated cases
Threats related to misuse and abuse
Explainability – In most cases, it is unknown how the applied algorithm makes decisions. They work as black boxes. In some countries where they use such systems for advising on court sentences, usually judges are unaware about details of algorithms. Judges must be knowledgeable in laws but they do not need to know neither about programming, nor about artificial intelligence. How much do judges and prosecutors need to know about the algorithm? How can the algorithm explain the decision?
Final decision made by an automated system – The use of artificial intelligence systems can significantly reduce the time needed to process information and make a decision in the judicial system. However, can the decision of a court rely solely on a decision made by an automated system? Should the final decision of a court be made by a judge? Usually, artificial intelligence algorithms make decisions based on previous cases. It may happen that in previous cases there have been incorrect sentences, for example, cases of discrimination. If that data is used, it is likely that the algorithm would decide again in a discriminatory way. The fact of having an automated decision system for the justice raises similar ethical challenges as the scenarios for automated medical diagnoses. The case of each person is different. In a scenario of 100 similar cases with the same verdict, how can justice be enforced by replicating the verdict for the case 101? Is it really the same? Cases are taken by similarity or approximation and this may fail.
Considering evolution of society – How can artificial intelligence systems deal with the evolution of societies? Machine-learning algorithms predict based on past data from previous situations and the calculation of probability. The question is how can algorithms consider evolution in current prediction models? Surely, the decision that a system can make today based on past data will not be the same as that taken by the same system within 20 years with the data produced in the new period. Thus, different artificial techniques, like those based on argumentation, should be explored and applied.
The following table summarizes ethical challenges associated with AI systems applied in court sentences.
|ID||CHALLENGE||RELATED TO *|
|C1||Ensuring that intelligent systems can explain the decision-making process and the criteria applied for making the decision||Principle O|
|C2||Ensuring that all the special issues, characteristics and background information related to a judicial case has been considered by the applied intelligent system and that the final judicial sentence is made by a person||Principle R|
|C3||Proving that the intelligent system considers the latest facts for making a decision and not just old data||Principle R|
*See Principles for more information about principles.