It is important to recognize that a well-run judicial system contributes to the formation of a favorable atmosphere that fosters national growth. The efficient administration of justice is just as important to the court's efficacy as its capacity to be impartial, firm, and fair at all times. Notwithstanding these vital functions of the court, Nigeria's legal system is sometimes unsatisfactory and sluggish moving. People no longer trust the courts because of this, as most people think that justice postponed is justice denied. In recent times, machine learning methods have been used for predictive purposes in several domains. In this work, the authors used 5585 records of precedent rulings from the Supreme Court of Nigeria (SCN) between 1962 and 2022 to construct a prediction model for the categorization of judgments. Primsol Law Pavilion, an independently owned data repository, provided the data. Following data annotation and feature extraction, three classification methods (decision tree, multi-layer perceptron, and kNN) were used to construct the model.
TopIntroduction
Maintaining the rule of law at all times and defending and upholding the constitution are among the judiciary's fundamental goals as the third branch of the government (Ahmad et al., 2021). Today's highly advanced and industrialized human society has produced tremendous advancements, particularly in the fields of technology and architecture (Bhilare et al., 2019). Among the most integral achievements of humanity are law and society. The development of the legal system started with the rise of human civilization. Communities bound to a certain legislature have experienced significant disruptions and anxiety as a result of the lack of openness and ignorance of the legal procedures. As data becomes more widely available, the necessity for established standards and methods becomes even more critical (Medvedeva, 2023). Many European national and international courts strive for openness by following the regulation to encourage the reuse and accessibility of public sector information and to post their records online (Marković, 2018). An exceptional chance to handle this material automatically on a broad scale is presented by digital access to a substantial body of published case law.
Computers are now machines that can learn without supervision and adapt to new inputs without needing to be reprogrammed. This is due to recent advancements in data collection, aggregation systems, algorithms, and processing power (Ashley & Brüninghaus, 2009). In the past, there have been many efforts to anticipate court judgments using statistical techniques, such as factor and linear regression analysis. In 1957, Kort made an effort to use quantitative techniques to forecast Supreme Court rulings. Probability theory and quantitative approaches were suggested by Nagel (1960) and Ulmer (1963) for the analysis of court documents. These systems, however, were noise-sensitive and did not transfer to other legal areas (Cui et al., 2022).
Machine learning algorithms' ability to forecast court decisions has gained popularity (Aletras et al., 2016). Artificial intelligence has been used to build legal decision systems that help attorneys by automatically anticipating verdicts (Kelbert et al., 2012). Big data analytics proponents contend that it will lessen human prejudice and provide the legal system a scientific, evidence-based approach (Simmons, 2018; Završnik, 2018). In order to help judges in certain instances, machine learning has been used to provide answers known as algorithmic decision predictors (Ashley, 2019). These algorithms foretell the outcome of a case, which would normally be decided by a court or judges (e.g., guilty/not guilty, plaintiff/defendant ruling) (Medvedeva et al., 2020). It is sometimes said that algorithmic decision predictors increase the consistency and predictability of judicial decision-making, which is required under the equality principle.
These arguments contend that decision predictors may help judges make decisions that are more consistent, knowledgeable, and free of prejudice (Chalkidis et al., 2019; O'Neil, 2016). Machine learning classifiers use a traditional encoding method that does not capture the overall connections between predictor variables in a machine learning-based data set, even though selecting efficient predictors and using machine learning techniques on large data sets can yield impressive results (Alghazzawi et al., 2022). The studies conducted by Bhilare et al. (2019) and Shaikh et al. (2020) emphasized the use of manually derived characteristics from the court case dataset before using a series of machine learning algorithms to construct a prediction model that was based on performance. There have been claims that this technique can unintentionally reinforce prejudice and bias that are already present in training data (Liu et al., 2018).