Introduction
Several countries across the globe including China, Estonia, Taiwan, Canada, the UK, Peru, and Mexico are relying on Artificial Intelligence (AI) systems such as Large Language Models (LLMs), Machine Learning (ML), Deep Leaning (DL), and Natural Language Processing (NLP) etc. and have integrated them into their justice system to assist judges, mediators and other adjudicators in the administration and deliver of justice. Instead of a human judge, judging by AI has become a reality in the judicial decision-making process. Judges from Colombia, India, Pakistan, the USA and UK admittedly applied AI to adjudicate legal matters although the scale of using AI in the judicial decision-making is possibly far higher.
AI brings revolution to the legal industry through cutting-edge innovation and enhanced service that also alters the customary processes of dealing with litigations. It becomes a powerful tool to deliver justice quickly and smartly. AI has huge prospect for lawyers, law firms, pro se (self-represented) litigants and marginal communities, court staff and judges to seek legal assistance and perform their duties seamlessly.
Even this algorithmic technology is capable to determine simple cases and transactional work. As a result, AI is being used in judicial decision-making processes now which is known as the judicial artificial intelligence decision-making system. This is a system that makes analyses and predictions about legal proceedings in court through knowledge graphs, LLMs, ML, DL, and NLP etc.
Nonetheless, there is apprehension that relying on AI to determine or resolve disputes and apply it in the judicial decision-making process will have heavy consequences as uncontrolled use of AI in adjudicating cases can vitiate the fundamental principles of justice such as judicial independence, fairness of trial, equality before law, human rights etc. Overall, justice is a complicated mechanism for determining division and resolving conflicts, encompassing individuals, property, time and location, value, emotion, custom, subjective and objective states, and behavioural modes which are challenging to address through an algorithmic judge.
Additionally, there are numerous concerns over depending on AI to decide lawsuits including diminished accountability of human judges, lack of expert knowledge of judges on this emerging technology, the questionable quality of AI outputs, challenges to transcribe legal texts into computable codes, inherent bias and discrimination in the algorithm, risk of manipulation of the system etc. As a result, public trust for the court will be decreased and people’s last resort to seek remedy will be put at risk.
Thus, this paper argues that the integration of AI in the judicial decision-making process will bring severe effects upon the whole justice eco-system despite having a potential to make the judiciary people-friendly. Nonetheless, it could be used in a planned, cautious and ethical manner in the administration of justice.
The Potentials of Artificial Intelligence in Courtrooms
AI appears as an advantageous tool for lowering effort cost needed to examine the legal instruments, find out appropriate provision of law and apply it to a given fact, and enhance accuracy by generating new information about predictions. Apparently, AI makes the life of both the litigants and legal professionals including lawyers, judges, court administrators and staff, less stressful by reducing effort and providing quicker service. It brings opportunity for underprivileged litigants who usually have limited access to justice as it transforms the traditional ways of delivering justice in a cost-effective manner.
Artificial intelligence is applied to predict judicial decisions and even to recommend suitable times to appear before specific judges. AI is demonstrating better capabilities in forecasting judicial outcomes compared to lawyers, as well as in anticipating the tendency of a convicted criminal to reoffend and violations of court orders relative to seasoned judges (Beer, 2021). Today’s AI tools can predict probable outcomes of cases even before the commencement of the proceeding by analysing huge number of similar types of previous case data. For instance, these tools help to identify the risks associated with the litigation and provide the probability of success in the court. It can forecast judge tendencies, opposing counsel behaviour, and courtroom dynamics. Therefore, this technique emerges as a game changer in the sector.
Judges and adjudicators can take assistance from AI during the litigation process and as a means of dispute resolution for summarising the case or evidence in a trial; producing summaries of relevant laws applied in a case; identifying relevant case laws; drafting court orders, judgments, decrees; determining issues in a suit; creating timeline for the cases, managing cases in court; answering legal questions; transcribing and translating various documents; analysing data, predicting the final outcome of the case and for various other purposes.
Judging by AI arguably has the potential to be fairer and more neutral than human judges as personal preferences, biases, corruption, and other socio-economic issues affect human judges. It is not practicable to anticipate that human judges will maintain an impartial, objective, and consistent disposition as they have already exhibited significant inconsistency, both collectively and individually (Crootof, 2019). Moreover, human judges are, by nature, expensive as they have prepared for years, take time to adjudicate, retire and also are limited in number, while AI can work for more than 8 hours a day tirelessly, doesn’t take time off and receives zero wages.
Therefore, some regard judicial artificial intelligence as a cheap, fast and scalable alternative to a human judge (Yalcin et al., 2023). For this reason, judicial AI has been implemented widely in different jurisdictions. A UNESCO survey reveals that 44% of judiciary professionals integrate AI into their daily tasks, using it to summarise texts, draft emails, prepare documents, and conduct legal research. Although, importantly, 91% of judicial operators disclose that their institutions do not provide training or official guidelines on the responsible use of AI in their daily work (UNESCO, 2023).
The Dangers of AI in Decision-Making
Artificial intelligence is a double-edged sword technology by nature, like nuclear fission, capable of illuminating cities or incinerating them (Barrat, 2015). The issue is not that robots are beginning to emulate human thought, but rather that humans are at risk of relinquishing their distinctive cognitive abilities and becoming subservient to machines (Yunpeng, 2021).
However, there are significant risks of deploying AI in the judicial decision-making system. In its present form, it has the defects of knowledge structure, application scenario, and potential ability. This paper will discuss some more dangers of artificial judicial decision-making processes in the following segments.
The accountability of judges can be compromised and weakened through the integration of AI-powered tools in the decision-making process. This is because judges are likely incapable of delivering or clarifying the reasons that an AI system produced while making a decision, as the vendor’s software does not provide detailed information about its functionality and are not transparent (Gravett, 2024). In that case, the judges will be unaware of the way by which the outcome was generated. Judges also have lack of intensive training and comprehensive understanding on how AI systems are technically working and how to utilise them best for their works.
Conclusions are derived from unintelligible algorithms that determine correlations between various data points. An AI’s actions cannot always be explained and, even if theoretically possible, it may be impractical to do so (the so-called, “Black Box” problem). The inability to explain how a model arrived at a particular decision can obviously hinder its acceptance in the legal context. In addition, the presence of these algorithmic black boxes significantly elevates the expense associated with identifying and rectifying faults in artificial intelligence, while the deployment of efficient and automated AI poses a tangible risk of widespread miscarriages of justice (Xu, 2022).
Often information about how an AI system functions are not revealed by the people concerned for reasons of operational secrecy, to protect trade secrets or to protect the privacy of personal information in training data. However, this lack of transparency, comprehensive explanation and understanding creates potential barriers for the users to rely on AI neutrality while also undermining the principle of reasoned decision and openness of justice.
Providing just and reasonable cause in the judgment is one of the fundamental principles of justice which may be hindered through the intervention of AI and other tech tools while adjudicating a dispute. Thus, customary accountability routes like the right to appeal and the judicial obligation to give a reasoned decision are less effective if judges only have the outputs of a non-transparent AI system.
The use of AI in judicial decision-making will create further chaos when different judges start using different software. Consequently, it will bring disparity in the adjudication process. Furthermore, AI tools are not capable of applying discretion like human judges in specified circumstances which may create injustices in many cases as tech tools are not eligible to evaluate each case with appropriate considerations (Buckland, 2023). Some cases may require progressive attempts and favour from the court to bring the marginalised into the mainstream, which a machine cannot do on a case-by-case basis.
Moreover, humans have the capacity to compensate for flaws and identify issues to devise solutions, whereas artificial intelligence lacks such innovative capabilities (Xu, 2022). Furthermore, AI is unable to quantify emotions and engage with individuals’ psyches. It also has deficiency in the flexible and creative application of justice values in accordance with particular circumstances. In contrast to artificial intelligence, which is devoid of autonomy, ruthlessness, intrinsic values, introspection, and the capacity to recognise potential errors, human judges possess emotional intelligence and, as active members of society, they share the ability to empathise with the public, thereby striving for empathetic justice (Song, 2020).
Further, there is question about the quality of output given by AI because it relies on vast past databases which may lead to inaccurate, incomplete, misleading or out-of-date outcomes (European Commission, 2024). Hence, there is high possibility that algorithmic judges would replicate the previous mistakes, discrimination and bias that contained prejudices of former cases. Also, AI is often trained with public source data which are not always authentic. Moreover, it does not have the self-capacity to evaluate and adapt with the social changes of the time. Many algorithms, at least partially, form spurious (random) correlations, making the value of their output dubious.
As a result, incorporating Algorithmic Decision Making (ADM) in judicial decision-making processes may impair core judicial values like the fairness of justice, diversity, efficiency, transparency and accountability, equality before law, the right to equal protection of law, the right to privacy etc. (Heinrichs, 2022). Also, ADM has the risk of discrimination on the ground of race, sex, ethnicity, age etc. by biased algorithms, selective use of technology by judges, and the possibility of susceptibility to different cognitive biases etc.
Another dimension of integrating algorithms into judicial decision-making is accurately translating the legislation into codes, commands and functions that a computer program can understand (Kennedy, 2024). Legal texts often require contextual understandings while computer programmers rarely have legal experience (Sourdin, 2018). Codes also need to be updated regularly due to amendments, new case laws, complex transitional provisions etc. So, the above-mentioned points need to be addressed before deploying AI in the decision-making practice to uphold the fairness and fundamental principles of justice.
Jurisprudential Questions Relating to Authority, Independence, Liability etc.
A jurisprudential question may arise relating to the legal authority to deliver judicial decisions. Who possesses the legal authority to make a decision by algorithmic judge? Is it the programmer, the policymaker, the human decision-maker or the computer or automated system itself? The answer is not clear in present legal frameworks.
The independence of the judiciary can also be questioned through the application of artificial judicial decision-making. Scholars raised concerned that the court that renders judgements through an artificial intelligence-based IT system cannot be considered independent in the context of adjudicative independence (Nowotko, 2021). Furthermore, any influence exercised by a state’s executive or legislative branch, such as over data centres used for digitising judicial decisions, the selection of training data for neural systems, or the design of algorithms employed in courtrooms, could potentially undermine the court’s independence (Gentile, 2022). It is apprehended that if an algorithm generates a judgement, it is not independent from the humans who created it, nor from the data processed and maintained by an automated system. The decision, rather than being solely the result of an individual court’s evaluation, is predetermined by the human programmer and the dataset supplied by them (Górski, 2023).
There is another concern as to who will take responsibility for the mistakes of AI systems. Is it the user or the creator? Judicial office holders are personally responsible for material which is produced in their name. However, clerks, judicial assistants, or other staff will sometimes help judges to furnish their judgment, and these workers may have used AI tools without informing the judges. Hence, there will be ambiguity in determining the responsibility of any wrong done in a judgment. As a result, institutional oversight is essential to ensure that AI is used responsibly, and that cautions have been taken to mitigate the risks. It is vital to have thoughtful human oversight at every phase of AI use in the judicial process. Additionally, the use of regular auditing by a superior authority can create an extra shield against irresponsible, overdependence, and unethical use of AI.
Judges are not generally obliged to disclose the research or preparatory work which may have been done in order to produce a verdict. However, they must disclose all information publicly as to how and up to what extent AI is being used in their decision-making process, if they take help from, or deploy, AI in their judicial functions. However, a court can utilise AI in administrative functions like track allocation, case and court management etc.
Maintaining procedural fairness is an additional challenge while using AI in any part of the judicial process. Enough checks and balances have to be maintained if AI undertakes functions ordinarily performed by a judge such as summarising fact, law or evidence to maximise accuracy as a simple task like summarising evidence could be skewed in favour of one party if the AI tool was not handled properly.
Possibility to Diminish Public Confidence
The public can fear that their personal data will be misused while handled by AI. Courts are usually dealing with confidential data, privileged communications and sometimes highly classified documents related to state security as well. Thus, the question is how can courts guarantee the security of state or clients’ privileged data when shared with AI algorithms.
How does the public keep confidence that the information shared with AI is securely protected and will not be used for subsequent purposes?Also, generative AI often produces fictitious case references (often called AI hallucination), incorrect interpretations, or quote overruled decisions (Dahl, Magesh, Suzgun, and Ho, 2024).
Moreover, there is a risk that AI judges could be hacked or manipulated, which could lead to wrongful convictions.There is also the potential for a power imbalance between those who use AI systems and those who are then affected by that usage. Consequently, public trust may be diminished in the court as the public may not trust AI judges to make fair and impartial decisions for them (Fine and Marsh, 2024).
Conclusion
It is evident from the present trend that AI is becoming progressively popular and the area of using AI in the legal industry is expanding. Alongside the advantages, we must think about the cost of applying advanced technology in decision-making as well. If we consider the cost of technological error which might lead to reinstitution of a suit, blunders, sanctions, dismissals, and utter chaos, then the cost, effort and time of the litigants will be so much higher than the usual manual process that this may cause people to refrain from relying on this digitisation.
Legal system needs human intervention in judicial decision-making processes because the core difference between human and AI is that AI lacks individual judgment to a particular situation, empathy, contextual understanding, and human intuition. Judging is unique work which requires human intelligence to do complete justice while AI may play a supportive role to conduct administrative duties and the initial analysis of a lawsuit, but AI cannot replace human cognitive, judgment, discretion and empathy. Moreover, institutional guidelines can protect the interest of justice-seekers by providing a parameter of responsible use of AI in judging.
References
Barrat, J. 2013. Our Final Invention: Artificial Intelligence and the End of the Human Era. New York: St. Martin’s Press.
Beer, M. 2021. Automatic Justice: The Role of AI in Judicial Determination. Available at: https://www.law.ox.ac.uk/events/automatic-justice-role-ai-judicial-determination
Buckland, R. 2023. AI, Judges and Judgment: Setting the Scene. Harvard Kennedy School, M-RCBG Associate Working Paper Series. 220. Available at: https://www.hks.harvard.edu/sites/default/files/centers/mrcbg/working.papers/Final_AWP_220.pdf
Crootof, R. 2019. Cyborg Justice’ and the Risk of Technological-Legal Lock-In. Columbia Law Review Forum. 119(7), pp. 233-251. https://columbialawreview.org/content/cyborg-justice-and-the-risk-of-technological-legal-lock-in/
Dahl, M., Magesh, V., Suzgun, M., Ho, E. D. 2024. Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models. Journal of Legal Analysis. 16(1), pp. 64–93. doi: https://doi.org/10.1093/jla/laae003
European Commission, 2024. The role of artificial intelligence in processing and generating new data. Available at: https://data.europa.eu/sites/default/files/report/The%20Role%20of%20Artificial%20Intelligence%20in%20Processing%20and%20Generating%20New%20Data-1-2.pdf
Fine, A., Marsh, S. 2024. Judicial leadership matters (yet again): the association between judge and public trust for artificial intelligence in courts. Discover Artificial Intelligence. 4, 44. Doi: https://doi.org/10.1007/s44163-024-00142-3
Gentile, G. 2022. AI in the courtroom and judicial independence: An EU perspective. EUIdeas. Available at: https://euideas.eui.eu/2022/08/22/ai-in-the-courtroom-and-judicial-independence-an-eu-perspective/
Górski, M. 2023. Why a Human Court? eucrim. 1, pp. 83-88.
Gravett, W.H. 2024. Judicial Decision-Making in the Age of Artificial Intelligence. In: Sousa Antunes, H., Freitas, P.M., Oliveira, A.L., Martins Pereira, C., Vaz de Sequeira, E., Barreto Xavier, L. (eds) Multidisciplinary Perspectives on Artificial Intelligence and the Law. Law, Governance and Technology Series. 58, pp.281-297. Doi: https://doi.org/10.1007/978-3-031-41264-6_15
Gutiérrez, J. 2023. UNESCO Global Judges’ Initiative: survey on the use of AI systems by judicial operators. UNESCO Available at: https://unesdoc.unesco.org/ark:/48223/pf0000389786
Heinrichs, B. 2022. Discrimination in the age of artificial intelligence. AI & Soc. 37, pp. 143–154 (2022). Doi: https://doi.org/10.1007/s00146-021-01192-2
Kennedy, R. 2024. Rules as Code and the Rule of Law: Ensuring Effective Judicial Review of Administration by Software. Law, Innovation and Technology. 16, pp. 170-193. doi: 10.1080/17579961.2024.2313801
Nowotko, P. 2021. AI in judicial application of law and the right to a court. 192 Procedia Computer Science. 192, pp. 2220–2228, 2224.
Song, G. 2020. On the artificial intelligence of judicial adjudication and its limits. Comparative Law Research. 5 pp. 80–92.
Sourdin, T. 2018. Judge v Robot? Artificial Intelligence and Judicial Decision-Making.UNSW Law Journal. 41(4), pp. 1114-1133. doi: https://doi.org/10.53637/ZGUX2213
Xu, Z., 2022. Human Judges in the Era of Artificial Intelligence: Challenges and Opportunities.
Applied Artificial Intelligence, 36(1). doi: 10.1080/08839514.2021.2013652.
Yalcin, G., Themeli, E., Stamhuis, E. et al. 2023. Perceptions of Justice By Algorithms. Artificial Intelligence Law. 31, pp. 269–292. Doi: https://doi.org/10.1007/s10506-022-09312-z
RaisuL Sourav is a Doctoral Researcher in Legal Analytics at the School of Law, University of Galway, Republic of Ireland; an Associate Professor of Law (on leave) at Dhaka International University, Dhaka, Bangladesh and a an Advocate (Lawyer) at the Supreme Court of Bangladesh. He holds a Master’s Degree in International Energy Law and Policy from the University of Stirling in the UK. He was a recipient of the prestigious British Chevening Scholarship in 2017-18. RaisuL regularly writes in academic journals, newspapers, and blogs, and presents his research in reputed academic conferences. He can be reached at M.SOURAV1@universityofgalway.ie His X Handle: @RaisuLSourav