Introduction
AI may be defined as the “allowing a machine to behave in such a way that it would be called intelligent if a human being behaved in such a way”. John McCarthy, widely credited with coining the term "Artificial Intelligence," provided the following definition for AI in 1956.2 Machine learning allows software programs to enhance their predictive capabilities without the need for explicit programming. Machine learning algorithms utilize past data as input to forecast novel output values. This technique significantly improved in effectiveness with the emergence of big datasets for training purposes. Deep learning, a kind of machine learning, is founded on our comprehension of the brain's organization.3 The utilization of artificial neural network architecture in deep learning serves as the foundation for current advancements in AI, such as autonomous cars, ChatGPT and AI enabled robots.4 Given India's substantial population and the accumulation of pending cases, the implementation of artificial intelligence (AI) can facilitate the judicial procedure and deliver expedited and more effective justice.5 Artificial Intelligence (AI) has the capacity to revolutionize the justice delivery system in India by enhancing the skills of legal researchers and judges, decreasing the backlog of cases, enhancing access to justice, and guaranteeing the quality and uniformity of judgments.AI reacts to feedback in a way that is similar to how humans normally react, since humans can think, judge, and have intentions. The technology has changed from following pre-designed and pre-configured codes to a more complex final product that can think like a person. This is what makes it really possible to change legal systems around the world.6
Usage of AI in Legal System of Different Countries
i. Estonia
A projected "robot judge" could resolve minor claims disputes under €7,000 (about $8,000) by utilizing an artificial neural network.7 Both sides will submit documents and other pertinent material, and the AI will render a ruling that can be challenged by a human court. Transcribing a single criminal case, which involves approximately 14 hours of audio recording, often requires an average of 112 working hours. Salme, a speech recognition tool developed in collaboration with IT company CGI Estonia and language-tech startup Tilde, has been implemented in Estonian courts to streamline and enhance this process. With 92% precision, Salme was implemented to optimize transcription efficiency, especially during lengthy court proceedings that extend above six hours.8
ii. United Kingdom
In 2020, the UK Ministry of Justice implemented the Digital Case System specifically for the crown courts. The platform provides live updates on ongoing cases and enables remote court engagement. Additionally, it facilitates the electronic submission of evidence, hence minimizing the need for paper consumption. The online portal's access criteria for criminal law barristers are provided by the Ethics Committee of the Bar Council.9 The Harm Assessment Risk Tool (HART) utilizes algorithmic methods within a policing framework to achieve improved outcomes in terms of public safety, legal considerations, and cost/resource management.10 The HART model utilizes behavioral predictors, along with demographic factors like as age, gender, and residential postcode, to estimate algorithmic risks. These factors might be considered indirectly associated with indicators of community impoverishment.11
iii. USA
Judges Substituting Speculation with Algorithm for Bail. The algorithm assigns defendants two scores: one indicating their tendency to engage in criminal behavior and another indicating their likelihood of not showing up in court. It also identifies individuals with a heightened risk of violence.12 After analyzing data from 1.5 million criminal cases, researchers discovered that less than 10 objective characteristics, such as age, criminal record, and past instances of not appearing in court, with more emphasis on recent offenses, were the most accurate indicators of a defendant's behavior. The user's text is a bullet point. I-CAN!, the Interactive Community Assistance Network in Orange County, United States, offers interactive modules that address the legal challenges commonly faced by self-represented litigants. Self-represented litigants are guided through the modules to locate the suitable forms for submission in court. This not only enables self-represented litigants to get a clearer comprehension of legal procedures, but it also results in time and cost savings for under-financed legal aid centers. The number of pleadings submitted per month is 4000.13
iv. Brazil
In late 2017, the Supreme Federal Court (STF) and the University of Brasília (UnB) collaborated on the development of the Victor Project. This project utilizes artificial intelligence and machine learning to address and alleviate the obstacles related to enhancing efficiency in the judiciary and effectively managing judicial processes. Originally employed for the examination of exceptional appeals received nationwide, particularly in relation to their categorization within overarching themes of larger prevalence.14 Victor has developed the ability to classify the papers and divide them into independent PDFs. In a PDF encompassing the full legal procedure, he has the ability to categorize it based on document kinds, such as the initial petition, defense, sentence, appeals, and others. Once the software has been appropriately divided into independent files, it proceeds to categorize and recognize these documents based on their nature, following the requirements set by the Supreme Federal Court.15
The AI system, SOCRATES, underwent training on a dataset consisting of 300,000 court rulings. AI analyzes new cases and categorizes them based on their similarities, allowing them to be evaluated collectively. The software is additionally employed in screening to prohibit the admission of certain categories of cases that are irrelevant to the court's responsibilities.16 The three primary constituents of SOCRATES are Collect, Analyze, and Explore.17 The Collect module enables users to acquire data from social media platforms, while the Analysis module facilitates users in doing fundamental analysis, including statistical analysis and textual analysis.18 The Explore module offers users various methods to visually represent the gathered data and the outcomes of research. SOCRATES 2.0 has recently upgraded specific back-end modules, such as the YouTube API version 3.0 and the Facebook API 2.3, in order to stay updated with
the latest social media services' application programming interfaces (APIs). Additionally, SOCRATES 2.0 is actively incorporating additional data sources.19
How AI Is Being Used in Indian Judiciary?
The Indian judiciary has embraced AI at an early stage. After establishing the necessary infrastructure for e-courts, including basic computer hardware, under the courts Mission Mode Project, the Indian judiciary has made significant progress in the past two years to fully use the potential of advanced AI technology.20 Justice Sharad Bobde, the esteemed Chief Justice of India, has consistently emphasized the importance of using AI-driven technology to enhance institutional efficiency. On November 26, 2019, Justice Bobde introduced the beta version of SUVAAS, a neural translation tool, on the occasion of the national Constitution Day. This event officially signified the introduction of artificial intelligence (AI) in Indian courts. judicial L. Nageswara Rao, who leads the Supreme Court's AI Committee, indicated that AI will be utilized for administrative functions and accelerating the judicial process.21
i. SUPACE:
SUPACE is an abbreviation for Supreme Court Portal for Assistance in Court's Efficiency. The E-Courts Project was initially conceived based on the "National Policy and Action Plan for Implementation of Information and Communication Technology (ICT) in the Indian Judiciary - 2005" proposed by the e-Committee of the Supreme Court.22 It is a tool that compiles pertinent legislation and facts and presents them to a magistrate. Its sole function is to analyze facts and provide them to justices seeking input for a decision; it is not intended to make decisions.23 The justices presiding over criminal cases in the High Courts of Bombay and Delhi will initially implement it in an experimental capacity.24
ii. SUVAS
The Supreme Court has created the Supreme Court Vidhik Anuvaad Software (SUVAS), an artificial intelligence-driven translation tool, with the aim of promoting the use of regional languages in legal proceedings.25 In 2023, the Supreme Court successfully translated all of its 36,000 judgments into Scheduled Languages through the E-SCR portal.26 The translation process began in January with 2,238 judgments and by the end of the year, over 31,000 judgments had been translated, marking an exceptional level of progress. Currently, the highest number of translated judgments are in Hindi at 22,396, followed by Punjabi (3,572), Kannada (1,899), Tamil (1,172), and Gujarati (1,112).27In 2023, despite a notable improvement in translation speed, legal professionals raised concerns regarding the effectiveness of translated judgments. This is particularly relevant when High Courts are unable to carry out procedures in regional languages, especially in states where Hindi is spoken.28 According to Article 348(1)(a) of the Indian Constitution, it is mandated that all hearings in the Supreme Court and every High Court must be conducted in the English language. Article 348(2) grants the State Governor the authority, subject to the President's approval, to authorize the usage of Hindi or any other official language of the state in the High Court of that particular state. Therefore, SUVAS was granted authority according to Article 348 of the Constitution.29
What Function Does Artificial Intelligence Serve Within the Context of Law?
Artificial intelligence may be utilized in diverse ways to cater to various requirements. There is an abundance of exaggerated marketing claims around the use of artificial intelligence in the context of courts.30 It has been asserted that this would significantly enhance impartiality, as AI, in contrast to human judges, does not experience fatigue or depend on glucose levels for functioning. That is mostly speculation.31 Nevertheless, the subject at hand mostly centers around our existing comprehension derived from evidence. The focus lies on well-established technology, such as artificial intelligence (AI) that has already demonstrated its practical use. Can robots, however, possess the ability to make judgments? Regarding this matter, the verdict is yet undetermined.32
Currently, courts are responsible for many random activities, such as making interim rulings, at different phases. However, there are a few essential jobs that may be accelerated utilizing computers powered by artificial intelligence (AI) to streamline the whole decision-making process. The following items are:
i. Examining and evaluating the contents of a written document
Already, corporations such as Microsoft have created software that can not only read and comprehend documents like a human, but also, and even more effectively, respond to inquiries based on the document's content. The human parity milestone was accomplished by a Microsoft Research Asia team utilizing the Stanford Question Answering Dataset, commonly referred to as SQuAD.33 This indicates that the machine's performance in reading and comprehending data was deemed equivalent. Indeed, the machine achieved a score of 82.650 on this evaluation, while its human counterpart, employing the exact same set of queries and answers, lagged behind by 82.304. Alibaba researchers have also made progress, as they have developed an AI-powered machine that achieved 82.440. Companies' increased investment in Machine Reading Comprehension will not only improve these scores, but also a machine's capability to aid humans in reading and comprehending enormous quantities of data. 34This capability of a machine can be utilized in courts to peruse and analyze parties' pleadings in order to highlight readily delineated points of agreement and disagreement. This instrument is also capable of analyzing and reading evidence presented by both parties.35 In cases involving voluminous pleadings and evidence, this will be of particular benefit to the court, as it will not only conserve the court's time but also allow the judge to reassess his own understanding of the situation to ensure he is making a more certain decision.36
ii. Providing a summary of the document's contents
A court must thoroughly examine various documents, including pleadings, oral and written evidence submitted to the court, relevant legal precedents, and applicable laws, as part of its regular process in reaching a decision. Consequently, a significant amount of the courts' time is consumed by perusing the contents of the documents. The court must carefully analyze every document in order to establish a preliminary opinion about a case, with the intention of issuing interim orders.37 While the judge will ultimately need to thoroughly examine the document before making a final decision based on its merits, it would be highly beneficial if a machine could provide a summary or abstract of the document to assist the judge in making interim decisions such as issuing temporary injunction orders in a civil litigation or determining a prima facie case in a criminal case when summoning the accused.38 In addition, a computer capable of efficiently summarizing the contents of a document would greatly benefit the court. It would allow for a larger number of precedents to be reviewed in the same timeframe, so enhancing the quality of legal study.39 Summarizer is an AI program now utilized for condensing the information of papers. Further sophisticated software may be created using a similar approach, specifically tailored to meet the courts' requirements for efficient material summarization.40
iii. Using visualization and predictive analytics for settlement:
Law firms and lawyers use big data analytics to display case data, identify cited cases, and determine positive or unfavorable verdicts. Courts can utilize predictive analytics to offer probable outcomes to litigating parties, leading to out-of-court settlements. This not only helps attorneys optimize their clients' ROI by predicting trial outcomes, but also saves court time by pushing parties to reach an amicable settlement.41
Challenges Related to AI Integration
This section highlights the ongoing and anticipated obstacles that have emerged worldwide in areas where certain applications of AI have already been implemented. These difficulties are frequently interconnected and, as a result, require comprehensive frameworks to be tackled, rather than individual and fragmented solutions.
i. Ensuring transparency and clarification
The sources and outputs of AI-driven technologies are sometimes known, but the system that changes them from one to the other is not. This problem of not being clear is also known as the "black box" situation.42 Because of this, the reasoning behind and mathematical codes for these algorithms are often kept hidden, making it hard to question the algorithm's foundation. The business sector's participation makes things even less clear and makes the problem harder to solve.43 Since these kinds of technologies are usually created by private companies, they are usually covered by trade secrets. This creates a legal black box. For example, Northpointe, a private company that built COMPAS.44 This problem with being clear leads to a related problem with being able to explain things. For example, even if the formulas were made public, most people would not know how to use them or understand how they work. This means that the people who make AI solutions don't have as much information as the people who use those solutions. This makes it impossible for users to look closely at or question them.
ii. Avoiding data and design biases that promote social inequality.
AI/ML systems have the potential to perpetuate prejudices, whether it be unwittingly or purposefully, which poses a threat to fairness within the legal system. Due to their frequent utilization of extensive datasets, these systems have a tendency to reproduce the same biases that were initially present in the original datasets.45
iii. Licenses and accountability-related queries
AI systems, in contrast to licensed attorneys, are not required to obtain a legal practice permit, therefore exempting them from adhering to ethical standards and professional rules of behavior. In the event that an AI system delivers erroneous or deceptive legal counsel, who will bear the responsibility or be held accountable for it? Who is being referred to: the developer or the user? According to a recent news source, a lawyer from New York employed ChatGPT for legal research and incorporated six case citations in a brief submitted to the court. Nevertheless, the opposing counsel was unable to locate any of the cases, and the lawyer had to acknowledge that he had not personally verified their authenticity. 46The judge levied sanctions on the implicated attorneys, resulting in a collective punishment of $5,000 imposed on their law practice. Hence, attorneys must exercise prudence while employing generative AI for legal research.47
The Risks of AI Taking on a Bigger Role in Court?
There are opposing viewpoints regarding whether AI can address or circumvent human biases by adopting a data-centric strategy. "Furthermore, AI has the potential to generate biased or inappropriate results due to various factors," explains Dr. Felicity Bell, a Research Fellow specializing in the Future of Law and Innovation in the Profession (FLIP) research stream at the Law Society of NSW.48 Although Al may have the authority to make legal decisions in certain instances inside the justice system, it does not imply that this is inherently advisable. Several artificial intelligence systems and prediction algorithms in machine learning are created by utilizing preexisting datasets or historical data.49 Although this strategy may seem obvious, its effectiveness heavily relies on the specific type and quality of the data provided. Discovering correlations, which are observable connections, within sets of data is a prevalent application of machine learning and big data. Regarding crime data, this could result in inaccurate positive identifications and diminish the data's efficacy in uncovering the root causes of crime.50 According to reports, The Al judge, created by research scientists at University College London, examined previous legal cases and rendered the same decision in 79 percent of cases. It has been shown that European Court judges prioritized non-legal facts over legal arguments. Occasionally, Al judges can provide a glimmer of light for a judiciary that is overwhelmed with responsibilities. An important problem with AI judges is that it gives rise to significant ethical considerations around prejudice and autonomy. There has been a debate regarding Al.51
The question of whether artificial intelligence technology can replace legal professions, such as judges, hinges on the belief regarding the possibility of formulating a behavioral pattern through the analysis of numerous comparable legal cases.52 If these patterns are effectively created, then artificial intelligence technology will be capable of adopting these patterns.
AI systems and predictive algorithms often undergo training through the utilization of pre-existing datasets or historical knowledge. Although this strategy may seem reasonable, its effectiveness is highly dependent on the specific characteristics and reliability of the provided data.53 . Humans frequently exhibit the same logical fallacy as a mechanical replica may potentially exhibit. An illustrious illustration is the association between a diminished income and an individual's inclination towards criminal behavior. While poverty may not directly cause criminal behavior, it can indirectly contribute to it by creating circumstances that increase the likelihood of crime. If inadequately addressed, such correlation flaws might lead to an AI-based law enforcement decision or judgment descending into a detrimental cycle of imposing excessively harsh or mild penalties.54
Conclusion
To summarize, the integration of artificial intelligence (AI) into India's legal system offers a dynamic field of opportunities and obstacles. The application of artificial intelligence, namely in the field of machine learning, has the potential to greatly enhance the efficiency of the judiciary, reduce the number of pending cases, and improve the availability of justice. The judiciary's dedication to adopting AI for office tasks and offering valuable support in decision-making processes is demonstrated by pioneering programs like SUPACE and SUVAAS. However, the integration of AI in the legal field is not without complications. Ensuring transparency and clarity, particularly in understanding the complex nature of AI algorithms, is a significant challenge. The obscure development and operation of these systems give rise to important problems regarding accountability and the ability to contest algorithmic conclusions. Furthermore, it is crucial to acknowledge and rectify any biases that are ingrained in the data and design of AI systems in order to avoid the ongoing growth of disparities in socioeconomic status within the legal field. Notwithstanding these difficulties, the potential benefits of AI in India's legal system are of great significance. The utilization of AI for tasks such as document analysis, summarization, and predictive analytics offers the potential to accelerate decision-making processes. The Indian judiciary's ongoing initiatives, such as the e-Court project, demonstrate a dedication to utilizing technology to bring about a significant change in the delivery of justice. In order to address the ethical and technological implications associated with the integration of AI, it is necessary to establish a cooperative approach involving policymakers, legal professionals, and engineers. It is crucial to achieve a careful equilibrium between utilizing the advantages of AI and upholding principles of transparency, fairness, and responsibility. India's legal landscape should prioritize the development of public trust while integrating AI. This means ensuring that technological advancements contribute to a legal system that is not only technologically advanced but also fair, accessible, and trustworthy.
*****
Footnotes
2. A. D. (Dory) Reiling, Courts and Artificial Intelligence, International Journal for Court Administration (2020).
3. Haroon Sheikh, Corien Prins & Erik Schrijvers, Artificial Intelligence: Definition and Background (2023).
4. Mohsin Kabir, M.F. Mridha & Md. Tanzib Hosain, Explainable AI Approaches in Deep Learning: Advancements, Applications and Challenges (2024).
5. Md. Arif Imam, The Integration and Impact of Artificial Intelligence in the Indian Judiciary, (Jan. 1, 2024), https://www.lawinsider.in/columns/the-integration-and-impact-of-artificial-intelligence-in-the-indian-judiciary.
6. Adam Zewe, Does This Artificial Intelligence Think like a Human?, (Jun. 6, 2022), https://news.mit.edu/2022/does-this-artificial-intelligence-think-human-0406.
7. Justice Suraj Govindaraj, Integrating Artificial Intelligence in Court Processes: Challenges/Opportunities and Issues/Possibilities.
8. Id.
9. Aditi Prabhu, Artificial Intelligence in the Context of the Indian Legal Profession and Judicial System, (2023), https://www.barandbench.com/columns/artificial-intelligence-in-context-of-legal-profession-and-indian-judicial-system.
10. Marion Oswald et al., Algorithmic Risk Assessment Policing Models: Lessons from the Durham HART Model and ‘Experimental’ Proportionality (2017).
11. supra note 6.
12. Id.
13. Id.
14. Victor Habib Lantyer, The Era of Artificial Intelligence in Law: Brazil in a Global Context (2023), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4650117.
15. Id.
16. supra note 6.
17. Peter Lubrich & Tiffany Vlemmings, SOCRATES 2.0 - New Services on Traffic Information and Traffic Management (2019).
18. Dongho Choi, Ziad Matni & Chirag Shah, SOCRATES 2.0: Bridging the Gap between Researchers and Social Media Data through Natural Language Interactions (2016), https://doi.org/10.1002/pra2.2015.1450520100139.
19. Id.
20. Ameen Jauhar et al., Responsible Artificial Intelligence for the Indian Justice System, Vidhi Centre for Legal Policy.
21. Id.
22. OptimizeIAS Team, SUPACE, (2021), https://optimizeias.com/supace-2/.
23. Vishalakshi Rituraj & SHYAM KRISHNA SINGH, SUPACE: AN ARTIFICIAL INTELLIGENCE SOLUTION FOR REFORMING INDIAN JUDICIARY (2021).
24. supra note 21.
25. ABHINAY LAKSHMAN, SC’s Translation Projects Picked up Good Speed in 2023, The Hindu, https://epaper.thehindu.com/articleshare?articleurl=https%3A%2F%2Fepaper.thehindu.com%2Fccidist-ws%2Fth%2Fth_chennai%2Fissues%2F65613%2FOPS%2FGQHC7DFA0.1%2BGTSC7DS95.1.html.
26. Shritama saha, The Supreme Court of India Gets A New AI Portal, SUVAS, (Dec. 22, 2022), https://analyticsindiamag.com/the-supreme-court-of-india-gets-a-new-ai-portal-suvas/.
27. supra note 24.
28. Supreme Court Vidhik Anuvaad Software (SUVAS), (2024), https://www.drishtijudiciary.com/current-affairs/supreme-court-vidhik-anuvaad-software-suvas.
29. Id.
30. Saloni Sharma, Surabhi Shrivastava & Pradeepika Verma, A Comprehensive Analysis of Indian Legal Documents Summarization Techniques, Springer Nature Singapore Pte Ltd (2023).
31. Sanjeev Ghanghash, Artificial Intelligence in Administration of Justice in India: Challenges and Opportunities, The Law Brigade Publishers (2022).
32. Id.
33. Parth Jain, Artificial Intelligence for Sustainable and Effective Justice Delivery in India, OIDA International Journal of Sustainable Development.
34. Id.
35. Shrivastava, supra note 29.
36. Jain, supra note 32.
37. Id.
38. supra note 30.
39. Khusi Patel & Aahini Gandhi, ANALYSING THE IMPACT OF ARTIFICIAL INTELLIGENCE ON LEGAL RESEARCH AND LEGAL EDUCATION, The Law Brigade Publishers.
40. Jain, supra note 32.
41. Id.
42. supra note 19.
43. Id.
44. Id.
45. Id.
46. supra note 8.
47. Id.
48. DAWN LO, Can AI Replace a Judge in the Courtroom? UNSW Sydney, 2021, https://newsroom.unsw.edu.au/news/business-law/can-ai-replace-judge-courtroom.
49. Kabir, Mridha, and Hosain, supra note 3.
50. Cary Coglianese, Maura R. Grossman & Paul W. Grimm, AI in the Courts: How Worried Should We Be?, Bolch Judicial Institute Duke Law School (2024).
51. Sachu Saji Thomas, Al AND REVOLUTION TO JUDICIAL SYSTEM, Indian Journal of Law and Legal Research.
52. Aamir Khan, AI-Powered Indian Judiciary: A Step Forward or Cause for Concern?, (Jun. 6, 2023), https://www.barandbench.com/columns/litigation-columns/ai-powered-indian-judiciary-a-step-forward-cause-concern.
53. Christopher McFadden, Can AI Be More Efficient Than People in the Judicial System?, (2020), https://interestingengineering.com/innovation/can-ai-be-more-efficient-than-people-in-the-judicial-system.
54. supra note 19.
