Use Of Artificial Intelligence As A Tool To Meet The Ends Of Law Enforcement And Justice

Authored By - Dr. Raju Narayana Swamy IAS

Introduction

            With the advent of big data analytics, machine learning and artificial intelligence (AI), the fundamental questions of law enforcement and justice are being reconsidered across the globe. Law is based on two important aspects – predictability and precedence and many are of the opinion that AI can greatly help align these processes. While disagreements are galore as to whether these technologies represent a panacea or whether they will further exacerbate social divisions and endanger fundamental liberties, the two camps agree that the new technologies usher in important consequences. In fact, there are three main ways in which technology is already reshaping the judicial system. First and at the most basic level, technology is assisting to inform, support and advise people involved in the justice system (supportive technology). Second, technology can replace functions and activities that were previously carried out by humans (replacement technology) – the concept of online courts being a classic example. Finally, at a third level, technology can change the way that judges work and provide for very different forms of justice (disruptive technology), particularly where processes change significantly and predictive analytics may reshape the adjudicative role. It is at these second and third levels that issues emerge in terms of the impact of technology on the role and function of a judge. Questions raised in this context include

1. Can AI enabled programmes extract the accurate position of law from a mass of precedents?
2. Can robots decide questions of law?
3. Who should be accountable for semi- automated decisions?
4. How should responsibility be allocated within the chain of actors when the final decision is facilitated by the use of AI?
5. Is the “due process of law” denied to the accused when AI systems are used at some stage of the criminal procedure?
6. Can judgements be replaced by data?

These questions are all the more relevant now that AI has made a lot of inroads within justice systems – in Estonia for adjudicating small claims (robot judges), in China, Russia and Mexico for giving legal advice/approving pensions, in Malaysia towards supporting sentencing decisions, in Austria for sophisticated document management, in Colombia and Argentina for identifying urgent cases within minutes, in Abu Dhabi for predicting probability of settlement and in Singapore for transcribing court hearings in real time- to name a few.

Artificial Intelligence : The Concept

            Among the several definitions of AI, one of the most relevant in the context of justice systems is the one given by the Commissioner for Human Rights:

“An AI system is a machine-based system that makes recommendations, predictions or decisions for a given set of objectives. It does so by

1. utilizing machine and /or human-based inputs to perceive real and /or virtual environments
2. abstracting such perceptions into models manually or automatically and
3. deriving outcomes from these models, whether by human or automated means, in the form of recommendations, predictions or decisions.”

Put in simple terms, AI is a type of computer technology which is concerned with making machines that carry out work in an intelligent way, similar to the way a human would. The technology has evolved from obeying pre-designed and pre-configured codes into a more sophisticated end product, imbued with human-like cognition. AI in order to work needs “big data”. Luc Julia, one of the creators of the digital assistant Siri evokes this image, “if a machine is to be able to recognise a cat with 95% certainty, we need about 1,00,000 pictures of cats.”[i] It needs to be mentioned here that AI comes in many different forms such as speech recognition and image recognition. There are two main strands to current AI technology – one being the rules-based approach and the other based on data analysis. The former underlies many of the document creation products that are used both by lawyers and the lay public, including products such as Legal Zoom. This technology more generally underlies what are called “expert systems” and faces inherent limitations. The latter looks for patterns in large bodies of data and finds relationships and correlations from which it can draw conclusions. This is the kind of AI that underlies products such as translation software and autonomous vehicles.

It needs to be mentioned in this context that the “strong” AIs of science-fiction literature do not exist. This type of AI which would be equipped not only with intelligence but also with conscience remains purely fictional. The machine learning systems currently being developed are described as “weak” AIs and are capable of extracting complex patterns and learning from large volumes of data efficiently and often with high levels of predictive accuracy.

Experiences in other jurisdictions

1. Estonia

Estonia has announced a plan to delegate some lower value claims to an online court powered exclusively by AI.[ii]The approach includes possible use of a de novo appeal to a human judge.[iii]While providing potentially enforceable actions, this amounts more to a form of AI powered mediation with the litigants free to pursue their legal claims if they are unhappy with the AI generated result. But this may differ from traditional mediation in that simply walking away may not be an option if the AI decree is enforceable – the disappointed litigant will need to invest in and commit to a human-driven judicial process in order to escape the decree.

1. China

China limits AI to specific kinds of easy cases where the decision parameters are simple and clear.[iv]In Zhejiang province, for example, several thousand dangerous driving and theft cases have been initially decided by AI software subject to review by a human judge. The large sample of similar cases and well –defined determinants of outcome have made this possible. Over time, the level of human review could conceivably be dialled down should experience show that these cases are really as easy as believed. The Chinese experience however points out that even for suggesting outcomes, not all cases are suitable for AI. In order to get to statistically significant results, there must be a large pool of cases with a limited number of factors that can affect outcomes. To put it differently, even in a country as committed to capturing data in its judicial system as China, not all kinds of cases are amenable. In fact, China's ambitious AI system that sought to model intentional murder failed as too few such cases existed to provide an adequate sample and as intentional murder presents in a multiplicity of ways making recognition and characterization more difficult.

China is also a pioneer in using AI to make sure that the resolution of a dispute by a particular court is in line with the results reached by other courts on similar facts and legal issues. Though China is nominally a civil law country, this process has the effect of bringing something like “stare decisis” to Chinese jurisprudence.    In some areas of China, AI robots greet visitors to the court house and help guide them to the appropriate location. There is no reason why this guidance cannot become more sophisticated and helpful over time thereby helping litigants to produce legal forms that are in accord with the requirements of the court and guide them with regard to the court process. Practical examples in this regard include a robot called Xiao Fa which was put in to operation at the law suit center at Beijing No.1 Intermediate People’s Court which can answer questions verbally or take queries on its screen with a key board and the AI enabled robot chatbot Fa Xiaotao using which Wusong Technology is working on digitizing the way courts function.

Mention also needs to be made here of the “smart court navigation system” and “intelligent push system” launched by the Supreme People’s Court in 2018, Beijing’s “rui judge” intelligent research system, Shanghai’s “206” criminal case intelligent auxiliary case system and Hebei’s “smart trial 1.0” trial support system. The smart court SOS in particular is a system powered by machine learning which is being used at various courts that connects to the desk of every functioning judge across the country. It automatically screens out cases for the purpose of reference, recommends relevant laws and regulations, drafts legal documents and alters perceived human errors if any in a court’s judgement. Moreover through facial recognition techniques, Chinese officials have evolved systems to figure out suspects from an ocean of individuals inside an arena which without man-made intelligence innovation would have been impossible. China is also experimenting automation in prisons. A jail that houses some of China’s most high profile criminals is reportedly installing an AI network that will be able to recognize and track every prisoner round the clock and alert guards if anything seems out of place.

1. USA

Judiciary in the US has already started using an AI programme called Public Safety Assessment (PSA) before deciding whether or not an accused should be released on bail.[v] This software calculates the risk of recidivism and flight ( ie) risk of defendant again committing the crime and risk of his escape from the clutches of law. The AI software calculates the risk score by taking into account several factors such as

1. whether the current offence is violent
2. whether the person had a pending charge at the time of the current  offence
3. whether the person has a prior misdemeanour/felony conviction
4. person’s age at the time of arrest
5. how many times the person failed to appear at a pre-trial hearing in the last two years

Other factors which can assist judges in arriving at bail, parole and probation decisions can also be incorporated into the algorithm.

            Mention must also be made of the SSL (Strategic Subject List) introduced in Chicago to predict those individuals who are likely to be involved in gun violence. US also makes use of “e Discovery” – an automated investigation of electronic information before the start of a court procedure.  e Discovery relies on machine learning. A more controversial tool COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) has been used to assess recidivism risk. COMPAS takes in 137 items of information and comes up with a risk score from 1 to 10. However, it has been criticized as biased against black defendants. In fact, the NGO Pro Publica analysed COMPAS assessments and published an investigative report arguing that the algorithm was racially biased.

No discussion on COMPAS will be complete without a reference to the judgement in Loomis vs Wisconsin (2016). The algorithm identified Loomis as an individual who presented a high risk to society and the first instance Court decided to refuse his request to be released on parole. In the appeal, the Supreme Court of Wisconsin decided that the recommendation from the COMPAS algorithm was not the sole ground for refusing his request to be released on parole and hence the decision of the Court did not violate Loomi’s due process right. The Court was in fact neglecting the strength of the automation bias. By claiming that the lower Court had the possibility to depart from the proposed algorithmic risk assessment, the Court ignored the social psychology and human-computer interaction research on the biases involved in all algorithmic decision making systems which shows that once a hi-tech tool offers a recommendation, it becomes extremely burdensome for a human decision maker to refute such a recommendation.

1. UK

A tool called HART (Harm Assessment Risk Tool) has been used by the UK to forecast which criminals are most likely to reoffend and suggest what kind of supervision a defendant should receive in prison.[vi] The tool uses random forest forecasting which is a machine learning technique. HART was developed in collaboration with academicians at Cambridge and is built on five years’ worth of data on people. It makes predictions on the basis of 33 different metrics, 29 of which relate to past criminal history and the remainder to demographic data such as the individual’s age, gender and postcode[vii]The usage of postcodes  as a metric of analysis has garnered repeated criticism of this tool as critics argue that such a situation amplifies existing patterns of offending.[viii]The application has also shown clear difference in opinion between human and algorithmic forecasts. In fact, comparison in the use of the algorithm found that the model and officers agree only 56.2% of the time.

1. Brazil

              Brazil uses an AI tool called VICTOR to conduct preliminary case analysis to reduce the burden on the Court.[ix] The tool supports the Brazilian Supreme Court by providing analysis of cases that reach the Court using document analysis and natural language processing tools. The goal of this tool is to accurately and quickly track resources that deal with issues of “general repercussions. “ This concept is intended to ensure that only questions that are truly relevant to the wider society are heard by the Court. To put it a bit differently, appeals that reflect only the unsuccessful party’s unwillingness to accept defeat are excluded.

            VICTOR has been highly beneficial for courts in Brazil. Earlier, the exercise was conducted by civil servants based on binding precedents and would take about 40 minutes for each law suit. VICTOR can do this exercise in five minutes.

            However, a criticism in the pipeline is that appellants are not informed when VICTOR is used as its pilot version randomly picks up appeals to evaluate. This is in possible violation of the Brazilian Data Protection Law which demands that decision making should be fair, transparent and informed.

1. A few other countries

                  In the Middle East , in collaboration with the private sector, Abu Dhabi Judicial Department as a part of their “Justice Intelligence” Project has been using technology to predict the possibility of settlement of cases.[x]The tools that are being used can predict the probability of settlement by upto 94% of the time. In Argentina and Columbia, a tool called Prometea has been used by the Public Prosecutor’s Office of Buenos Aires and the Constitutional Court of Columbia respectively to identify urgent cases in just two minutes which would normally take a human being 96 days.[xi]In Singapore, a speech translation system has been deployed by Courts. The tool utilizes neural networks trained with language models and domain- specific terms to transcribe court hearings in real time, thus allowing judges and parties to review oral testimonies in Court instantaneously. [xii]In countries like Russia and Mexico, robots are providing services like legal advice to citizens and aiding judges to identify if pensions should be granted. In Austria, AI is being used for sophisticated document management and as a digitization assistant of existing analogue files. In Malaysia, AI is being used to support sentencing decisions.[xiii]

The Indian Saga

            AI in the Indian judiciary is still in its nascent stages of development. The advent of AI within Indian Courts was initiated on 26th November 2019 with the launch of a neutral translation tool called SUVAAS (Supreme Court Vidhik Anuvaad Software) which has been trained using machine learning processes. It has the capability of translating English judgements and daily orders into nine vernacular scripts. The Supreme Court’s AI Committee is also working on a composite new tool named SUPACE (Supreme Court Portal for Assistance in Court Efficiency) which will target different processes like data mining, legal research, projecting case progress etc.

            The AI-powered work flow of SUPACE has four parts:

1. File Preview

The case files, typically available as PDFs, can also be converted into text. There is also a search tool to browse through all files

1. Chatbot

The text and voice enabled chatbot helps to give a quick overview of the case in a matter of minutes by answering simple questions such as “ What is the matter about?” or “Which fundamental rights of the petitioner are violated?”. Chatbot can switch between documents to fetch the right answer while allowing the user to check the source of the answer. This bot suggest further questions to be asked for better understanding and the entire question summary can be printed by the user.

1. Logic Gate

This fact extraction system of the chatbot is divided into four parts : Synopsis, FAQs, Evidence and Case Law. These give information about the case such as overview, chronology, judgement and so on. With enough training and refinement of the algorithm, there will come a time when any and every question, factual or contextual, will be answered by the chatbot.

1. Notebook

This is the integrated word processor which truly makes the tool an end –to- end system. A brief summary of the case can be prepared by collating all information auto-extracted from the database using the AI .In addition, voice dictation can be used to prepare notes on this comprehensive drafting tool. Therefore without typing a word, a summary document is ready as a soft or hard copy.

On the research front, IIT Kharagpur has released an AI- based method to automate reading of legal case judgements. It is based on deep neural models to understand the rhetorical role of sentences. It also needs to be mentioned here that the pandemic has led to a surge in discussion around increasing digitization through the e Courts Project, creation of virtual courts and the potential of Online Dispute Resolution (ODR) and that within this conversation, AI has become an increasing talking point. However, a word of caution is needed here – while pilots are promising, there is a need to identify steps for scaling these technologies and their adoption.

Is AI a Panacea?

            The question revolves around a fundamental issue: Will we as a society ever be willing to delegate fundamental rule - making powers and assign assertion of the legitimacy of the state to non-human entities ? Only by matching the real potential of AI with the full range of judicial functions can we give a non-hyperbolic assessment. The reality is that the current capability of AI is limited to specialised tasks and the roles of judges are so generalised that there is no near-term possibility of AI wholly and satisfactorily displacing judges in high stake cases. The function of AI with the best evidence so far of success is structuring large amounts of information which could make administration of justice more efficient. In other words, AI can be used partially – where human discretion may not be needed. An evolutionary step forward can be advancement in predictive big data policing instruments. There can be a two pronged strategy behind this. First, advancements in AI promise to make sense of enormous amounts of data and to extract meaning from scattered data sets. Second, they can be aimed at regulation of society at large and not just the fight against the crime. A classic example is ‘function-creep’ , Singapore’s total information awareness system programme.

            AI tools can be used to penetrate deeply into the preparatory phase of the crime which is yet to be committed as well as to scrutinize already committed crimes. A distinction is however to be made between tools focusing on risky individuals (heat lists) and tools focusing on risky places (hot spots). A leaf can perhaps be drawn out of the experience of the Dutch children’s rights organization Terre des Hommes which was the first NGO to combat webcam child sex tourism by using a virtual character called Sweetie.

            AI is not a kind of magic; it is technology with the capabilities and limitations inherent therein. The intelligence of AI is limited by the design and input of human programs. An excessive reliance on such systems may result in legal issues being decided by computer programs. The fact is that “algorithmic partiality” is entrenched in the method AI algorithms work.[xiv]For example in 2015, Google’s photograph application inaccurately labelled a photo of two dark individuals as gorillas since its calculation had not been prepared with suitable pictures of people of dark skin. In another matter, the AI judged beauty contest for the most part picked white members as victors since its training was done on pictures of white individuals. [xv] These tools have the capacity to become prescriptive, potentially over shadowing case specific reasoning and reduce judicial decisions to purely statistical and algorithmic outcomes. In fact, there are two significant issues with a pervasive integration of AI within the justice system – value lock-ins which can stagnate legal and jurisprudential evolution and the alteration of the judiciary’s constitutional role under the doctrine of separation of powers. The impact of AI systems may have distorting effects on the fundamental cornerstones and architecture of liberal democracies including the limitation of political power by the rule of law. Human rights may also be impacted through the use of automated processing techniques and algorithms. To be more specific, in an AI –centric judiciary while fostering the rule of stare decisis, it is a plausible fall out that precedents become stagnant. Moreover as Yutaka Matsuo, a Japanese scholar pointed out, civil litigation especially divorce or property inheritance disputes may be better handled by people because it involves a lot of emotional factors.

The Road Ahead

            The role of a judge is a complex one. It can incorporate activism, interactions with people, dispute settlement, case management, public and specific education activities, social commentary as well as adjudicatory functions that may be conducted with other judges or less commonly in some jurisdiction with lay people. Given this variation, it is difficult to determine how developments in AI may reshape the judicial role. Human judges are temperature-saavy and as participants in social life, they have the same ability to empathize with the general public and are committed to achieving empathetic justice. To put it a bit differently, against the backdrop of AI’s full penetration into the judiciary, the concern is not that “machines are starting to think like humans, but that humans are beginning to lose their unique ability to think and become vassals of machines.” Everyone and every case is unique. Each requires human judgement and the vital and very natural ability to empathy that AI systems cannot provide. Although AI will play a key role in the trial process in the years to come, it will be in a subordinate position – only assisting the judge to handle the case – the judge continuing to be the key and core of the trial work. Thus co-bots rather that robots will play a role when we intend to apply AI to the field of law. The road ahead is perhaps one of utilizing AI in addition to people – as airplane pilots use autopilot.

            Technological advances in the form of AI based programs will benefit the judges with increase in memory, enhancement in the ability to manage and process information as well as reduction in occurrence of fatigue. In international arbitration, the use of AI has been predicted for a variety of tasks, including appointment of arbitrators, legal research, drafting and proof-reading of written submissions, translation of documents, case management and document organization, cost estimations, hearing arrangements (such as transcripts or simultaneous foreign language interpretation) and drafting of standard sections of awards. There are several processes in the administration of law that are repetitive and mechanical (such as scheduling of hearings and e-filing) and the use of machine learning in these mechanisms can greatly enhance efficiency and improve case flow management. Decision making in areas like rate of taxation can be covered by AI and thus technology can play a significant role in tribunals like ITAT (Income Tax Appellate Tribunal). AI models like computer vision and natural language processing can be used to generate statements and evidence documents.  Leveraging such technology will help to reduce human dependency and errors in information input. To put it a bit differently, the initial focus of use of AI when it comes to improve the work of judges must be on document/case management and research.

            In fact, the main categories where AI can be used at present are:-

1. Advanced case – law  search engines
2. Online Dispute Resolution
3. Assistance in drafting deeds
4. Analysis (predictive, scales)
5. Categorization of contracts according to different criteria and detection of divergent or incompatible contractual clauses.
6. Chatbots to inform litigants or support them in their legal proceedings.

The following tabulation speaks volumes for itself.

Application and Examples of AI in Legal Sector[xvi]

Over a period of time, these uses can lead to more sophisticated ones that require considerable methodological precautions as well as uses to be considered following additional in depth scientific studies. A day may also come when AI can be put to use successfully in areas considered today with most extreme reservations- use of unbiased algorithms in criminal matters in order to profile individuals being a classic example.

At one level, AI is an alien form of intelligence and will be – even if it achieves generalized capabilities – no more like humans than reptilian visitors from another galaxy would be. To have such an intelligence create and extend laws, despite being so far removed from being a member of the body politic, comes up against the legitimacy of the judicial system. Whether our societies are ready to accept that involves issues far beyond technological capability. This is especially true in today’s milieu wherein the neutrality of AI remains constrained and where justice is like the top of a mast, swinging violently at the slightest movement of the hull.

References

1. Zichun Xu (2022) Human Judges in the Era of Artificial Intelligence: Challenges and Opportunities,  Applied Artificial Intelligence, 36:1,2013652.

1. Christopher Rigano (2019) Using Artificial Intelligence to Address Criminal Justice Needs, NIJ Journal 280

1. Mvea U (2018), Japan Considers Crime Prediction System using big data and AI, Japan Times, 24 June 2018 (Online)

1. Zavrsnik A (2019), Algorithmic justice: algorithms and big data in criminal justice settings, Ekr.J.Criminol.

1. Dewan S (2015) Judges Replacing Conjecture with Formula For Bail, The New York Times

1. Ferguson AG (2017), The Rise of Big Data Policing : Surveillance, Race and the Future of Law Enforcement, NYU Press, New York

1. Freeman K (2016), Algorithmic Injustice : How the Wisconsin Supreme Court failed to protect due process rights in State Vs Loomis, NCJ.Law Technol 18(5)

1. Tania Sourdin and Archie Zariski (eds)(2018), The Responsive Judge : International Perspectives, Springer Nature
2. Aryan A (2019), Law firms take baby steps in AI at increase efficiency and cut costs http://www.nishithdesai.com

1. Branting K (2003), An Agenda for Empirical Research in AI and Law in evaluation of Legal Reasoning and Problem Solving Systems

1. Hali AC (2018), How law firms can benefit from Artificial Intelligence, https://www.lawtechnologytoday.org/2018/11/how-law-firms-can-benefit-from-artificial-intelligence/

1. Hubert Dreyfus (1992) What Computers Still Can’t Do : A Critique of Artificial Reason, Cambridge, The MIT Press.

[i] L.Julia,L’intelligence artificielle n’existe pas, First Edition, Paris 2019

[ii] Erie Niiler, “Can AI be a Fair Judge in Court ? Estonia Thinks So “, WIRED (March 25, 2019,7.00 AM,https://www.wired.com/story/can-ai-be-fair-judge-court-estonia-thinks-so)

[iii] Tara Vasdani, From Estonian AI Judges to Robot Mediators in Canada,  UK, The Lawyer’s Daily, https://www.lexisnexis.ca/en-ca/ihc/2019/06/from-estonian-ai-judges-to-robot-mediators-in-canada-uk.page

[iv] Jinting Deng, ”Should the Common Law System welcome Artificial Intelligence : A Case Study of China’s Same – type Case Reference System” 3 GEO.L.TECH REV 223 (2019)

[v] Arnold Foundation Launches Expansion of Public Safety Assessment Tool https://the crimereport.org/2018/04/25/arnold-foundation-launches-expansion-of-public-safety-assessment-tool/

[vi] Marion Oswald and others “Algorithmic risk assessment policing models: lessons from the Durham HART model and Experimental proportionality” (2018)2 Information & Communication Technology Law

[vii] Patricia Nilsson, “UK police test if computer can predict criminal behaviour”, Financial Times 6th February 2019.

[viii] Alexander Babuta, Marion Oswald and Christine Rinik, “Machine Learning Algorithms and Police Decision-Making: Legal, Ethical and Regulatory Challenges” (Royal United Services Institute for Defence and Security Studies 2018.

[ix] Daniel Becker and Isabela Ferrari, “ VICTOR: the Brazilian Supreme Court’s Artificial Intelligence: a beauty or a beast? “ https:// sifocc.org/app/uploads/2020/06/ Victor- Beauty- or-the- Beast.pdf

[x] ADJD leverages judicial insights to transform the justice delivery process (SAS Institute)www.sas.com/en-ae/customers/adjd-judicial.html

[xi] Juan Corvalan and Enzo Cervini, Prometea experience: Using AI to optimize public institutions (Cendap, 1st May 2020) and Irma Isabel Rivers, the implementation of new technologies under Columbian law and incorporation of artificial intelligence in judicial proceedings ( International Bar Association, 5 November 2020)

[xii] Michelle Chiang “State Courts and A STAR’s Institute for Infocomm Research Collaborate to Develop Real-time Speech Transcription System for Use in Courts”(State Courts, Singapore, 14 December 2017)

[xiii] Olivia Miwil, “Malaysian judiciary makes history, uses AI in sentencing”, New Straits Times, 19 February 2020.

[xiv] Dickson, B (2018)”What is algorithmic bias?” Available at https://bdtechtalks.com/2010102126/racist-ai-deep-learning-algorithms/

[xv] Artificial Intelligence beauty contest does not like black people, The Guardian, 8 September 2016

[xvi] Geethanjali Chandra, Ruchika Gupta and Nidhi Agarwal (2020), Role of Artificial Intelligence in Transforming the Justice Delivery System in COVID-19 Pandemic, International Journal on Emerging Technologies,11(3)