The integration of Artificial Intelligence in the Indian judiciary: In conversation with CEO of MCIL, Manthan Trivedi
The challenges that plague the judiciary in India are not new, and the pandemic has only accentuated them. To counter the ever-burgeoning pendency of cases, it is only a matter of time before Artificial Intelligence is adopted to aid judges in overcoming the burden of case load that they are currently facing.
Earlier this year, former Chief Justice of India SA Bobde had envisioned the use of Artificial Intelligence (AI) in the judiciary. During that speech, he expressed his delight at the launch of the Supreme Court Portal for Assistance in Court Efficiency (SUPACE).
ManCorp Innovations Lab (MCIL), the brainchild behind this AI-powered solution has in the past assisted the Jharkhand and Patna High Courts with similar solutions to meet the challenges being faced by the judiciary.
Bar & Bench caught up with MCIL Chief Executive Officer Manthan Trivedi to talk about the journey of the deep tech AI startup, which is studying varied industries, especially the legal fraternity. In this interview, he speaks about the idea that inspired the start up and the AI assistance rendered thus far to the Patna and Jharkhand High Courts as well as at the Supreme Court of India.
Trivedi has a diverse background, having studied hospitality, management and analytical science. After studying data-based decision making at Harvard University, he came back to India in 2016 to start a venture that “used data for understanding decision-making to identify problems and solutions to resolve them.”
“I returned to India to try and see if similar systems can be used in India or if they are already used in India, and how we can improve upon them. We essentially set out to bring about data-based decision making and were giving multiple talks and seminars on this subject, at which point we met individuals from the judiciary who posed questions on how such systems could help the legal fraternity,” he recalls.
Trivedi was part of a team that conceptualized a digital mechanism to footprints of all activities done online. The idea was to give the judges in India an interface on which they could work so that minimum time is spent in reading files to extract facts and preparing synopses, etc.
“The system is integrated with machine learning capabilities where it would follow the footsteps of the user and try to replicate and automate that process, incrementally decreasing the overall time taken in order to deal with these cases. That is how the concept of AI and judiciary began,” he said.
Even before registering as a company, the team at MCIL relentlessly engaged in identifying the different bottlenecks that India as an economy was facing, while also understanding the different industries that were willing to adopt a new methodology for their daily functions.
“We saw interest from the legal fraternity and especially from the judiciary in integrating AI for swifter functioning. It was very interesting for us; we understood that the functioning of the legal fraternity has an overarching impact across industries, so it was very intriguing for us to approach this,” Trivedi said.
As a matter of serendipity, towards the end of 2018, the team at MCIL chanced upon a tender issued by the High Court of Jharkhand requesting for a smart court solution using the latest technologies.
"We found it very interesting as at that point, we were already looking for projects in the public sector and this came about as good news. We pitched for this tender and got awarded the same,” he fondly reminisced.
Studying the problems faced by the Jharkhand High Court, they learnt that there was a dearth of judges and very high volumes of criminal cases. The company developed had two pieces of technology:
1. Optical Character Recognition (OCR) - which converts scanned documents into computer readable text, corrects the orientation, etc.
2. ChatBot - controlled by both voice commands and text.
In order to make this operational, an exhaustive list of about 120-150 questions that would naturally crop up in a judge's mind when looking at criminal cases, especially cases of murder, were framed. Thereafter, a bunch of similar cases were labelled in the system, which made reading of files a smooth and quick endeavour.
Elaborating on how the system assisted judges in dealing with criminal cases, Trivedi said,
"The system will be able to tell you the specifics of a particular case, like number of victims, number of accused persons, what are the crimes committed by the accused, what was the severity of the injury caused to victim, how long did the police take to reach the scene of the crime, what was the cause of death according to post mortem, whether the blood samples collected matched the blood samples on the weapon match the wound, etc. This was the first project we undertook with the judiciary.”
Soon after this, in 2019, the Supreme Court formulated the AI Committee, which discussed the use of AI in judiciary, and the bottlenecks faced by members of the Registry and judges across various courts. The members of the Committee include retired judges, as well as sitting judges of various High Courts and the Supreme Court.
“We got a holistic perspective on multiple problem statements and we would try and solve these one by one,” Trivedi recounted.
In the meanwhile, the Patna High Court shared its problem statement on the need for developing a mechanism for speedy allocation of cases. The team at MCIL developed a product where the system auto-allocated cases looking at the historical record of the kind of cases dealt with by each Bench, and based on their rate of disposal, the algorithm would determine which judge could be allocated a particular type of case.
Finally, the solution called SUPACE for the Supreme Court came to fruition.
“This is an entire end-to-end system which has everything from e-office, user and task management to automatic conversion of documents, automated facts extraction, ChatBot, automatic creation of synopsis,” he elaborated. The launch of the product took place in April 2021 in the presence of the then CJI SA Bobde.
The many technologies developed by MCIL could prove crucial for judges, especially at a time when all courts have gone online and the use of paper has largely reduced.
A suit of technologies were developed around Optical Character Recognition (OCR), Natural Language Processing (NLP), Chatbot and Workflow, to name a few. Elaborating on the concept of ‘Workflow’, Trivedi mentioned that it seamlessly helps an individual replace everyday essentials that make a physical office operational with a digital platform.
“We devised a solution where everything from managing documents to assigning tasks to communicating via emails, video calls, instant messaging, reading and highlighting scanned documents, researching on any aspect in the files, adding notes, making illustrations, tables and charts and drafting documents or performing simple tasks like approval and rejections - all of these can be done on a digital platform which also enables the software to track an individual’s digital footprint, eventually leading to automation,” he added.
When one uses the system for 6-9 months, then 40-50% of the work being done on the platform gets automated, thereby giving a lot more mental space and time required to do work that requires actual application of mind. This is the product named ‘JUDi’ - an entire smart office solution powered by AI.
The product is currently being tested with law firms and the team at MCIL is getting their feedback on it.
“JUDi is primarily for the legal fraternity - since we have already worked with the judiciary in the past, we have some tricks up our sleeves, like some AI algorithms that are already relevant to the legal fraternity. So we are venturing there first, but at the same time, we are looking to introduce these technologies to other industries as well," Trivedi said.
The Natural Language Processing (NLP) technology is one that understands human language and is currently being trained to understand legal language. Everything from process automation to spell check, defect identification, auto-completion of forms, filtering of spam or categorizing documents, developing search engines and also auto-generating synopses for cases, is covered by NLP. In order to verify the accuracy of the results produced by the technology, Trivedi said,
“There are 3 stages to this -
1. There are industry standard tests which give you a technical evaluation of how accurate the models are.
2. A team of lawyers use the system day in and day out and verify the information retrieved by the AI, and
3. We do multiple test runs with cases and get feedback from the end users, and there is a mechanism on the platform where the user can point out inaccuracies and suggest amends so the system auto-corrects itself. These are ways we ensure the accuracy is high."
Speaking about further developing products that are already available to the stakeholders, Trivedi said,
“There are 2 things - One is a routine meeting with members of the Committee and the stakeholders to whom we have delivered our products. There is also a continuous feedback loop where we understand the various emerging roadblocks or if anything else is required. At the same time, we are developing JUDi, which has additional features that came out much after delivering these projects. So there is a constant development of our products to support the workflow of our end users."
Additionally, the team at MCIL conducts training sessions with law clerks and law researchers of each chamber much before the launch of a platform, to familiarize them with the functions and usage know-how. There is also a grievance redressal mechanism in place to resolve any queries that a user may have. The team also conducts one-on-one sessions with each judge’s chamber to understand the unique challenges faced by them and suggests ways in which a particular bottleneck can be circumvented.
“We are currently in talks with other High Courts as well for deploying similar AI-powered solutions to meet the challenges faced by each High Court, and they are looking forward to procure the system. In fact, we conducted a conference on ‘AI and Judiciary’ where over 6,000 members from the judiciary were present including all members of the AI Committee of each High Court. However, since each High Court functions in a different context, there are different requirements of what the AI needs to do, but the underlying infrastructure for all courts is ready,” he said.
On a lighter note, when asked about the AI technology replacing law clerks and researchers, Trivedi said,
“Honestly, I don’t foresee that happening in the next 5-10 years, practically speaking. Even when we get to that stage, AI is a machine after all. It requires human supervision. Especially the system we have developed tries to capture the digital footprint of an AI user. If there is no human user, the system may become redundant given that the law is an ever-evolving field and a reflection of fast-changing society.
This system only tries to become a digital paralegal to reduce the workload of juniors or law clerks in mundane administrative and clerical tasks like compiling documents, extracting information, etc. Initially it tries to follow the steps of the paralegal and eventually imitates it. So tomorrow, each judge may have the strength equivalent to 50-100 paralegals which can expedite the process of preparing for matters. This is where the system will try to supplement and facilitate the function.”