In the past decade, there has been considerable discussion around the design, development and deployment of artificial intelligence (AI). For instance, in India, the NITI Aayog recently published an approach paper on the need to harness AI in a responsible and ethical manner. The Indian judiciary, which has already created basic information and communication technology infrastructure under the eCourts Project, is now looking to leverage AI’s potential as well.
Given that the foundations of an ecosystem are currently being laid, it is desirable that the Supreme Court should ideally steer the deliberations over such a roadmap. Without such oversight, its development might be haphazard and uncoordinated. Recognising this objective, and to aid the judiciary’s efforts in this pursuit, this section enumerates a broad agenda under three broad categories namely, conceptualizing the integration of AI in the justice system; setting up operational support to enable such integration; and deploying AI in the justice system.
Conceptualising the integration of AI in the justice system
The foremost exercise that must be initiated by the AI committee of the Supreme Court is to determine the short, medium and long-term usage of AI. This exercise must establish clear ground ethical rules for the responsible design and deployment of AI. In parallel, it must also ascertain the logistical capacity of the judiciary to integrate such technologies (for instance, how can judicial data be archived and made more openly accessible). The conceptualization stage must include, inter alia:
a. Adopt a governing charter establishing key principles to safeguard due process, constitutional and legal rights, and address concerns of transparency, bias, and lack of accountability in AI driven technologies. For instance, the Institute of Electrical and Electronic Engineers (IEEE) has come up with Four Principles for the Trustworthy Adoption of AI in Legal Systems. Even the European Commission For the Efficiency of Justice (CEPEJ) has come up with an ethical Charter on the use of Artificial Intelligence in judicial systems.
b. Undertake extensive stakeholder consultation to ensure transparency and confidence of the legal fraternity in this entire process. It is pertinent that lessons from the COVID-19 pandemic guide future digitization endeavours of the judiciary, and the justice system at large. Crucially, the buy-in of different stakeholders is vital and inputs must be solicited on the scope, potential uses, and challenges that will be faced in this process of AI integration. The AI Committee, in its expanded form as mentioned below, can be responsible for establishing these channels of engagement. The first document for consultation can be the AI Charter.
c. Promote research on governance of AI for the justice system, to address critical challenges that AI presents. There is a need for high quality, interdisciplinary research evidence to facilitate better informed policy decisions on the integration and use of such emerging technologies. The AI Committee can incentivise and commission independent research studies in this regard, with a focus on the Indian context, and grassroot realities. While there is some international literature that exists on similar questions, India’s own social and cultural settings are subjective, requiring indigenous research inputs.
d. Plan for capacity building through adequate training and skill development. A key deficiency that typically impedes scalability of automation is the absence of accompanying capacity building. Lack of adequate training can also foster suspicions about the tech intervention, and erode confidence in it even before its implementation. AI for all its sophistication, is advanced technology. In a country like India, it is crucial to not merely train the users of such technology, but also the recipients. This multi-stakeholder educational undertaking will be complex and would require adequate planning which must be executed prior to actually deploying such technologies.
Setting up operational support to enable integration
The actual development of AI-driven technologies for the justice system will require careful planning. Particularly, there are three main action points for the judiciary to initiate:
a. Expand the Supreme Court’s AI Committee to oversee the integration of AI within the justice system. This expanded AI Committee should ideally consist of a core-group of sitting or retired judges of the Supreme Court or High Courts. It should be headed as it currently is, by a sitting Supreme Court Judge. Critically, expert members like technologists (with specialisation in AI design and innovation), ethicists, policy researchers and academics, should be inducted on a professional basis. Though some such members are a part of the committee today, they are neither multi-dimensional nor fully engaged. The rules of their engagement, including eligibility, remuneration, tenure, number of expert members etc., should be determined by the core group of judges. The objective is to have a permanent entity that is working singularly on the task of this AI integration, in lieu of a body that periodically meets to discuss this agenda.
b. Publish openly accessible datasets as they are sine qua non for any meaningful AI innovation. As discussed earlier in this paper, presently, both the judiciary and other wings of the justice system collate and archive a huge amount of data. However, there are no overarching protocols which dictate the sharing and usage of these. The AI committee can consider a data trust created by the Supreme Court for the collection, storage, and sharing of non-sensitive judicial data which can aid innovation of ML (machine learning) algorithms, while safeguarding public interest. The data trusts should subscribe to the data protection laws that have and will be introduced in India. In order to establish such a trust, the core group of the AI committee will need to decide the institutional framework including the potential board of trustees and the terms of licence for use of its datasets. The data trust can be set up in different institutional formats, say as a non-profit company. Incorporation must be determined on which institution is ideally suited to perform their core functions while maintaining a fiduciary responsibility towards public interest.
In addition, the judiciary should consider evolving an open data policy regarding access to its data, in a manner which protects personal privacy, while allowing technologists to harness the potential of existing datasets. The creation of viable open datasets requires the judiciary to find trustworthy partners that can facilitate this process. This should be done by the e-Courts Committee of the Supreme Court in consultation with the AI Committee.
c. Harness public private partnerships (PPPs) to design and deploy AI tech interventions. A PPP model, which engages private social corporations, can ensure oversight of the State, while tapping into the resources and expertise of the private sector. Until now, the NIC has been the foremost institution responsible for the overall digitisation of public sector entities in India, over the past two decades. However, design and development of public-centric AI, as in the case of developing AI for the justice system, will be a costly and logistically complex endeavour. It is common knowledge that ecosystems like these are not created proprietorially—they need a coordinated public-private partnership to build a platform on top of which innovators can create various products. Expressions of interest from suitable private players to build the AI use cases listed in Section I of this paper should be sought.
Phase-wise deployment of AI in the Indian justice system
To accomplish the twin-pronged objectives of improving administrative efficiency and decision-making processes, the actual deployment of AI must be done in a phased manner. This will include piloting AI interventions, reviewing their progress, and building on these first-generation technologies in an iterative manner. There needs to be a planned and incremental expansion of the deployment of AI, instead of its haphazard and piecemeal usage. Key points that need to be implemented for this phased design and deployment are as follows:
a. The first generation of AI pilots are already in progress with both SUVAAS and SUPACE (an AI tool to assist judges in legal research) being tested in different courts under the auspices of the AI committee. Further, AI innovation can build on this to create more ML and deep learning algorithms for greater process automation. The focus of the first-gen AI tech interventions can be on improving administrative efficiency, as has been discussed before in section I of this paper. This would crucially require identification of the different administrative processes in the entire life-cycle of a case, and prioritising which of these can either be completely automated (like scheduling hearings), or which are proving to be bottlenecks (like issuing summons). Once identified, such proceedings can benefit from the use of narrow, ML algorithms to automate, streamline and make basic court processes more efficient.
b. Creating feedback loops and impact evaluation frameworks to understand what works and what does not. Vital in improving the future generations of AI for the justice system, is to ensure that pilots and first-generation AI interventions are independently evaluated. The impact evaluation, while being conducted under the auspices of the AI committee, must be undertaken by expert technology auditors. Furthermore, such scrutiny must be periodic and not a one-time occurrence, to ensure that the quality of technology deployed, meets the highest standards, and conforms with ethical best practices as they emerge and evolve within the broader discourse on governance of AI.
c. The second generation of AI tools can target decision making processes by inducting more sophisticated AI technologies like case query tools, intelligent analytics, research augmentation, computational tools directly aiding in judicial decisions, and legal robotics. As discussed earlier in this paper, improving decision making is a key objective behind the ongoing interest in use of AI. For judges, intelligent case query tools, or algorithms that collate and analyse case-law and theoretical legal research, can be an immense aid. These tools would not only expedite judicial decisions but arguably, also ensure a comprehensive review of existing jurisprudence which is more time consuming and laborious as a manual process. In addition to lawyers and judges, this phase of AI development will also directly benefit litigants by providing them easier access to basic legal services, and an interactive repository of rudimentary legal information. It is pertinent to mention here that the second generation of AI innovation for the justice system will benefit from the constantly evolving field of AI, and may arguably even rely on more advanced techniques than ML.
This phase will have two parallel programmes—one, to aid judges and lawyers, and two for the general public and potential litigants. As also discussed earlier, some of these “narrower” computational tools will also aid in the development of more complex and accurate tools for aiding predictive justice. This kind of integration of advanced artificial judicial intelligence, will be pivotal in transforming the justice system in India. Within the judiciary itself, the easier use cases could focus on task specific ML algorithms through pilot programmes aiming to determine their efficacy, ease of usage, and potential shortcomings which need to be addressed before scaling or branching into other use cases.
While attaining the twin objectives can commence the Indian justice system’s journey with AI, the potential that AI possesses to bring about change is transformative. In the medium to long term (3 years), it will be necessary for the AI committee to revisit this roadmap and identify a renewed set of next steps that cater to and are responsive to the evolving needs of India’s justice system. At its core however the development of the AI ecosystem in India must remain open to all innovators in the Indian start-up ecosystem based on development of a set of standards customised to the Indian justice system.
This is part-2 of a two-part series on AI and judiciary wherein excerpts from the paper ‘Responsible Artificial Intelligence for the Indian Justice System’ are published with the permission of Vidhi Centre for Legal Policy. You can read the entire paper here.
Disclaimer:Ameen Jauhar, is a Senior Resident Fellow and Team Lead at the Centre for Applied Law and Technology (ALTR) at Vidhi. Vaidehi Misra, is a Senior Resident Fellow at the Justice, Access and Lowering Delays in India (JALDI) initiative at Vidhi. Dr. Arghya Sengupta, is the Founder and Research Director at the Vidhi Centre for Legal Policy. Dr. Partha P. Chakrabarti is a Professor at the Department of Computer Science and Engineering (DSE) and a Joint Faculty Member of the Centre of Excellence in Artificial Intelligence at the Indian Institute of Technology, Kharagpur (IIT-K). Dr. Saptarishi Ghosh is an Assistant Professor at the DSE at IIT-K. Dr. Kripabandhu Ghosh is an Assistant Professor at the Department of Computational and Data Sciences (CDS) at the Indian Institute of Science Education and Research (IISER), Kolkata. Views expressed are personal.
Read all the Latest News and Breaking News here