Can artificial intelligence help reform Indian courts?

Source: Hindustan Times dated 29.11.2019

-- Ameen Jauhar (Senior Resident Fellow working on judicial reforms at Vidhi Centre for Legal Policy)

AI can aid judicial reform. But do remain cautious about abuse of big data sets, and biases in technology.

Minority Report was a classic Steven Spielberg sci-fi film. Employing tech-noir, the film exhibited a dystopian plot showcasing the dire pitfalls and consequences of predictive law enforcement. The movie conceived a futuristic technology, mixing psychics and premonitions, to pre-empt crime, with a suspect apprehended using a special department labelled, quite literally, “PreCrime”. Similar themes surrounding the deployment of intelligent machines to aid in law enforcement and criminal justice, which in turn go awry, have consistently featured in popular culture. These seemingly grandiose notions of artificial intelligence (AI) are rapidly finding themselves at play in real life.

The new Chief Justice of India (CJI), SA Bobde, has, in no uncertain terms, expressed the judiciary’s desire not to miss the AI bus. He has talked about the Supreme Court (SC)’s internal group of experts, currently working to determine avenues for deploying AI in our courts. Though the technology is novel, the focus is on the age-old, intergenerational crisis of excessive judicial backlog. With AI, the judiciary wants to experiment and deploy a transformative technology, which supersedes all its earlier efforts of integrating information and communication technologies in courts, under the e-courts project.

AI can potentially create unique technological applications to exponentially improve the efficiency of judges and lawyers. These interventions will go beyond merely identifying case status updates or finding the appropriate orders and judgments. The automation conceived through AI is far more sophisticated, allowing, for instance, judges to determine precise answers to their queries in a bail application, using a case-query tool. Rapid progression in machine learning and natural language processing techniques have opened the floodgates for newer tools.

While all this sounds fascinating, and gives life to hitherto fictional plots, there are considerable challenges underlying the deployment of AI-driven technology in Indian courts. For the judiciary, there are two immediate concerns, which warrant more deliberation and concrete governance frameworks.

First, there is the question regarding the collection and utilisation of big data. Machine learning inherently operates with large data-sets serving as fuel for the engine, informing the algorithm of the various correlations, patterns, and analyses of extensive and meticulous data sets. While the SC recognised the individual’s right to privacy in its landmark Puttaswamy judgment, the contours of this are amorphous. The vacuum of a statutory framework for data collection and protection renders the use of copious data sets susceptible to abuse.

The second issue, needing a finer and more pragmatic appraisal, is the presumed unbiased nature of an AI-driven tool. Among many exponents, there is a seeming consensus that AI in courtrooms can dispel the biases of judges. The problem with this reductionist conclusion is factual inaccuracy. For all its technological superiority, AI today is indeed afflicted by biases. An official report from the Obama presidency’s archives identified different types of biases, from the use of inaccurate, incomplete, or antiquated data sets, to the personal bias of programmers designing AI-driven tools, seeping into the final product. AI is, thus, vulnerable to existing biases, and given the extent of its usage, it has the capacity to perpetuate these systemically, if it is deployed in an unregulated manner.

The question then remains whether AI should be avoided in the courts or are there steps that can ensure the maximisation ofits potential, while minimising the detrimental fallouts. The answer must always be the latter.

To optimise the utility of AI in the Indian judiciary, there are four broad steps that must be taken. One, undoubtedly, the first-generation AI tools, as have been highlighted by the CJI in his media interviews, will prove to be a watershed. However, it is imperative that the process is iterative and incremental, yielding more sophisticated and diversified AI-driven technology for the Indian judiciary.

Two, to undertake this steady expansion, it is critical to facilitate the requisite user feedback through appropriate channels. These feedback loops must be supplemented with periodic, impact evaluation studies.

Three, as the AI industry is rapidly evolving, so must policies and governance frameworks for this technology to remain effective. Given its innate complexity, the use of AI must certainly be experimented only through an evidence-based and research-driven approach, and not through experiential intuition.

Four, a long-term transition into an AI-driven justice system requires all stakeholders to have a firm grasp over its technologies. Therefore, streamlining of training modules and workshops for evolving more sophisticated AI, must be effectuated simultaneous to its deployment in the courts. AI still remains uncharted territory. These steps will ensure its seamless integration and allow future generations to build on this edifice in the coming years.