AI’s best use case may actually be in our judicial system
4 min read 23 Mar 2023, 11:28 PM ISTLandlords in Bangalore often ask for obscenely high security deposits and many choose to leave their houses vacant in fear of squatters
PremiumFrom funny answers generated by ChatGPT to beautiful images created by Midjourney, much has been said about the potential of AI over the past few weeks. Its greatest application, however, at least in the Indian context, is not imagery or the promise of faster coding through GitHub co-pilot, but is likely to be what it could do if adopted by the Indian judiciary.
By all measures, India’s mechanisms for dispute resolution and contract enforcement are abysmal. There are a whopping 47 million pending cases, with a million added every year. According to the World Bank’s Doing Business Report 2020, India ranks 136th out of 190 countries in the enforcement of contracts. This is significantly lower than China (46th). Also, the average time taken to enforce a contract in India is nearly four years (1,445 days), more than four times the global average of 358 days.
The first-order problems caused by India’s slow judicial system are somewhat understood, but its second-order impacts are under-appreciated. These include:
Friction in economic transactions: Our lack of effective contract enforcement means a breakdown of transactional trust, and we cope with the risk of a counter-party reneging on a contract in other ways. This either leads to additional cost or decreases the volume and velocity of transactions. An example of this is observed in the property rental market. Landlords in Bangalore often ask for obscenely high security deposits and many choose to leave their houses vacant in fear of squatters (implying a market breakdown).
A vicious cycle that limits state capacity: Economist Karthik Muralidharan summarises this phenomenon in a memorable quote: “In India, the legislative arm of the government makes laws and policy commitments that are beyond the capacity of the executive arm of the government to deliver. This then often leads to the third arm of government (the judiciary) holding the second arm (executive) in contempt of the first arm (legislature)! The time spent by the executive in addressing the judiciary’s attempt to hold them accountable further reduces their time and attention to actually deliver services. In this way, the three arms of the government are each doing their job as they see it, but end up tying the system in knots and further reducing overall capacity to deliver!"
How can AI help solve these problems?
For starters, few domains lend themselves as neatly to the application of AI as the legal one. From petitions and court proceedings to judgements, every piece of information is well-documented for at least 75 years. However, despite this information being available in text, we still lack a clear understanding of the nature of languishing cases.
For example, we don’t know what type of disputes account for most pending cases. We cannot identify what the different categories are, how they’re changing or what their root causes are. This is partly because it would require an analysis of millions of judgements and tens of millions of petitions.
AI can help resolve these challenges.
Analyse and categorize cases: AI can analyse both rulings and filings to identify the major categories in which disputes arise. It can even be used to provide in-depth root-cause analysis for these disputes, which in turn could inform procedural and substantive changes. For example, if it turns out that most disputes are over land and mostly involve compensation, then our dispute resolution mechanism could be changed to include arbitration or a specific ombudsman for such common cases.
Provide a feedback loop: This loop between the judiciary and legislature lies broken. If clear data can be presented on the caseload impact of every new piece of legislation in near real-time, it would provide much-needed information on how to improve the design of a scheme.
There are, of course, significant hurdles to be surmounted before judicial AI becomes a reality. Relevant data must be available in a machine-readable format, and there are several domain and language specific nuances that AI needs to be trained for. But optical character recognition and Indic language translation tools have matured. Researchers, including Namita Wahi at the Centre for Policy Research and the Open Nyai initiative have already applied similar AI-focused methodologies successfully to research projects in the Indian legal domain.
Who knows, the killer co-pilot app in India might not be GitHub for programmers, it might be one built to help judges and clerks in courts improve the speed at which rulings are delivered.
China has already implemented a similar system, Xiao Zhi 3.0 (‘Little Wisdom’), which claims to have helped to cut a judge’s average workload by over a third and saved Chinese citizens 1.7 billion working hours from 2019 to 2021.
A unique confluence of factors at play also fuels hope. AI has captured the popular imagination and some early experiments have already proven their viability in the Indian context. Also, we have a Chief Justice of India with a clear two-year term who is widely expected to implement significant reforms in the judiciary.
Several researchers have found a positive link between judicial pendency and economic growth. Quantifying that relationship is not easy, but an estimate suggests that even a 10% improvement in judicial efficiency could help unlock at least ₹4,000 crore for India’s GDP.
Joseph Sebastian works at Blume Ventures
