AI in governance — there’s nothing artificial about it

Defence Minister Rajnath Singh launched an Artificial Intelligence (AI)-powered grievance management application in New Delhi on July 15, 2021, developed by Ministry of Defence with the help of IIT-Kanpur. This was the first AI-based system developed to improve grievance redress in the Government in India.

The AI tool has the capability to understand the content of the complaint, and identify repeat complaints or spam automatically. Based on the meaning of the complaint, it can categorise complaints even when key words normally used for such search are not present in the complaint. It enables geographical analysis of complaints in a category including analysis of whether the complaint was adequately addressed or not by the office concerned. While the use of AI in the Government of India is gaining momentum, most of the initiatives taken by state governments — where public interaction is relatively more — have been in the news for all the wrong reasons.

More desirable use of AI

The use of AI solely to put citizens under surveillance, scrutinise their activities, monitor their financial transactions to ensure better tax compliances and, thereby, extract more revenue to the government is a disturbing trend. The less prevalent and more desirable trend is using AI to augment citizen services. One seems to miss the obvious point that decision-making and file disposal are the most crucial functions of any state government department, and the use of AI can streamline the process and make it more efficient and substantially fair.

The government generates a massive amount of paperwork, and file disposal is a complex process that involves several stages. It involves the receipt of the file, its verification, the processing of the file, and its eventual disposal. The process is often time-consuming and requires significant human resources. However, with the implementation of AI, the process can be automated, making it faster, more efficient, and cost-effective. A 2014 study showed that healthcare officials spent 37.5 per cent of their time on documentation and administration, compared to just 9 per cent on substantive work. AI automation of mundane administrative tasks can free up time to build relationships and apply ‘human mind’ to solve problems of the citizens in a more humane manner.

AI and government’s problems

The use of AI in file disposal can be achieved through the implementation of Machine Learning algorithms. Machine Learning (ML) is an application of AI that involves the use of algorithms that can learn from data and improve their performance over time. By applying ML algorithms to the file disposal process, the system can learn from the data and make decisions based on the patterns it identifies.

Once the Acts, rules, regulations, precedents, relevant court orders and all the previous files related to a particular subject are fed to an AI system, it becomes competent enough to replace the section officers, under-secretaries, deputy secretaries, and special secretaries who support a secretary to the government. It may sound far-fetched but a minister willing to spend time and effort in running their office may not even need a secretary!

The only ‘glitch' for the political executive would be that if AI is fed with the right information it would stick to rules and suggest a morally and ethically right decision — always. Perhaps, this is what is deterring many state governments from relying on AI for better governance, and instead focus on peripheral activities that don’t really benefit the public.

Hila Mehr of Harvard Ash Center has identified six types of government problems as appropriate for AI applications: resource allocation, handling large datasets, to augment in the event of shortage of experts, to handle predictable or procedural scenario involving repetitive tasks where inputs/outputs have a binary answer, and handling diverse data — where data takes a variety of forms (such as visual and linguistic) and needs to be summarised regularly.

Instead of focusing on these areas, most State-level AI initiatives are targeted at citizen surveillance, and are vendor driven. Such an approach is understandable in countries such as China or in West Asia, but a thriving democracy like India should be focusing more on initiatives that are citizen-centric.

AI for the better

The liberal democracies across the world are marching ahead in this sector. New York City proposes to work with IBM’s AI platform (Watson) to build a new customer-management system to speed up the time and process of answering questions and complaints about city services on their 311 platform. This is similar to the work in Surrey, British Columbia, where Watson helped power the MySurrey app to quickly answer citizen questions. The app is used to address 65 per cent of questions that already have answers on city websites. Watson, which learns over time, studied over 3,000 documents about 16 city services, and can answer 10,000 questions. In North Carolina, government chatbots (auditory or textual) answer nearly 90 per cent of calls which are just about basic password support, allowing operators to answer more complicated and time-sensitive inquiries.

Japan has gone further with ChatGPT making its debut in the parliament. Kazuma Nakatani, of the Constitutional Democratic Party, asked ChatGPT: ‘What kind of questions would you ask the prime minister if you were a member of the lower house of parliament?’ He then used those responses to form questions for Prime Minister Fumio Kishida during a discussion around a draft amendment related to Covid-19 pandemic policy.

Better judgment

If the executive and the legislature are impacted by AI, can the judiciary be far behind? The use of AI for legal and judicial functions has a massive potential to help free up human resources to focus on higher value-adding tasks as well as bring down costs. DoNotPay, a legal services chatbot, founded by Joshua Browder in 2015 deployed AI in place of an attorney to defend a case of speed-ticket in February. In the United Kingdom, the cost of hiring a lawyer to defend against a speeding ticket can be between £200 and £1,000. With AI-powered drafting and argument generation, very soon, litigants will argue their own case with a slight prompt from a chatbot in their smartphone. It may also be remembered that the accuracy of AI is far superior to human lawyers, and it is only getting better.

The major applications of AI for legal purposes can include tasks such as document automation, contract review, legal research, legal analytics, and litigation prediction. The accuracy and objectivity of AI in arriving at considered fair judgments could potentially beat out corruption in the judiciary. A protocol could be created whereby AI arrives at a judgment based on all the inputs, and the adjudicator must mandatorily record reasons for any deviation from what the AI suggests. The same methodology can be adopted for arbitration and quasi-judicial functions of the government, including tax assessment and issuance of permits and licences. Thus, AI becomes a tool in assisting the adjudicator as well as prevent arbitrary deviation from the norm.

A flawed approach

AI, however, doesn’t have a legal ‘mind’ of its own, and it can only be an assistive tool in matters that require taking cognisance of crime or requiring application of mind while adjudicating. The overenthusiasm in deploying AI for telephone/camera surveillance for crime detection, and using such AI-driven data for imposition of punishment, including fines, is fundamentally flawed as such measures may not stand the scrutiny of law.

Apart from requisite statutory amendments, the AI ‘thought process’ in government will have to undergo regular audit, and be subjected to oversight mechanisms if it has to be deployed in such a manner that it directly impinges the rights of citizens.

Data standardisation

One of the key challenges in deploying AI to assist decision making is the lack of data standardisation. The government generates a massive amount of paperwork, and the data is often stored in various formats and systems. This can make it challenging to implement AI, as the system needs to be trained on standardised data. To address this challenge, the government needs to invest in data standardisation, and make it a priority. Popularising NIC’s (National Informatics Centre) e-office platform and its protocols can be a significant step in this direction.

AI is a complex technology that requires specialised skills and knowledge. The government needs to invest in training its employees and building a team of experts who can implement, train, and maintain the AI system because it follows the principle of ‘garbage-in-garbage-out’. The data fed to the system must be accurate in every sense. The national administrative academies and state training institutes also must understand the paradigm shift.

Prime Minister Narendra Modi directed the setting up of an Artificial Intelligence/ML Lab in the Lal Bahadur Shastri Academy of Administration during the Valedictory Function of 96th Common Foundation Course. Set up in April 2022, it is the only such initiative in any of the training institutes of India.

The use of AI in government is an exciting development, and with the right investments and policies, it has the potential to revolutionise the way governments work. Critics would sarcastically comment that a bit of intelligence in governance is always welcome, even if it is artificial. But the real requirement is common sense, and the ability to ride the AI-tiger, adhering to the constitutional and legal framework, without losing sight of the rights of citizens.

(Prasanth Nair is an author and Special Secretary, Government of Kerala.)

Disclaimer: The views expressed above are the author's own. They do not necessarily reflect the views of DH.