Artificial Intelligence (or AI) in governance is an idea whose time has come. The necessity is there – India’s citizens are growing ever more impatient and continuously demand for greater efficiency in the delivery of public services. Traditional systems, on the other hand, are creaking and seem to be unable to cater to these changing times.
Conditions are also ripe for AI in governance – thanks to the use of IT there is a lot of data in the government today in machine-readable form and technologies have reached a level where they can rival any human on a real-time and cost-effective way. All that is needed now, is that leap.
AI would fundamentally transform the way in which governance is done. It is often complained that the implementation of government schemes remains confined to official documents and dusty stacks of paper. What if there is a way to actually check if things are happening on the ground? Take for example, the Swachh Bharat Mission (SBM). To make sure that the toilets are actually built, the government has built a mobile app where a government functionary will have to go to the toilet site, click a photo of the toilet and the beneficiary and upload it to a central server. Connectivity issues are taken care of by giving offline photo-clicking mode and uploading the photos when the person comes back in a 2G network. This curbs malpractices to a great extent.
However, if there are 10 crore toilets to be built under the SBM, this would result in 10 crore photographs. If the government is presented with 10 crore photographs by its employees and local contractors, it is almost impossible to check whether each and everyone of these photos show a complete toilet or half-built toilet. If the toilet is in use or is stashed with hay? Or even if it is a toilet or a bird? And is the same beneficiary appearing on multiple photos or if 100 photos have been uploaded from a government functionary who is sitting in her office? Clearly, this task would be impossible.
So, presently, our system relies on people to do random checks in order to create a deterrence effect. This is what the Indian system has relied upon for the past 70 years and its outcome is clear for all to see today. Government work is never-ending and unfortunately there is a lack of capacity within the government.
Now, coming back to our toilet example, what if there was a way to process each of these 10 crore photos and generate an alert whenever the photograph is not that of a completely built toilet which is actually in use (not stashed with hay or other stuff) and the same beneficiary doesn’t appear in multiple photos or multiple photographs don’t get uploaded by a government functionary sitting in her office? Won’t cheating and malpractices go down by an order of magnitude as people realise that each photo would be scrutinised and not just a small sample? More than that, wouldn’t it be a relief to know we actually have 10 crore functional toilets on the ground and not just on paper? This is possible with AI and machine learning – and it is possible to do this in a very cost effective, very accurate way.
Sceptics, however, argue that in rural and remote India, Internet penetration is very low and as a result AI will have limited or no applicability there and will actually create a digital divide. However, contrary to this, the need and applicability for AI is more urgent in the remotest areas of the country than in the heart of the capital. Because it is in these remotest areas that traditional governance systems are totally broken. Physical infrastructure in rural areas is poor. Generally no one wants a posting there – most people there would be on punishment postings and as soon as they come, they would start spending their energies in getting a transfer back to the mainstream areas. As a result there are problems of severe under-staffing, lack of morale and poor quality in the government workforce and the monitoring of government schemes and implementation suffers. In the SBM example, in Delhi and state capitals, there would be lot of people to check if toilets have been built; we don’t really need AI there. But who will check in the tribal areas of Rajasthan or Chattisgarh? Imagine if in these areas, the government schemes start functioning as they were supposed to do! AI will bridge the development and digital divide, not accentuate it.
Likewise, again contrary to what the sceptics say, the scope of AI is immense in traditional sectors such as agriculture. For example, consider the government run crop insurance scheme. In this crop insurance scheme, if the yield is below a threshold, it would trigger an insurance payout to the farmer. To determine the actual yield, millions of crop cutting experiments would be carried out – many more than what are mandated today. As per the scheme guidelines only, even the ones done today “lack reliability, accuracy and speed”. So mobile app solutions could be developed where geo-tagged photographs of the crop cutting experiment would be uploaded like SBM. Again, wouldn’t it be wonderful if we have a solution to check these millions of photographs to see whether an actual crop cutting experiment has been carried out by the same person who was supposed to carry it out or has the crop cutting experiment work been sub-contracted to unskilled persons who simply went there and clicked selfies?
On similar lines, the government runs Kisan Call Centers which receive lakhs of calls every month. What if we are able to get timely warning from the call centre data that says in Maharashtra, this year the distress level among farmers is unusually high due to some factor? Perhaps then the administrative machinery can be activated timely on a war scale to prevent farmer suicides? Or say based on soil and environmental condition reports from our satellites and based on what crop is sown in a particular area, we are able to predict that this year vulnerability of this crop to this pest is higher and perhaps we can supply additional required pesticide there and send targeted SMS / agronometric advisories to the farmers in that region? All these things have not been taken from some science fiction movie but are very much available, proven and economic technologies. Similarly, there is a kisan suvidha app – the flagship app of the agriculture department – where among other things, a person can upload 3 photos of some disease / pest infected crops and our scientists would tell what the problem is and what the remedies are. However, as with the case with almost anything in our country, the rush is huge, there are thousands of queries and there isn’t enough capacity to answer all the queries manually. As a result many queries go unanswered and people’s faith suffers due to which they would stop using it in future. Again, AI can help here. Even if the farmer himself doesn’t have a smart phone, even in remotest areas of the country, today someone will be having a smart-phone and there would be a 2G connectivity nearby if not within the village. So these are practical solutions, solutions which can work given the huge social capital in our rural society.
Finally a word on another common misconception – that AI will lead to job losses. One bane of our country is systems don’t work here. AI can make them work. It can leapfrog us in terms of development; it can bring immense prosperity to the nation. Government today is over-burdened and there is lack of capacity to do the multitude of tasks it has taken upon itself. AI is an important answer to capacity building.
Human beings have significant capacity to adapt. It would be unwise to compare us to horses, which after the mass production of cars were rendered jobless. Productivity gains nearly always create more and better jobs than the ones which are lost due to them. This is not to say that the jobs argument is not important — only that it will happen 50 years from now, not today. And, if it has to happen, it will happen regardless of whether government uses AI or not. But its massive productivity gains cannot be ignored.
This really is the case for AI in governance: Can we afford to ignore it?
Gaurav Agrawal is part of the Indian Administrative Service’s 2014 batch and studied computer science at IIT Kanpur.