Using artificial intelligence in governance and government initiatives

Multiple governments today wish to use AI to make their processes more efficient. This could be a game changer specifically for developing economies which suffer greatly from problems of lack of capacity, infrastructure and corruption. Easier said than done, governments really struggle on how to find the right projects where AI could be super effective, how to build solutions and finally implement it. This is a short policy note on doing this in an efficient manner. It may take 5-7 years to implement these programs, but the methodology with yield real long term benefit. The policy note is written in context of India, but is applicable to other countries as well.

Stage 1: Identify the right questions to apply AI and develop their solutions (2-3 years)

Everyone wants to apply AI in their processes, businesses and administration, but they do not know on what problem to address and how. It is easy to use commoditized technologies such as speech recognition or image tagging. However, identifying problems in one’s process and applying AI on one’s own data to build solutions is non-trivial. Many trials fail!

We need to involve the research community and practitioners to frame and run multiple experiments that can help  identify and develop solutions for the right projects. The way to do it is as follows:

-          Roll out call-for-proposals for AI-related projects through Department of Science and Technology (DST):
o   Hire AI Program Managers in DST. Get people who have experience in applied AI research
o   Launch generic AI RFPs in different domains
o   Program Managers need to actively interface with people in the different domains to understand the problems in great depth
o   Announce problem specific RFPs in different domains based on problem understanding
-          Research groups to co-bid for projects together with public/private organizations in the domain.
-          Allow private parties in software development/data science/data engineering to participate in some projects.
-          Allow researchers from foreign universities to collaborate on some projects. Funds to be paid to foreign universities can be procured from alternate sources.
-          Make it mandatory that all data used in these projects should be added to open-source projects.
-          All code generated should be open-source.

This will create a number of different experiments, create data-sets and solutions. It will also naturally spur entrepreneurs to take up some of these ideas and build companies out of them.

Stage 2: Identify high impact projects (1-2 years)

In stage 2, the government should identify which of these projects could have the highest impact. Various government departments can examine the solutions developed and develop  hypotheses on projects could be  as high-impact. At this stage, the government may involve public policy researchers to do an impact evaluation of selected technologies. They will validate the usefulness of the technology and also help in streamlining processes. In this way, the most promising technologies may be identified and taken up for implementation.

Stage 3: Implement solutions (2 years)

In this stage, the government should implement the identified projects for better governance, in government initiatives and public programs. The government should classify projects in two buckets and have implementation strategies accordingly:

-          The government should identify those projects that can become part of the government system and where the government doesn’t have to get into selling or drive adoption. For instance, a AI technology which could make the driving test and licensing process more effective. This can be directly integrated into government IT systems. The government should implement these projects through third parties using a tender process. Various IT solution companies with AI practice may bid for them. Here the projects should be IT services projects which become part of the current government systems.
-          Projects that require selling/adoption should be left to the entrepreneurs. For instance, AI technology that may help farmers understand soil or seed quality. Governments do not do the best in scaling adoption of such offerings – private businesses on the other hand do well at it.The government has already created viability for these projects through Step 1 and Step 2. It could further encourage entrepreneurs by providing competitive funding or subsidies or by becoming the first customers for such products/services.

First published: 14th April 2018