The Learning State: How information becomes insight

The vision is that these systems will enable citizens, government officials, and politicians to gain access to all the information that they need to play their individual roles in a democratic society and enable the necessary exchange of knowledge for effective program implementation. Those managing food subsidies can monitor the movement of grains via geo-tagged trucks, while the MGNREGS MIS informs administrators of payment delays, and citizens use online grievance portals to register complaints.

Such information has the potential to be extremely valuable. One reason why these investments are happening now is that the cost of collecting and sharing information has gone down enormously over the last decades, thanks to the IT revolution. It is possible now for a mother at work to watch her toddler at play in a playschool by connecting her cellphone wirelessly to a CCTV at the school through one of multiple available apps. And she might for the first couple of days, but the novelty wears off fast. And then? The app will sit unused, unless there is a specific concern (say, the child is sick).

The problem is that information by itself is not insight; it has the potential for insight. Knowledge is not gathering mounds of information. It is processing that information and translating it into useable propositions — for example, “my child really looks tired, perhaps I should plan to leave early” — that makes people and organisations learners. The mother stops watching the CCTV feed because she needs to focus her mind on other tasks — she already has a tough time fending off the thousand other distractions that life throws at all of us. Processing the volume of information that a CCTV can generate every minute can clog the mind so fast that she needs to take defensive action.

Thus, so it is for organisations. It is not enough to generate information — someone has to process and translate data into something useable, and the more information you generate the harder that necessarily becomes — the more to sort through and discard. And if you have no guidance on how to do that sorting, then more information may actually hurt. Consistent with this, process evaluations suggest that in the wake of the MIS expansion, the local bureaucracy at the district and state levels is drowning in MIS data, with neither the capacity nor the inclination to process it.

With the growing complexity and sophistication of India’s social protection systems and schemes, building the capability of the local state to learn and reflect on information is critical. To build a genuine learning state — a state where everyone, citizens, bureaucrats and politicians use the information they need to generate insight and hold each other accountable, we need three building blocks.

* First, we need to curate the information that gets highlighted. That does not mean censorship — citizens should be able to access any information that they could reasonably need — but some filters have to go into choosing what to give prominence and why, based on a clear theory of how the information would be used, by whom and why that would matter enough to deserve the priority.

This might seem obvious, but both of us have been witness to the many extremely far-fetched theories of change that get used to justify MIS interventions; in particular, the vision of a citizen who gets up in the morning and quickly files a MGNREGS problem report, then tweets about the policeman she observed taking a bribe on her way to work, spends her afternoon responding to a call from a government call-centre about PDS delivery and grain quality, followed by an evening reviewing the accounts of the municipality before turning in (to dream of accountability we’re certain), is far-fetched. Even the most committed citizens have many other life problems to deal with, and activism is at best a part-time activity.

* Second, we need to have clear designation of how the information will be used to provide incentives to actors within the system — if the news is bad, responsibility for it has to be clear. The MLA report cards developed by Satark Nagarik Sangathan and published in various newspapers before state elections, are a good example of prioritisation (a small set of numbers, prominently displayed) and clear designation — it says, more or less explicitly, “you are about to vote for your MLA, here is what he did”.

Research shows that this intervention changes how the voters vote, rewarding the best performers according to the report card, and hurting the worst. This example also makes clear that whenever we prioritise information there are hard choices to be made. The report cards, to be effective, left many things out and an MLA could potentially complain that this was unfair to her achievements. There is no perfect resolution here, but there is a clear trade-off between effectiveness and being comprehensive.

* Third, it is important to test whether the information is doing its job, whether it is being used to provide the required intelligence and insight. Even if the information collected is salient, it may lead to no change in behaviour or local action. For example, the government of Karnataka built an ambitious MIS to biometrically track real-time attendance of nurses at health centres. The pilot hoped to hold front line staff accountable by making their attendance transparent. However, the reform made limited long-run impact as state officials, local-level bureaucrats, and locally elected bodies were reluctant to use the better-quality attendance data for enforcement due to a fear of generating discord among the staff.

To summarise, learning is not a logistical task. We have too many examples now showing us that an MIS is not magic. The mere existence of call centres and information infrastructure does not mean much without effective use and design. Even with the best technology and information monitoring, local administrators may choose not to convert the data they produce and own into knowledge and a body of actionable evidence, unless they have the people, time, and training to do so. Without such resources in the design and use of MIS for social programs, information will become relentless noise, a nuisance rather than an opportunity.

Abhijit Banerjee is Ford Foundation International Professor of Economics at MIT and Director of the Abdul Latif Jameel Poverty Action Lab (J-PAL). Shrayana Bhattacharya is Senior Social Protection Economist, The World Bank.