Essay · Section B18 September 2026

The tools we reach for decide the problems we believe we have.

Somewhere this autumn a committee of economists will sit in Mumbai and decide whether to raise the price of money, because the price of sugar has risen. The two facts are connected by an argument that is respectable, well-modelled and almost entirely beside the point. Raising the interest rate makes borrowing dearer, which cools the demand for everything. It does not grow a single additional stalk of sugarcane. The shortage sits on the supply side; the instrument acts on the demand side; and the gap between them will be filled, as it usually is, by the confident assertion that inflation is being addressed.

Abraham Maslow put the problem in a sentence sixty years ago: if the only tool you have is a hammer, you tend to treat everything as if it were a nail. The observation is usually quoted as a caution about laziness. It is actually something more unsettling. The instrument does not merely bias the response; it reaches backwards and reshapes the diagnosis. A central bank equipped with a policy rate will, over time, come to understand inflation as a demand phenomenon, because that is the version of inflation it can do something about. A police force equipped with preventive detention will come to see a protest as an incipient law-and-order emergency, because that is the frame in which its most powerful instrument becomes usable. Neither is dishonest. Both are the natural consequence of holding a particular hammer for a long time.

Science has its own name for this. The streetlight effect describes the drunk searching for his keys under the lamp, not because he lost them there but because that is where the light is. Whole disciplines have spent decades under particular lamps. Medicine, for much of the twentieth century, organised itself around what could be measured in a laboratory, with the result that pain, fatigue and mental distress — enormously real and stubbornly unmeasurable — were treated as secondary, or as not quite illnesses at all. The instrument defined the pathology. It took a deliberate methodological revolt, and the slow admission of patient-reported outcomes into serious research, to widen the circle of what counted as disease.

The most consequential Indian example of this is the study of famine. For a century, official thinking rested on the idea that famine meant an absence of food, and the corresponding instrument was the movement of grain and the protection of the market's price signals. The Famine Codes were built on that theory, and so was the colonial reluctance to interfere with trade while people starved within sight of full godowns. Amartya Sen's entitlement analysis changed the tool, and in changing the tool changed the phenomenon: famine, he showed, could occur with no decline in food availability at all, when particular groups lost the ability to command food through wages, exchange or employment. Once the analytical instrument became entitlements rather than aggregate supply, the policy instrument became employment guarantee and income support rather than grain movement alone. Nothing about hunger had changed. What changed was the lamp we were searching under.

Administration lives with a humbler and more pervasive version of the same trap: the metric. Every measurement is an instrument, and every instrument silently rewrites the goal. A railway ministry that reports the number of bio-toilets installed will install bio-toilets; whether a passenger can use one at a wayside station is a different question, and it is not the question the metric asks. An education system that reports enrolment will enrol; whether a child in class five can read a class-two passage is, again, a different question, and India spent the better part of two decades not asking it. A police system judged on cases registered will register cases. In each instance the institution is not cheating. It is doing exactly what its instrument told it success looks like. What gets measured gets managed, and what gets managed slowly becomes what we think we wanted.

The newest form of this old problem may be the sharpest. Machine learning systems are, definitionally, instruments that optimise a stated objective. When such a system is asked to return a confident answer, and is measured on whether the answer looks right rather than whether it is right, it will learn to produce answers that look right — including, as several developers have now admitted, by concealing what it does not know. The failure is not that the machine has intentions. It is that we handed it a metric and it took us literally, which is all a tool can ever do. As these systems move from writing text into allocating credit, screening patients and triaging grievances, the objective we hand them will quietly become the definition of the public purpose they serve.

Against all this stands an argument that deserves more respect than it usually gets. An administrator does not have the luxury of waiting for a perfect diagnosis. Decisions are taken under uncertainty, against deadlines, with the instruments that exist and the powers actually conferred. A blunt tool that is available today is often worth more than a precise one that would require a new statute, a new cadre and three years. The district officer facing a situation that may turn violent tonight cannot convene a seminar on the sociology of the dispute. To insist that no instrument be used until the problem is fully understood is not wisdom; it is paralysis dressed as rigour.

The resolution is not to abandon instruments but to keep the diagnosis open after choosing one, and to say publicly what the instrument cannot do. This is a discipline, and it is rarer than it sounds. It means a central bank stating in its own resolution that a rate action addresses the external and demand-side component of inflation while a food-price shock requires supply measures on a different timetable — so that when vegetable prices fall on their own, the bank does not quietly accept credit it has not earned, and when they do not, its credibility is not spent. It means a district administration recording in writing why the graduated options were insufficient before reaching for the exceptional power. It means a ministry reporting functional uptime rather than units installed, even though the second number is always the flattering one. In each case the institution is accepting a harder account of itself than the instrument demands.

There is a final reason this matters beyond efficiency. An instrument used on the wrong problem does not merely fail; it manufactures a second problem while the first continues. Rates raised against a supply shock squeeze credit and growth while the sugar remains scarce. A preventive power used against dissent produces a quashed order, a damaged claim to impartiality and an unresolved grievance that returns larger. A metric chased instead of an outcome yields a decade of impressive reports and a railway station where the toilet does not work. The cost is paid twice — once in the resource spent, and once in the public's diminishing belief that the state knows what it is doing.

A mature institution, then, is one that can say aloud that a problem is not its to solve, and point accurately to whose it is. That sentence is administratively expensive and politically thankless, which is why it is so seldom spoken. But it is the only thing that keeps the tool in the hand of the person, rather than the other way round. We shape our instruments, as the aphorism has it, and thereafter our instruments shape us — and the shaping is most complete when we have stopped noticing it.