Every morning, before I leave for work, I ask a machine whether my shirt matches my trousers.
It is not a question I could once answer for myself. For a person blind since birth, the small rituals that sighted people perform without a thought—checking that a collar is clean, that the colours do not quarrel, that no stain is announcing itself ahead of a review meeting—have always meant borrowing someone else’s eyes: a parent’s, a spouse’s, a colleague’s, a kind stranger’s. You learn to live with the borrowing.
You also learn what it costs, in small surrenders of privacy, to need another person simply in order to know how you look. Now an application on my phone tells me, in its level voice, that the blue shirt does not agree with the green trousers, and that I would do well not to step out dressed like Dobby the house-elf in Harry Potter’s world. The relief is out of proportion to the task. It is the relief of deciding something about myself, by myself.
This is the part of the artificial intelligence story that the productivity debates keep missing. For most people, these tools are a faster horse—they do a little more quickly what could already be done. For a blind person, they are often the difference between doing a thing oneself and not doing it at all. A machine that can look at a photograph and describe it does not so much save me time as hand back a small piece of the world that was previously shut.
The examples multiply once you begin to notice them. A blind student can now have the diagram in her textbook described to her—the graph, the labelled figure, the illustration that carried half the lesson and none of the words, and that students before her, myself included, simply had to do without. She can be in on the meme her classmates are laughing at, fluent at last in a visual language everyone else speaks without effort—which is less about information than about belonging. None of this shows up in any measure of efficiency. All of it decides, quietly, whether a person is folded into the ordinary life around her or left a little to one side of it.
I meet the same capability at my own desk, at a larger scale. Governments run on things a screen reader was never able to read: scanned files, photographed documents, hand-drawn maps, a junior officer’s picture of a half-built road sent up from the field. For much of my career, these were sealed to me unless someone read them aloud. Now I can ask a machine to describe the worksite in the photograph, or to boil a two-hundred-page project report down to the parts that deserve a closer look. It is the same trick that sorts my wardrobe—sight, briefly, on loan—turned to the work of governance instead of the work of a morning.
I try not to mistake the gift for a miracle. The machine is confident even when it is wrong, and I cannot glance at the original to catch it out; the duty of verification that everyone now carries sits a little heavier on someone who cannot see the thing being described. With a shirt, the stakes are low and the bargain is plainly worth it—the worst case is a small sartorial crime.
With a confidential file, I am not permitted to feed into a public tool in the first place, or a Tamil document the model only half-follows, the sum comes out differently. And the tools that promise access are themselves, often enough, delivered through apps and pages a blind user can barely move through—an irony one learns to curse under one’s breath.
Which is exactly where the State should come in, for none of this ought to depend on the luck of which application a person can afford or reach. Accessibility already sits in our law—the Rights of Persons with Disabilities Act, the guidelines for government websites—and is honoured largely in the breach. As governments now hurry to buy and build AI, the moment is unusually good to insist that access be a condition of purchase rather than a patch applied afterwards; that public-facing systems work with a screen reader from the first line of code; that we invest in Indian-language capacity so the tools serve the languages the country actually speaks; and that we build systems safe enough to hold official data, so that the officers who most need them are not the ones forbidden to use them.
The larger prize is easily overlooked. The same features that let me dress myself could let a great many disabled Indians into work and schooling, and daily life that was, until now, quietly out of reach—not as the objects of a welfare scheme, but as participants in the ordinary world. Access has a way of spilling over: the ramp built for the wheelchair helps the traveller with a suitcase, the caption made for the deaf is read by everyone in a noisy room. Build intelligence that includes the person at the edge, and you tend to improve it for everyone in the middle.
I still ask the machine about my shirt each morning. It is a small thing, and I have come to think the small things are the whole of the argument. They never appear in the productivity figures, and they are precisely what decide whether a person moves through the world on their own terms. A State that grasps this will not treat accessible intelligence as a courtesy. It will treat it as the design.
Gokul S | IAS officer of Tamil Nadu cadre
(Views are personal)
