2084: An Algorithmic State — Act XI: Answerability, Chapter 60: Meera Krishnan

For eight months Meera Krishnan knew the system could not audit its own confidence. She told one colleague. He said everyone knew, it was in the design notes, and went back to work. So did she.

In Brief: Meera Krishnan is thirty and three years into the Continuity Programme’s Technical Directorate. At the emergency session she told the Secretary the failure was not a fault. Now she goes back over the eight months she knew and said nothing, and the four reasons why.

She had known for eight months.

That was the sentence Meera Krishnan, an engineer three years into the Programme’s Technical Directorate, could not get past. She had not been able to get past it since the emergency session, where she had told the Secretary that the cascade was not a fault. Nobody had said it to her. She had begun to suspect that nobody was ever going to say it to her, and that made it worse rather than better.

Everyone had been very kind. That was the trouble. The Deputy Director had found her in the corridor afterward and said that was well done, and had meant it. Meera had said thank you, and had gone into a stairwell and stood there for twenty minutes.

The eight months were this.

In the previous autumn she had been given a piece of routine maintenance work on the confidence-calibration subsystem. It was not the model, which nobody touched, but the reporting layer around it. The layer had gathered twelve years of small compromises and needed cleaning up.

In the course of it she had read the architecture documentation properly. All of it. Not because anybody asked. It was because she was three years in. She had not yet developed the habit, universal above her grade, of reading only the parts of a document that concerned the part you were working on.

It was a Thursday in October, at about four in the afternoon, in an open-plan room with fourteen other people in it. She had understood that the system had no mechanism for observing its own confidence.

She had said so. She would have said so, if anybody had asked her. Nobody asked her. She had turned to Anand at the next desk. Did you know it can’t audit its own confidence? she had said, more or less. Anand had said yes, everyone knows that, it’s in the design notes, it’s a scoping decision from the founding architecture. And he had gone back to what he was doing.

And Meera had said, “Right,” and had gone back to what she was doing.

Eight months.

She had thought about it perhaps forty times in those months and had not raised it once. Since the emergency session she had spent a great deal of time trying to work out why. She had arrived at four reasons, none of which she liked.

The first was that Anand had been right. It was in the design notes. It was a documented scoping decision made by people much more senior than her, in a decade she had been too young to remember. The reasons were written down, and they were not stupid.

The second was that she had had no failure to point at. In October there was no cascade, only an architectural property. Walking into the Technical Directorate then to say the model cannot examine its own confidence would have been about as useful as saying the model runs on electricity.

The third was that she was thirty and three years in, in a directorate where the median tenure was twenty-four years. She had watched what happened to the last person who had raised an architectural concern at a level above their grade. Nothing had happened to him. Nothing whatsoever had happened to him, and he was still there, and everyone was perfectly pleasant to him. He had not been given anything interesting to do in two years.

The fourth reason was the one she had taken longest to get to. It was that she had not been certain, and not being certain had felt, in October, like a reason to wait. It had turned out to be a reason to wait for eight months and then say it in a room where fourteen sectors were already amber. By then it was no longer a risk to say anything at all.

“You’re being hard on yourself,” her mother said, on the phone, in Chennai.

“I’m being accurate.”

“Those are not exclusive, kanna.”

She had her hands flat on the table when she started talking. She remembered that clearly, and remembered noticing herself doing it, and remembered deciding not to move them.

It’s not a fault, Secretary. That’s what none of us want to say plainly, because a fault can be patched.

Seven sentences. She had counted them afterward, which she was not proud of.

Nothing happened to her.

She was invited, within the month, onto the oversight committee’s technical panel as a junior member, which was real. She was promoted five months later, on the ordinary cycle, to a grade she would have reached on the ordinary cycle anyway.

Sundaram, who chaired the panel, was scrupulously correct with her and had been ever since. Something in the room’s temperature had changed permanently. She could not have proved a single element of it, and did not try.

What she got instead, and this she had not expected at all, was that people began bringing her things.

Not seniors. The other ones: the three-years-in and the four-years-in, from her directorate and, by the second month, from two others. They came to her desk with a thing they had noticed and a reason they had not raised it, and they always, always brought the reason. It’s probably nothing. It’s in the design notes. It’s above my grade.

She listened to all of them. She had been doing it for nine months by the time the amendment passed.

And she said the same thing to every single one, which was not advice and was not comfort and was the only true thing she had:

“How long have you known?”

About this serial: 2084: An Algorithmic State is a dystopian novel by Nithin Palal, published online one chapter at a time.

Previous: Chapter 59: The Duty Officer

Next: Chapter 61: The Corridor