White Noise in the Mind

There is a particular kind of evening where nothing is actually pending for the day at work. Whatever is left can wait till morning. I am gradually retiring for the day - slow tired ride back home, changing into night gear, asking for food, and then with some scrolls on Instagram and Inshorts, I close my eyes to sleep.

As I focus on my breath and rehearse my routine to be conscious of my physical being ( I also say 'returning to present' at times), I realise my mind has a series of chatter. Chatters that come and go and ones that I am not even paying head to. Yet, they keep occurring in my mind.

I needed a right word for this 'chatter'. I happen to recall my MBA lessons on 'white noise'. And there it is - 'Noise'.

What a student of finance learns and then forgets

Anyone who has sat through econometrics, even badly, even only to clear the paper, has met white noise. It comes up early, gets defined in a line or two, and most of us never think about it again.

Basically it is defined as something that cannot be explained or defined quantifiably. For instance, say the monthly sales of a product. You have a belief about what drives it, so you list those things out. Price. Household income. The season. What you spent on advertising. Whether there was a festival that month. Each one becomes a variable, you collect the data, you fit the model, and out comes an equation that says sales are some combination of the things you thought mattered.

Then comes the part that is actually the point of the exercise.

Whatever the model could not explain is left over, and that leftover is called the residual. You don't discard it. You go and look at it, carefully, because the residual is where you find out whether you were honest about your own assumptions.

If there is still a pattern sitting in the residual — if it climbs every March, if a high month tends to be followed by another high month, if the whole thing drifts in one direction across the years — then your model is incomplete. Something real is operating out there and you left it out. So you go back, find the missing variable, and model again.

But when the residual has no pattern in it at all, no memory of what came before, no direction, no rhythm, just hovering around zero and behaving the same way at every frequency, then it is white noise. And white noise is where you stop. Apologies for bringing 'statistics' in spirituality and meditation.

You stop because there is nothing further to decipher out of it. What is sitting there now is the part of the world that carries nothing about itself, and no amount of further effort will make it say something.

That's really the whole idea. White noise isn't the enemy of the econometrics model. It is the sign that the model is done.

The flatness is the problem

There is one property of white noise that I keep coming back to, and it is the one that made me want to write this at all.

It is flat across the spectrum. Equal power at every frequency. Nothing inside it is louder than anything else, and that flatness is exactly why nothing can be recovered from it. There is no peak to lock onto.

Which is an uncomfortably accurate description of that evening. A mind that is working properly is loud about one or two things and fairly quiet about the rest. When everything is equally loud you cannot pick anything up and act on it, so you go around in circles, and the circling itself starts to feel like diligence.

Fischer Black wrote a paper in 1986 that he simply called "Noise." He was writing about markets, and his point was that an enormous amount of trading happens on noise, and that the people trading on noise are quite convinced they are trading on information. That is the sentence to sit with. Nobody knows they are in the noise while they are in it. The noise arrives dressed as analysis. It has the texture of thinking, it uses the same vocabulary, it produces the same furrowed forehead.

Kahneman and his co-authors later took the same word in a different direction — noise as scatter, the unwanted variability in human judgement. Two experienced people looking at the same file and reaching different conclusions. Or the same person reaching a different conclusion on a Tuesday than he would have on a Friday, with nothing to explain the difference except the day.

Both of those are in the evening somewhere. Content that carries nothing, and a self that keeps giving slightly different answers depending on nothing much at all.

What the model does, and what we could do

The row worth staying on is the second last one.

Nobody in a statistics department is trying to abolish the residual. There is no technique for it, and there is no ambition toward it either. The entire skill lies in recognising which part of the data is saying something and which part is not. And once that recognition happens, the noise doesn't get smaller. It stays exactly the size it was. It just stops having a claim on you.

But who is doing the recognising

Which brings up the thing I don't have a clean answer for.

In the model there is a statistician standing outside the data. The residual does not classify itself, obviously. Somebody has to look at it and say, this part is carrying information and this part is not.

In the evening, that standing-outside is precisely what has gone missing. There is nobody at that position. There is only the circling, and the circling has quietly occupied the chair of the one who was supposed to be watching it.

This is very old ground and it is not mine. Patanjali did not call them thoughts, he called them vritti, whirls, and the sutras open by proposing their stilling — which almost everyone reads as suppression, and which I don't think is suppression at all. It is nearer to what the statistician is doing. Look at the whirl. Ask what it is carrying. Most of the time it is carrying nothing, and the looking is the entire operation.

The drishta is a position more than a technique, and it is available while you are walking to the car, or waiting for a heavy file to open, or sitting acrossThe drishta is a position more than a technique, and it is available while you are walking to the car, or waiting for a heavy file to open, or sitting across from somebody who has been talking for too long. The cushion at five in the morning is where people go to practise it. Most of the noise, though, does not show up at five in the morning.

You are not the mind, you are not the body. This gets repeated so often that it has gone smooth and sounds like a slogan, until some evening when it doesn't — when a thought is halfway through its whirl and something in you notices, without any effort at all, that it isn't carrying anything. The thought keeps going. It doesn't stop just because you saw it.

But for that second, or less than a second, you had stopped modelling it.

And then the next one arrives at the same volume as the last, and the whole thing has to be noticed again.