Ever notice that “forecasters” on local TV stations begin by talking about the weather already experienced in the region: “This morning we had above average temperatures,” “We started the day below average, but warmed up to normal,” etc. Are you happy with the history lesson? Isn’t it a bit like looking at a Selfie immediately after taking it, and saying, “Wow! I was much thinner back then,” or “Look at that hairdo. What were we thinking?”
Don’t look for a change in the weather forecast presentation. You’ll continue to learn what you already know at the beginning of the TV segment. But then, maybe the logic of presentation is based on showing data already gathered as the substance of the prediction. If a cold front passed by this morning, then colder air should prevail this afternoon. Call it deductively reasoned forecasting based on an a priori probability. Or, say it is a matter of Liouville’s theorem that says the phase-space distribution function is constant along a path. Gosh! It’s all so complicated, all this using mathematical models of the atmosphere and seas to make a weather prediction.
You might think that supercomputers in the hands of bright meteorologists could use a grid of data to make an accurate prediction and that inaccurate predictions should be rare. Your experience tells you otherwise. Weather forecasters can do pretty well (although sometimes they fail at this) at making historical statements about the weather (“This morning we had…”). And that’s because using real time data in complex formulae of even more complex theorems is impossible at our current level of data-gathering and computing power. There are just too many places in horizontal and vertical planes to monitor. Sometimes the local farmer (or you) might do as well at forecasting as the TV forecaster.
Think about the data sources in making a prediction: There are the horizontal ones that stretch over layers in the atmosphere on a moving planet, and there are the vertical ones that run from surface to stratosphere. Then there’s advection (wind), solar and terrestrial radiation, sea and land ice area, ground and sea temperatures, humidity, pressure, ocean currents, cloud cover and type, and precipitation (rain or snow). Beginning a weather forecast by telling what we already know is pretty much like saying, “I have some good news and some bad news. I’m going to give you the good news first to soften you up before I hit you with a chancy prediction, especially if you want a long-term prediction.”
That’s largely the way we do our predicting about human behavior. We use an a priori probability that we variously call stereotype or character or history, and we deduce from what we have already experienced. But like the weather’s complex causes, human behavior also has numerous unknown data points. Of course, sometimes we get “it” right. We just have a feeling that someone will precipitate a certain behavior sometime in the near future. As for long-term predictions, however, we are no better than weather forecasters. Just too much data to crunch.
That’s good news. Our inability to predict long-term behavior reveals itself in the changed behavior of many formerly characterized by their addiction to substances or social pressures inimical to their health and the well-being of others.
Would any of us want to be so predictable that our lives are essentially already lived at the beginning of the forecast? Would any of us want the mistaken forecasts of the past to be the accepted forecasts of our personal futures?
Probably not. We like to think that however complex weather forecasting is, its complexity pales in comparison to our own, and that no accumulation of data points, no set of formulae, and no a priori forecasts, even with the most sophisticated supercomputers, can unfailingly describe.
Don’t look for a change in the weather forecast presentation. You’ll continue to learn what you already know at the beginning of the TV segment. But then, maybe the logic of presentation is based on showing data already gathered as the substance of the prediction. If a cold front passed by this morning, then colder air should prevail this afternoon. Call it deductively reasoned forecasting based on an a priori probability. Or, say it is a matter of Liouville’s theorem that says the phase-space distribution function is constant along a path. Gosh! It’s all so complicated, all this using mathematical models of the atmosphere and seas to make a weather prediction.
You might think that supercomputers in the hands of bright meteorologists could use a grid of data to make an accurate prediction and that inaccurate predictions should be rare. Your experience tells you otherwise. Weather forecasters can do pretty well (although sometimes they fail at this) at making historical statements about the weather (“This morning we had…”). And that’s because using real time data in complex formulae of even more complex theorems is impossible at our current level of data-gathering and computing power. There are just too many places in horizontal and vertical planes to monitor. Sometimes the local farmer (or you) might do as well at forecasting as the TV forecaster.
Think about the data sources in making a prediction: There are the horizontal ones that stretch over layers in the atmosphere on a moving planet, and there are the vertical ones that run from surface to stratosphere. Then there’s advection (wind), solar and terrestrial radiation, sea and land ice area, ground and sea temperatures, humidity, pressure, ocean currents, cloud cover and type, and precipitation (rain or snow). Beginning a weather forecast by telling what we already know is pretty much like saying, “I have some good news and some bad news. I’m going to give you the good news first to soften you up before I hit you with a chancy prediction, especially if you want a long-term prediction.”
That’s largely the way we do our predicting about human behavior. We use an a priori probability that we variously call stereotype or character or history, and we deduce from what we have already experienced. But like the weather’s complex causes, human behavior also has numerous unknown data points. Of course, sometimes we get “it” right. We just have a feeling that someone will precipitate a certain behavior sometime in the near future. As for long-term predictions, however, we are no better than weather forecasters. Just too much data to crunch.
That’s good news. Our inability to predict long-term behavior reveals itself in the changed behavior of many formerly characterized by their addiction to substances or social pressures inimical to their health and the well-being of others.
Would any of us want to be so predictable that our lives are essentially already lived at the beginning of the forecast? Would any of us want the mistaken forecasts of the past to be the accepted forecasts of our personal futures?
Probably not. We like to think that however complex weather forecasting is, its complexity pales in comparison to our own, and that no accumulation of data points, no set of formulae, and no a priori forecasts, even with the most sophisticated supercomputers, can unfailingly describe.
RSS Feed