12 modern delusions that must be challenged: #12

Number 12: Human behaviour can be predicted.
In the name of a supposedly "scientific" criterion of knowledge,
scholars are berated for not predicting the end of the cold war, the
rise of Islam, 9/11 and much more besides. Yet many natural sciences -
seismology, evolutionary biology - cannot predict with accuracy
either. Human affairs themselves, even leaving aside the matter of
human intentions and will, allow of too many variables for such
calculation. We will never be able to predict with certainty the
outcome of a sports contest, the incidence of revolutions, the
duration of passion or how long an individual will live.

My reply:
Isaac Asimov is probably to blame here, for writing the 'Foundation' sci-fi novels, wherein a galactic utopian civilisation is built on the work of mathematician Hari Sheldon, who constructs a model of humanity that can predict large-scale trends with complete accuracy. But this was written before the widespread work began on chaos theory, which says that very small changes in beginning conditions can result in very large changes to the state of the system over time. This means that unless you can know with arbitrary precision, the state of the system now, you cannot predict its behaviour over time. Most complex systems are also chaotic. The problem is not one of modelling (which is what people expect the scientists to do), but of data gathering (which is what they *should* actually do, and build the models after to help confirm or disconfirm their hypotheses).
Even if you could have sensors placed around the earth's atmosphere and oceans in a grid with 1cm intervals, and even though the mathematical models for fluid dynamics are perfectly accurate , you cannot measure the current state of the weather with a fine enough precision to predict more than a weeks worth of weather with accuracy. You can't even predict the behaviour of liquid in a small pot with accuracy using fluid dynamics.

Thus the only way to gather enough data on our 'reality' that you could predict its behaviour, is to measure and store its quantum state. Now if you have a large enough quantum computer (say one with the same number of quantum states as the reality you're measuring), you could store this measured quantum state and run it in simulation to 'predict' the behaviour of the 'real' system. There are three problems with this. Firstly, the computer would run no faster than the reality so it wouldn't really be a prediction at all. Secondly, perhaps more importantly and for various reasons that I mostly forget, in order to replicate a quantum state you have to destroy the original - so there would be no reality left to observe the results of the prediction, except inside the prediction. Thirdly, the quantum computer that you built was originally part of the quantum state that you're trying to simulate, so in trying to measure it and simulate it, you need to replicate it inside itself, and then destroy it outside of itself.

Hopefully that's enough paradox to adjudge the whole idea of predicting human behaviour as pointless.