Causal forces and scientific explanations
But it gets worse, for the universe of causal forces may not be subject to a simple, though vast, equation. There can be cases in which a tiny change causes a cascade, so that a butterfly might cause a tornado in Beijing—the famous butterfly effect. And as Stephen Wolfram (among others) has so convincingly shown, even simple rules can lead to literally unpredictable results.
Now, of course it’s hard to make up an example in which we actually care where any one piece of confetti falls. But it has an important effect on how we think about scientific explanations. We have liked science’s models because they are deductive: If you know the formula for gravity and you know the mass of two objects, you can deduce their mutual pull and the relative movement of each of them. Deductions are logically incontestable, so long as the premises and variables are correct. We have good reasons to believe that science’s models are true and correct; science itself provides the methodology for coming to belief about these models. But if you can't get the variables right, then you can’t do the deduction with any certainty. So, we can have nailed the law of gravity but still not be able to predict where confetti will fall.
Similar to confetti
And everything is similar to confetti falling. We can only deductively predict things we care about—the path of a space probe, the effectiveness of airbags— by accepting that those results are approximate and probabilistic exactly the way that deductions are not supposed to be. Science will not give up on hypotheses. But it already is becoming more willing to accept results based on the sorts of statistical analyses performed by machine learning. And it may be that when science does rely on theories and laws, we will recognize that no matter how ironclad they are as generalizations, their application to a world of confetti will always and necessarily render them approximate and probabilistic.