The Problem of Induction
David Hume can rightly be called The Empiricist. He more carefully and thoroughly explicated the implications of empiricism than anybody before or since. Even the mighty Immanuel Kant (whom we will repeatedly reference as we continue in this course) spoke glowingly of Hume and said that Hume “awakened me from my dogmatic slumber.”
By “dogmatic,” Kant was referring to the rationalism articulated by Descartes, and you’ll remember our distinction between rationalism and empiricism. So, Kant was a Cartesian rationalist prior to reading Hume, and Kant took Hume to have nailed the coffin shut on rationalism. It is hard for a philosopher to imagine higher praise at that time than to have Kant say that you had totally changed his whole world view!
Hume realized that science is an entirely empirical enterprise, and it can rightly be said that Hume was the first true philosopher of science. And philosophy of science even today is in one way or another responding to or attempting to answer Hume. But Hume got a whole range of things spot-on-right, and the implications of his realizations haunt science to this day.
Certainly the most pressing problem for science that Hume articulated better than anybody in history is the so-called “problem of induction.”
You’ll remember that inferences divide into two types: deductive and inductive. You’ll also remember that deduction is a wholly “formal” enterprise, while induction requires an assessment of “content” in order to take a shot at specifying “how likely” it is that a conclusion does “follow from” a set of premises. Hume realized and rigorously argued that induction is not even a “rational” approach to inferences at all.
The two major sorts of inductive arguments are to: 1) argue from particulars to a conclusion about generals/universals; 2) argue from past events to a conclusion about the future. Science employs both sorts of inductive inferences.
Particularly with the second sort of inductive inferences, Hume found that the idea of causation and induction are closely linked. We infer that one event causes another through employing induction, and we see this linkage in scientific studies. Consider this example.
The 19th century cowboys widely feared rattlesnakes on the prairie, and they widely believed that if you got bit by a rattlesnake you had better start drinking whiskey. In fact, 2 ounces of whiskey every two hours was the minimum recommended dose. The US Army, in fact, had this dosage written in their field manual, so widely believed was the curative effect of whiskey on rattlesnake bites!
What inductive inferences could justify such a belief?
Well, consider the state of medical knowledge (and “research”) at that time. Consider that most rattlesnake bites are actually dry bites (wholly unknown at that time). Consider that the American rattlesnake is much, much less dangerous in general than are venomous snakes in many other parts of the world, so even venom-injecting bites are usually not fatal. These and other facts lead to a “sensible” (although ignorant) inductive argument that argues from the past to the future:
1) Joe, Bob, Marley, Sam, and every other cowboy I’ve ever heard of that got bit by a rattler started feeling really sick.
2) They all drank large amounts of whiskey right after the bite.
3) Not one of them died from their bites.
4) I’ve heard of other people getting bit that didn’t drink whiskey as they should have (stupid teetotalers).
5) I’ve heard of those teetotalers dying from their rattler bites.
6) I just got bit, and I’m no teetotaler!
————–
7) I should drink a lot of whiskey to keep from dying of my rattler bite!
Something along those lines was going on in the mind of virtually every rattler-bit cowboy of that era. And, guess what? He would drink whiskey, and he would recover just as had the many, many others who had been bit and recovered after drinking whiskey. So, his story would go down in the records as yet another “proof” that whiskey cured rattlesnake bite. Voila! Another “confirmation” of a pet theory!
But here is the key thing that Hume articulated: Induction “works” in many cases, thereby producing the sense that induction is “reliable,” but it does not always “work,” and it is impossible in principle to know whether any particular employment of the same inductive argument will “work” at that time.
Hume noted that we necessarily argue inductively about induction itself:
1) Induction has worked reliably in the past.
————–
2) Induction will work reliably in now and in the future.
But Hume asks: How “likely” is it that the conclusion follows from the premise? And the standard “answer” would be: “Well, it’s been working… uhh… quite a bit… most of the time.”
So, the “problem of induction” in a nutshell is that to justify our perception that “it works,” we necessarily argue in viciously circular fashion using the very inferential form that is at question!
Inductive inferences (if they are “good” ones) do (amazingly) seem to “work” most of the time. In fact, they “work” often enough, as a percentage of the time, that we often rely on them unthinkingly. We don’t question if the floor will be underfoot when we take our next step. We don’t question if our car will slow down when we hit the brakes. We don’t question if the fridge will be cold inside when we open the door. In fact, the vast, vast majority of expectations we have about the future, all of which are based upon inductive inferences, do turn out to result in true conclusions.
So, we have a deep psychological dependency upon induction, and that dependency seems to be “well founding” most of the time. However, Hume wants us to think critically about what is really happening in our thinking as we contemplate this dependency, and he wants us to realize that it really is “brute psychology” rather than “reason” that makes us think inductively.
Now, scientists (and many philosophers of science) respond to Hume’s points this way: Okay, sure, if you want to get really ticky-tacky about it, then, yes, induction is not a wholly-reliable, wholly-“rational” process. Who cares? It does work most of the time, in fact the vast majority of the time. So, I can’t strictly “prove” that the sun will rise tomorrow, but that does not mean that I am being “irrational” to believe that it will! In fact, I would be irrational to not believe that it will just because I cannot “prove” that it will! There is a very high degree of regularity in how the universe works, so it is not at all “unreasonable” to expect that the future will resemble the past. Thus, the force of the “problem of induction” is greatly reduced, because causality does in fact hold as we think it does. And that is because our good inductions do properly connect up with causality!
There are two major problems with this response:
First, on strictly empirical grounds, exactly zero account of “causality” can be given. Causality is a very opaque notion anyway, but on strictly empirical grounds it is an impermeable notion. Hume rightly articulated this. So contemporary scientists and philosophers of science are really “helping themselves” to a notion they have no explanation of whenever they blithely refer to causality. Whatever this thing is they think is producing the “regularities” in the universe, they have no account of it.
Second, again and again the notion of “works” arises. But pragmatism does not equate to truth. That a method “works” is not the same as that method being justified or correct. This is a point we will revisit in depth next week. But for now I’ll just assert that pragmatism does not equal truth.
So, in a nutshell now, this is no actual answer to Hume’s points, because this “answer” turns out to be a bunch of hand-waving and helping themselves to concepts they have no right to employ. It is really arguing, again, inductively about induction, which is not a justification of induction; it is merely an appeal to pragmatism and the (completely unsupported) claim that “what works is good enough.”
I will readily agree that “what works is good enough” for ENGINEERING. But it is not good enough for “science,” especially when science is now acting as though it is doing metaphysics, indeed the only good metaphysics!
If science were intellectually honest about what it is really doing, and it were humble and mitigated in its conclusions, then there would be no problem here. But when science claims that it is doing the only good metaphysics, then it must provide far better justification for its conclusions than a punt on Hume followed by the lame appeal to pragmatism: “It works well enough.” For science to really do metaphysics, it must answer Hume far, far better than this!