My understanding of the problem of induction is this. You want to hypothesise that the future is like the past (let us call this hypothesis H), and so justify the scientific practice of making inferences about the future from present data. However Hume points out that just because in the past the future was like the past doesn't mean this will continue to be true, and to assume otherwise is to assume the very thing you're trying to prove. Thus any apparent evidence that the future will be like the past is really circular and so doesn't count as evidence.

But it seems to me this is not really problematic, and what it really shows is that a certain kind of reductionism is false. What Hume's argument shows is that "the data don't speak for themselves," i.e. you always need to assume some hypotheses to have any idea what counts as evidence for or against something. If you assume H, one finds that the evidence, interpreted according to H, fits H, and so overall you have a "coherent picture." In other words, it implies a certain form of coherentism and/or Quinean holism. This doesn't mean that we can assume any hypothesis we like, because there is no guarantee that any hypothesis or set of hypotheses will form such a coherent picture. The reason Hume wasn't able to produce any evidence that the future will be like the past is he imagined that you could evaluate evidence for a single hypothesis without making any assumptions at all, which seems to me to be the kind of reductionism Quine attacked in Two Dogmas.

Do philosophers agree with this interpretation?

  • 3
    You only have the problem of induction if you demand that all theses are to be rationally and/or empirically grounded. Then the problem is that grounding induction in the "uniformity of nature" is either baseless or circular. You can, of course, drop the grounding requirement and say that induction can be left ungrounded because it is not needed anyway, as in coherentism, but that is not "implied" by the problem. Something like that was proposed by Popper, but it is one option among many and it has its own problems.
    – Conifold
    Commented Aug 14, 2021 at 17:24

3 Answers 3


On my reading, your formulation of coherentism here is: Hypothesize H; if the results of that assumption fit with that assumption, that corroborates H. That's problematic because it opens the gates for self-confirming hypotheses, confirmation-bias, etc. But I think some of what you're saying here is close to Kant's response to Hume's formulation of the problem of induction, which is in my view the most insightful response on offer thus far: "The future resembles the past" (as Hume put it, meaning by it, more precisely, "given the exact same conditions, the exact same result occurs") isn't just a hypothesis; it's a precondition of reasoning; if you ditch that principle, anything follows from anything, and then you can't infer anything about anything. I think H ("given the exact same conditions, the exact same result occurs") isn't one hypothesis among others in a holistic web of beliefs; it's a precondition for there being any epistemic relations between different beliefs in that web.


Conifold's comment answers the question, but it was rather brief, so here is a more detailed reply.

Basically, Hume didn't just point out the circularity; he proposed a trilemma of which the circularity was just one of the horns. Hume suggested that there are three possible ways that we might know about causes (by which he seems to mean what today we call laws of nature).

  1. Rational knowledge--knowledge that come prior to any experience.
  2. Empirical knowledge--that comes directly from experience or observation.
  3. A rule of induction--knowledge that we can inductively infer from multiple experiences.

Hume argues that our knowledge of laws of nature cannot be rational (that is, prior to experience) because we can always imagine that things might have been different, and can't give any rational reason that things are one way rather than another. Our knowledge cannot be empirical, because we have not directly observed or experienced a law of nature; we can only experience individual instances of that law being played out.

The last possibility is that we can use a rule of induction to infer from individual experiences to a general law, but whatever this rule of induction is, how do we know it? We can't know it rationally, for the same reasons as previously, and we can't know it empirically for the same reasons already given, so we can only know it from some rule of meta-induction. Then that rule of meta-induction has the same problem. We are caught in a vicious circle.


Quine is a pragmatist, a physicalist. Under such view, metaphysical considerations are not necessary, causality or induction are just assumed to be true. A priori/a posteriori or synthetic/analytic discussions have no place: we just know, and we can use language to produce scientific knowledge. In Two Dogmas..., Quine is just justifying the unnecessaryness of metaphysical considerations.

From my personal interpretation of Quine, he simply argues that language holds no truths (which is true: the dictionary is a set of circular references). That happens because he dismisses the essential role of metaphysics in reason, making his discourse addressing purely linguistic, and not epistemological/epistemic facts, but therefore making him a physicalist. Remember that Kant was searching to formalize metaphysics (or at least, what was "pure", pure as in Critique of PURE Reason). Quine does the opposite, falling down precisely on the problem Kant remarks: Quine just takes metaphysics for granted. To me, that's Quine's biggest issue.

Quine's approach follows precisely what science is founded upon: empirical truths. Assuming that the future will ressemble the past; so, it can be stated that entropy cannot decrease in a closed system. Never. So, the 2nd Law of Thermodynamics holds. ALWAYS. Because in science, the future ressembles the past. ALWAYS. If the future might not ressemble the past for just an instant, one can NEVER accept that E=mc^2. Such equation is not a probability: it is a scientific (empirical) truth.

But the problem of induction has metaphysical foundations. From a metaphysical approach (therefore, excluding Quine), one cannot expect the future to ressemble the past. You seem to have the rest of the problem clear from here on.

Remark C. D. Broad: "Induction is the glory of science, and the scandal of philosophy".

Do philosophers agree with this interpretation?

I would say most will not.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .