Professional Documents
Culture Documents
Penultimate draft. Final version forthcoming in Philosophia. All citations should be to the published version.
Matt Lutz
In this paper, I defend an inductivist solution to Hume's Problem of Induction against the
popular counterinduction parody argument. Once we examine the structure of the inductivist position
Induction is a reasoning process that we use to draw probabilistic inferences about how things
will stand in unobserved instances. Hume famously wondered how induction could be justified. All
belief-formation seems to come from one of three processes: direct observation, demonstrative
reasoning (which yields a priori knowledge), and inductive reasoning (which yields a posteriori
knowledge of empirical generalities and, thereby, knowledge of unobserved particular instances). This
forms a trilemma: if these three processes are our only sources of belief, there seems to be no way for
us to know that induction is reliable. On the first horn of the trilemma, we attempt to justify induction
by observation. But we cannot know that induction is reliable through observation because induction is
a mode of reasoning that gives us knowledge of unobserved instances. On the second horn, we attempt
to justify induction by demonstration. But it is not a priori certain that unobserved instances will be
like observed instances; it is always conceivable that the future will not resemble the past. And on the
third horn, we attempt to use induction to justify induction. But we cannot do this, either; that is
circular reasoning. So, with no way to justify induction, we are forced to skepticism about induction.
The inductivist challenges the third horn of Hume's trilemma. According to the Inductivist, we
can use induction to justify induction. There may be a kind of circularity involved in doing this, but we
ought not view this kind of circular reasoning as viciously circular. The Inductivist views this sort of
I will not directly address the question of whether the kind of circular reasoning that
inductivism manifests is vicious. Instead, my concern will be with a second objection, the
counterinduction parody, as formulated by Salmon (1957) and popularized by Skyrms (1966). Hume
contented himself with the observation that circular reasoning seems to be vicious. But, after some
began to warm to the idea of an inductive justification of induction (Black 1949), Salmon argued that
something must be wrong with the inductive justification of induction because it can be successfully
parodied. If justification can be justified inductively, argued Salmon, then counterinduction can be
justified counterinductively. And since it is absurd to say that counterinduction can be justified
According to inductivism, induction can be used to justify itself, in something like the following
way: We are looking for a reason to believe that the past will resemble the future. Here is the reason
that we can provide: in the past, the future has always resembled the past. And the fact that, in the past,
the future has always resembled the past gives us a reason (by induction) to think that, in the future, the
future will continue to resemble the past. This follows from the principle of induction: that the future
will resemble the past. Thus, the proposition that induction is rational can be rationally deduced from
propositions about the past, by aid of the inference rule of induction. The remarkable success of past
Counterinductive inference is the inference rule whereby one infers that the future will not
resemble the past. For the inductivist, the fact that the sun has risen every morning is a reason to
believe that the sun will rise tomorrow, and the strength of this reason is proportional to how often the
sun has risen in the morning. The fact that the sun has an unbroken track record of rising is what gives
us such good reason to think that the sun will rise tomorrow. For a counterinductivist, on the other
hand, the fact that the sun has risen every morning is a reason to not believe that the sun will rise
3
tomorrow, and the strength of this reason is proportional to how often the sun has risen in the morning.
The fact that the sun has an unbroken track record of rising is what gives us such good reason to think
that the sun will not rise tomorrow – or so says the counterinductivist.
rational, and no one would be rational to use counterinduction in their reasoning. But counterinduction
can be justified: counterinductively. In the past, the future has always resembled the past. And the fact
that, in the past, the future has always resembled the past gives us a reason (by counterinduction) to
think that, in the future, the future will not resemble the past. This follows from the principle of
counterinduction: that the future will fail to resemble the past. Thus, the proposition that
counterinduction is rational can be rationally deduced from propositions about the past, by aid of the
inference rule of counterinduction. The remarkable failure of past counterinduction gives us a strong
counterinductive reason to believe in its future success. The lesson of the parody is that a rule's ability
to support itself is not sufficient for being justified in adopting that rule. And that's a huge problem for
inductivism.
In stating this parody argument, I have been speaking rather loosely. In what follows, I will
show that this argument sounds compelling only because I have been speaking loosely. In the next
section, I will state precisely what I take the inductivist position to be. In Section 3, I will show that this
2. Distinctions
We begin by drawing a distinction between three things that are often conflated under the
common term 'induction.' Understanding the differences between these three things will be essential to
In a deductive argument, the premises entail the conclusion of that argument. To believe the
4
conclusion of an argument on the basis of premises that entail it is a certain kind of reasoning process,
called deduction. In an inductive argument, the premises give a reason to believe the conclusion even
though those premises do not strictly entail the conclusion. The inferential step that a reasoner performs
when moving from premises to conclusion cannot be accurately described as deduction; we therefore
call it induction. Thus understood, induction is a kind of process that one goes through in one's
reasoning. While an agent may go through many different processes in her reasoning, what makes a
piece of reasoning count as inductive is that it is reasoning in accordance with a certain rule. To wit:
The important thing to consider about the rule of induction is that it is a proper rule. The logical
form of the Rule of Induction is an imperative. It does not describe the world as being any particular
way.
However, reasoning in accordance with the Rule of Induction will only tend to be reliable if the
world is a certain way. In this way, the Rule of Induction seems to presuppose that a certain kind of
claim about the world is true. This is a claim that Hume called
resemble those, of which we have had experience, and that the course of nature
It is this principle that Hume famously sought justification for. Hume's concern was that he did
not see what reasoning process one could use to gain justification for believing the Principle of
Induction. The Principle of Induction is a claim about the world that cannot be known directly by
observation or by deduction. But this means that we can only justify the Principle of Induction by
reasoning in accordance with the Rule of Induction. And this, thought Hume, was unacceptably
1 Hume refers to this principle several times, sometimes including a 'must' to indicate that this principle is necessary,
sometimes not. The version of the Principle stated here omits the modal term.
5
circular; reasoning in accordance with the Rule of Induction presupposes the truth of the Principle of
Induction.
accordance with that Rule will tend to be reliable only if that Principle is true.
induction; see Alston (1986, p. 1-4) for a discussion of this conception of presupposition.) The
inductivist position, then, is that reasoning in accordance with the Rule of Induction can justify belief in
the Principle of Induction, despite the fact that the Rule of Induction presupposes the Principle of
Induction.
Note, however, that while the Rule of Induction presupposes the Principle of Induction, it does
not directly justify the Principle of Induction. The Rule of Induction is a description (in imperative
form) of a method for forming beliefs. It serves only to pick out a set of reasoning-tokens (the tokens
that accord with the method). The Principle of Induction is a descriptive, general principle: it says
something about the way the world is. Neither the Principle of Induction nor the Rule of Induction is a
normative epistemic principle that states conditions under which one can be justified in believing a
claim. But the inductivist's conclusion is that we are justified in believing the Principle of Induction.
We're missing the most important element of the inductivist position: the normative epistemic element.
In order for the Rule of Induction to justify belief in the Principle of Induction, we need to rely on
Successful Prediction: For any Rule of inference, if that Rule instructs you to
believe that P, and you receive evidence that P, that evidence provides some
Successful Prediction is an epistemic principle stating sufficient conditions for a certain kind of
evidential support. If you reason in accordance with a Rule, then later gain evidence that the Rule made
a successful prediction, this is evidence that the rule is reliable. And, given Presupposition, evidence
6
that a rule is reliable will be evidence in favor of any Principle that rule presupposes. The kind of
evidential support in question provides “some confirmation” (not conclusive reason) for thinking that a
Principle is true. Thus, Successful Prediction is a kind of inductive principle. The confirmation is also
defeasible because it is always possible to receive additional evidence which would undercut or
overturn the evidence provided by successful predictions. This is a characteristic of all non-conclusive
evidence.
Although Successful Prediction is an inductive principle, it does not only apply to the rule of
induction. Successful Prediction also implies, for instance, that when I get reason to believe that I have
hands, this is some confirmation that the world is the way that my senses present it to be, because the
Rule “trust your senses” presupposes that the world is the way my senses present it to be. In this way,
particularly p. 6-7). Indeed, Successful Prediction can feature in an inductive justification of any mode
Fitting Successful Prediction into a larger theory of epistemic justification falls outside the
scope of this paper, and as such it is somewhat schematic; I take no stand here on how much
confirmation successful predictions provide, or the circumstances under which this confirming
evidence would be defeated. The point is that Successful Prediction is distinct from both the Rule of
Induction and the Principle of Induction. The Rule of Induction is a reasoning process; the Principle of
Induction is a general, descriptive claim about the world. Successful Prediction is a normative
epistemic principle that states one way in which a claim can be inductively justified.
I have just presented an expanded version of the inductivist position. I don't intend to argue here
that there is nothing wrong with this inductivist position – that is outside the scope of this paper.
7
Instead, my goal is to argue for the more modest thesis that the counterinductivist position is not a
successful parody of the inductivist position. Now that a distinction has been carefully drawn between
the Rule of Induction, the Principle of Induction, and Successful Prediction, along with the role that all
three of these principles play in the inductivist position, we can see why the counterinduction parody
fails.
The inductivist position is that belief in the Principle of Induction can be justified by using the
Rule of Induction, even though the Rule presupposes the Principle. The counterinductivist parody
would then have to say that belief in the Principle of counterinduction can be justified by using the
observed instances, infer that the regularity does not hold in unobserved instances.
experience do not resemble those, of which we have had experience, and that the
The Counterinductivist claims that the Rule of Counterinduction justifies belief in the Principle
of Counterinduction, even though the Rule presupposes the Principle. But this can't be right. A rule's
presupposing a principle does not suffice to show that that Rule justifies the principle. Successful
Prediction (or something similar) is needed for that. But the counterinductivist cannot accept
Successful Prediction as part of the justification of the Principle of Counterinduction. This is because
Unsuccessful Prediction: For any Rule of inference, if that Rule instructs you to
believe that P, and you receive evidence that ~P, that evidence provides some
8
The Principle of Induction and the Principle of Counterinduction are both empirical
generalities; they both describe general ways that the world can be. But Successful Prediction and
Unsuccessful Prediction are not particular facts about the world, knowable through observation, nor are
they empirical generalities. They are normative epistemic claims, that state sufficient conditions for a
Counterinduction describes a failure of patterns to continue; and as Hume correctly points out, we have
no a priori reason to prefer one empirical generality to the other. But we cannot say the same thing
about Successful Prediction and Unsuccessful Prediction. Those are not contingent, empirical
principles concerning the progress of events in the world. They are normative epistemic claims that
state sufficient conditions for a certain kind of evidential support. This makes Successful Prediction a
truth of pure reason, knowable a priori through deductive reasoning. Theories are (somewhat, ceteris
paribus) confirmed when they make successful predictions. By similar token, Unsuccessful Prediction
is a philosophical falsehood par excellence: a necessarily false claim that is knowable to be so a priori.
Theories are (somewhat, ceteris paribus) disconfirmed when they make unsuccessful predictions.
These are both, plausibly, conceptual truths.2 The success of our predictions is what distinguishes
The counterinductivist claims that, at a given time, both an inductivist and a counterinductivist
might be equally rational in believing in the Principle of Induction or the Principle of Counterinduction,
respectively. But how could a counterinductivist come to rationally believe that the Principle of
2 I believe that they are conceptual truths because I am, like Hume, a skeptic about the synthetic a priori. But my
argument doesn't rest on this assumption, so I won't defend it. Kantians may hold that this is an instance of synthetic a
priori knowledge.
9
Counterinduction is true? Evidence accumulates over time, and, over time, The Principle of
Counterinductivist will say that this very disconfirmation provides a reason to believe the Principle of
Counterinduction. But that is incoherent; when a Principle is disconfirmed, that's a reason not to
believe it.
The root problem for the counterinductivist parody is that the counterinducitivst does not only
propose a (highly implausible) claim about the way the world will be in the future. The
counterinducitivst is also committed to the idea that some beliefs may be justified counter-inductively.
Thus, the counterinductivist must, at some point, advance a principle that states (ceteris paribus)
sufficient conditions for “counterinductive justification.” Neither the Principle of Counterinduction nor
the Rule of Counterinduction is such a principle, because neither is a normative epistemic principle.
justification; it is a principle of inductive justification, and it does not support the Principle of
justification that supports the Principle of Counterinduction, but we can know a priori that it is false.
One might object that I have mischaracterized inductivism, in the following way: One might
claim that the Rule of Induction is not a mere description of a method, but an epistemic norm that
essentially justifies any principle that it presupposes. Thus understood, the Rule of Induction would, by
itself, serve to justify the Principle of Induction, thereby obviating the need for an appeal to Successful
Prediction. An inductivist position of this kind can therefore be parodied without having to rely on
Against this objection, I have two responses. First, my goal in this paper is only to state an
10
inductivist position that cannot be successfully parodied by the counterinductivist. The fact that a
Indeed, if it is true that versions of inductivism that are not committed to Successful Prediction can be
successfully parodied, that's a very good reason to prefer versions of inductivism that are so committed.
But I don't wish to rest content with this first response, because the objection fails for another reason.
The reason that I separated the Rule of Induction from Successful Prediction in Section 2 was to
separate a simple imperative – the Rule of Induction – from an epistemic normative principle,
Successful Prediction. Claiming that the Rule of Induction is normative would, in effect, combine the
descriptive elements of the Rule of Induction with the normative commitments of Successful
Prediction. By similar token, then, a modified Rule of Counterinduction would not only describe
counterinductive reasoning, but also state that reasoning in this way is justified. But that's a substantive
easily avoided. That principle is essential to the “counterinduction is justified counterinductively by its
own failures” gambit, and that principle – not any descriptive claim about the way the world will be in
3 [ACKNOWLEDGMENTS]
11
Bibliography
Alston, William P. (1986). Epistemic circularity. Philosophy and Phenomenological Research 47 (1):1-
30.
Black, Max (1949). The Justification of Induction. Proceedings of the Tenth International Congress of
Philosophy 2:791-793.
Skyrms, Brian (1966). Choice and Chance: An Introduction to Inductive Logic. Dickenson Publishing.
Strawson, PF (1952). An Introduction to Logical Theory. John Wiley & Sons, Inc.
Van Cleve, James, (1984). Reliability, Justification, and the Problem of Induction. Midwest Studies in
Philosophy 9 (1):555-567.