The Regress Problem

One of the traditional problems in epistemology is the Regress Problem. This problem arises out of a few seemingly trivial ideas about “reasonability” (what I will be labeling below as the “Initial Definition of ‘Reasonable Belief’”).

Initial Definition of “Reasonable Belief”

Remember, one of the concerns of epistemology is to give criteria for “reasonable belief”. Epistemologists attempt to specify what it is that makes a belief reasonable. To put it another way, epistemologists try to spell out the difference between a reasonable belief and an unreasonable one.

Now, this might seem like a rather silly thing for anyone to study. In part it might seem silly because the answer seems obvious. Isn’t the obvious difference between a reasonable and an unreasonable belief simply that a reasonable belief is just one that you have reasons for, and an unreasonable belief is one that you don’t have reasons for?

This obvious answer does capture part of what we think about reasonability. For example, let’s say that I am of the belief that the U.S. should not have invaded Iraq. Upon hearing of this belief, you might press me to cite my reasons in support of the belief. If I was completely unable to give you a single reason why we shouldn’t have invaded Iraq, you would probably think that my belief is unreasonable.

But notice that the obvious answer above doesn’t capture everything that we require of a reasonable belief. While it’s true that a complete lack of reasons does make a belief unreasonable, we don’t think that a belief is reasonable simply in virtue of having any old reasons supporting it. Often we assess how good the reasons themselves are.

Using the Iraq example again, let’s say that when you press me for my reasons I tell you the following: We shouldn’t have invaded Iraq because Iraq made a pact with a race of space aliens. After the attack, Iraq contacted their alien friends who deployed a deadly squad of pink robots to annihilate the U.S. The pink robots will be here soon.

In this example, I do have some reasons in support of my belief that we shouldn’t have invaded Iraq. But I imagine that most of you would still think my belief is unreasonable. And I imagine that you would say that my belief is unreasonable precisely because my reasons themselves are crazy. My reasons themselves are unreasonable.

So, in light of this last example we might propose the following definition of a “reasonable belief”.

A reasonable belief is one that you have good (reasonable) reasons for.
An unreasonable belief is one that you either have no reasons for, or only have bad (unreasonable) reasons for.
In what follows, I'll refer to this definition as our “Initial Definition of ‘Reasonable Belief’” (and sometimes simply as our “initial definition”).

The Regress Problem

To repeat, our Initial Definition of “Reasonable Belief” seems obvious and fairly trivial. It’s the kind of thing that most people propose when asked to define “reasonable belief”. However, this initial definition leads quickly to a serious problem. The problem is that this definition can never be satisfied. That is, on this definition of “reasonable belief” it looks like no one can ever have a reasonable belief. Put roughly, in order to have a reasonable belief according to our initial definition it looks like we need to have an infinite number of supporting reasons.

To illustrate the problem, let’s consider, in abstract, what it would take for one of my beliefs to be reasonable. Let’s label my original belief as “B”. (Here B just stands for some belief, say my belief that “The U.S. should not have invaded Iraq”.)

According to our initial definition, for belief B to be a reasonable belief I must have some reasons in support of B. Let’s keep the case simple and assume that I just have one reason in support of B. Let’s label this reason “R1”.

But not just any reason R1 will make belief B reasonable. The reason has to be a good one. (Remember the Pink Robots example above.) According to our initial definition, R1 itself must be a reasonable belief.

But for R1 to be a reasonable belief it too must satisfy our initial definition. This means that I need some new reason R2 in support of R1. But again, not just any R2 will do, R2 also has to be reasonable. But this means that I need a reason R3 in support of R2. And of course R3 will need a reason and so on:

So, given our Initial Definition of "Reasonable Belief", for any belief to be reasonable, it needs to be supported by an infinitely long chain of reasons. This is the Regress Problem, because it’s not possible to have an infinite number of reasons in support of any belief. (Though I am a pretty bright guy, it seems plausible to think that I am limited to a finite number of beliefs.)

Responding to the Regress Problem

To repeat, the regress problem arises out of our Initial Definition of "Reasonable Belief". One way to take the regress problem is as challenge. To solve the problem we need to modify (or add to) our initial definition in a way that avoids any regress. We need to give a new definition that doesn’t require us to have an infinite number of supporting reasons for any reasonable belief.

It’s worth noting that there have been a number of attempted solutions to the regress problem. But in class, following Plantinga's article, we looked exclusively at the way Foundationalists have attempted to solve the regress problem.