No Principled Constraints

by Caleb Ontiveros

There's a kind of argument type that I call the no principled constraints objection. It's a kind of argument that appears in philosophy here and there. But it's not a successful objection.

I'll give three examples and then show why the argument type doesn't work.

The first two have to do with generality.

Here is a simple view about justification: a belief is justified if and only if it is produced by a reliable belief generating process. I believe that there's a laptop in front of me - a functioning visual system produced this belief. If I believe that a pink elephant is in front of me because I took hallucinogenic drugs, that belief wouldn't be justified. Hallucinogenic drugs aren't known for producing accurate perceptual beliefs. So, determine whether a particular belief is justified or not by determining whether the process that produced it is reliable. This view is called reliabilism. Here's a potential problem with this view:

  • There's no principled way to constrain "reliable belief generating process" to play the relevant epistemic role.

Less abstractly, when I look at my laptop, there are many candidate belief generating processes:

  • Perceptual processes.
  • Perceptual processes operating over objects in my house.
  • The perceptual process that is specific to recognizing laptops.

Each of these has different levels of reliability. For example, I'm better at recognizing laptops than at recognizing objects in the world generally. Yet there are likely many different candidate belief generating processes. What's to say that one is the relevant process that matters for justification? In the drug case, is the relevant process my general perceptual process or my perceptual process operating while drugged? The former is generally reliable, while the latter is not. But what principled way is there to handle the many cases like this? Philosophers have used this as an objection against reliabilism.

A similar problem arises for Kantian ethics. Roughly, according to Kant, ethics is about acting under a maxim that can be rationally universalized. This is the first categorical imperative. The second categorical imperative states that one should only treat people, including yourself, as ends and not merely as a means. According to Kant, these are the same in some sense. But let's stick with the first. If you're acting ethically, you're acting under rules that you can rationally endorse others acting under as well. Classically, Kant said that it is always wrong to lie because you can't rationally universalize a maxim that allows for lying. This is silly, but there is a question about what maxim you're endorsing when you lie in a permissible way. Is it:

  • It's permissible to lie to people to stop them from doing grave wrong.
  • It's permissible to lie if you're hiding political fugitives.
  • It's permissible to lie when telling the truth would be deeply offensive.

Like with reliabilism, there are many different candidates, and one's choice matters for the theory. The no principled constraints objection states that there's no principled way to constrain the relevant concept of acting under a maxim. Hence, Kantian ethics cannot be successful.

Here's another example concerning the principle of indifference. The idea is that if you don't have any evidence about some number of hypotheses, then you should assign equal probability to them. The problem is that it can be hard to know how to carve up the hypothesis space. Suppose a factory produces tiles with sides that have a length between 1 to 3 inches. That's all that you know. What is the probability that a given tile produced by the factory has a side between 1 to 2 inches? A reasonable answer is 1/2. Now, suppose there's a factory that produces tiles with the area of 1 to 9 inches. That's all that you know. What is the probability that you receive a tile with the area of 1 to 4 inches? Typically, if people are asked that question, they'll say 4/9. Should you divide up the hypothesis space by the length of the sides or by the possible areas? The idea is that there is no principled way to carve up the probability space and so the principle shouldn't be used.

The general idea behind these objections is:

  • There is no principled way that a concept (or set of concepts) can be constrained to play the relevant role.

Candidate hypotheses can vary along many dimensions, as can maxims and belief generating process. The problem is that it's challenging to come up with a principled way for constraining what counts as satisfying the concept and what doesn't. If there's no principled way to constrain the concept to play the relevant role, then the philosophical account fails.

Arguments of this form don't work. No mechanism showing that there can't be a principled constraint is given in any of the objections above. A priori, there are many ways to constrain a concept and it's likely that at least some of them are principled. Given that there are many ways to constrain a given concept, we shouldn't expect any of the above arguments to work without an account of why exactly constraining the concept can't be done in a principled way. Of course, reliabilism, Kantian ethics, and the principle of indifference may be false for other reasons. I assure I am no Kantian.

To motivate the idea that these concepts can be constrained, let's return to the principle of indifference. Suppose you know that someone took a trip of 100 miles in their car. The trip took between 1 - 2 hours and Sue's average speed was between 50 and 100 miles per hour. What's the probability that the trip took between 1 hour and 1 1/2 hours? If your carve up the possibilities by duration, the answer is 1/2. If you carve up the possibilities by speed, the answer is 2/3. Which should you prefer? The time that it takes to travel a given distance is explained by speed, not a range of possible times. So, speed is explanatorily prior. The idea is to constrain hypotheses by the explanatory priority of their components. Hence, the probability that the trip took between 1 - 1 1/2 hours is 2/3. Variants of the tile case given earlier are tricker, but this is just one example of answering the charge against the principle of indifference.

Here's an example of a related argument of a different type (though there are related statements that don't):

  • Phenomenal properties are of a fundamentally different kind from physical properties.
  • If two kinds of facts are of a fundamentally different kind, then they cannot ontologically reduce to one another.
  • Phenomenal properties cannot be ontologically reduced to physical properties.

To be clear by phenomenal, I don't mean "really great facts," I mean facts about conscious experience. A justification for the first premise is that phenomenal properties are intrinsic, while physical properties are structural and functional. That's a controversial idea, but that's the kind of mechanism one needs when running a no principled constraints objection.

The fact that it's hard to see how to constrain a given concept is a problem for a view - one that should be solved. But it doesn't show that the view is likely false.