No Need for Explanation

Here, I explain why Phenomenal Conservatism is better than Phenomenal Explanationism.*

[ *Based on: “No Need for Explanation,” Asian Journal of Philosophy 3 (2024), https://doi.org/10.1007/s44204-024-00187-2. ]

1. Background

Phenomenal Conservatism (PC) holds that if it seems to you that P, and you have no reason to doubt P, then you thereby have some justification for believing P. I’ve defended this in earlier papers and posts (1, 2, 3, 4, 5).

Phenomenal Explanationism (PE) holds that a seeming that P provides justification for P when and only when the truth of P is part of or implied by the best explanation for one’s having that seeming. (See Kevin McCain & Luca Moretti, Appearance and Explanation, 2021.)

Someone invited me to write a paper about this.

2. Replies to Objections to PC

Here, I respond to objections to PC raised by McCain and Moretti (M&M).

2.1. Explaining defeat

M&M objected that PC doesn’t explain how defeat works, i.e., how additional information can defeat the justification that an appearance would normally provide for believing P.

I think PC can easily explain this. Justified beliefs are, basically, beliefs that make sense to adopt if you want to have true beliefs and to avoid false ones. If P seems true to you and you have no reason to doubt it, then it makes sense, from your own point of view, to adopt the belief that P.

But also, if you then acquire some evidence of ~P, or you just acquire evidence that your appearance that P isn’t a reliable indicator of the truth in this case, then it no longer makes sense, from the standpoint of the goal of having true beliefs and avoiding false ones, to believe P.

2.2. Reflective awareness

Maybe this case poses a problem for PC: imagine that S is taken, blindfolded, into a room. In advance, S knows that one of the room’s walls will be green, while another will be white but illuminated by green light such that it will look exactly like a green wall. Upon removing the blindfold, S sees a wall in front of her that looks green. What should she believe?

It seems that S should have a 50% credence that the wall in front of her is actually green, and a 50% credence that it is instead white but illuminated by green light. How would a phenomenal conservative explain this? Why can’t S outright believe that the wall is green?

I think the answer is that the information [Half of the green-looking walls in this room are not actually green] counts as a defeater for [This wall is green]. This is because [Half of the green-looking walls in this room are not actually green] implies that an appearance-based belief that a wall in this room is green would not be reliable (sc., it would be only 50% reliable). In general, the information that your belief-forming method is unreliable counts as an undercutting defeater for that belief’s justification.

2.3. Bootstrapping

The problem for PC

Suppose that PC is true, and suppose that I formed the belief that P based on an appearance that P, which I am aware of (and I have no defeaters). Then it seems that I could make the following inference:

1. P.

2. It appears to me that P.

3. Therefore, my appearance is accurate.

(1) is justified by the appearance that P. (2) is justified by introspective awareness. (1) + (2) support (3), so it looks like this argument justifies (3). If that works, then I could do this repeatedly, for many different appearances, and thence conclude that in general, my appearances are a reliable guide to the truth.

Many people find it counter-intuitive that I could acquire justification in that way for thinking that my appearances are reliable. It seems objectionably circular or something.

The (alleged) problem for PC is that it’s unclear what a Phenomenal Conservative could say is wrong with these “bootstrapping” arguments (as epistemologists call them).

The problem for everyone

If this is a problem, I think it is a problem for virtually any theory, not just PC. Suppose your theory is that x is justified in believing y whenever Phi(x,y) holds. Phi(x,y) can be any condition whatever that relates a person to a proposition.

Now suppose S believes P, which is justified because Phi(S,P) obtains, and S knows this. Then S can construct the following argument:

1. P.

2. Phi(S,P).

3. Therefore, Phi(S,P) → P.

S could go on to collect many more instances like this (with different propositions), and thence conclude that the obtaining of Phi(S,x) is a reliable indicator of truth. And the same people who had a problem with the earlier bootstrapping inference will have the same problem with this one. So the bootstrapping problem, if it’s a problem for PC, is a problem for any theory of justification that cites a condition on justification that people sometimes know to obtain.

What’s wrong with bootstrapping

Briefly, bootstrapping inferences fail to raise the probabilities of their conclusions. Suppose your initial estimate of the reliability of your appearances (or whatever you’re using to form beliefs) is 90%. Then, in each of the bootstrapping inferences above, you would have 90% credence in premise (1). When you deduce (3), then, (3) will have only 90% or lower probability.

In general, if you have a belief-forming method that you initially estimate to be x% reliable, you cannot infer that it is more than x% reliable using premises that come from that very method, because those premises will themselves be only x% likely to be true. Since bootstrapping inferences don’t raise the probabilities of their conclusions, we appropriately judge them as failing to justify their conclusions.

3. Objections to PE

3.1. Necessary truths

PE says that an appearance justified P whenever P is a consequence of the best explanation for that appearance. However, every necessary truth is a consequence of every proposition (or set of propositions) whatsoever. So any necessary truth will be a consequence of the best explanation for any appearance, no matter what the appearance is and no matter what the explanation is.

For example, the Banach-Tarski Theorem in set theory says that it is possible to decompose a sphere into a finite number of parts, rearrange the parts (using rigid translations and rotations, with parts allowed to move through each other), and reassemble them into two spheres, each the same size as the original. This is very non-obvious and requires a clever proof to justify.

But on PE (as articulated by M&M), all you have to do to be justified in believing the Banach-Tarski Theorem is see a squirrel. Whatever the best explanation is for your squirrel appearance, it will entail the Banach-Tarski Theorem, since the Banach-Tarski Theorem is a necessary truth.*

(*The theorem requires the Axiom of Choice. If you reject it, just take some other example of an abstruse and non-obvious necessary truth.)

3.2. Unreflective observers

There are many observers—such as animals, children, and education majors—who fail to reflect (and perhaps cannot reflect) on their own appearances and what is the best explanation for those appearances. For these observers, facts about the best explanation for their appearances would play no role in how they form beliefs.

Given this, it seems that the Phenomenal Explanationist would have to say that these unreflective observers lack justified beliefs. (If certain facts about explanation are essential to the justification for P, then for a person to have a justified belief that P, it seems that the person’s belief must somehow be based on those facts about explanation.)

PC avoids this problem, because the Phenomenal Conservative only requires you to have an appearance with no grounds for doubt. If this appearance causes you to form a belief, that belief will then be prima facie justified. This works fine for unreflective observers.

3.3. Flexibility

PE holds that you have justification for believing whatever is part of or a consequence of the best explanation for your appearances. At first glance, this is just slightly more complex than PC, so this is perhaps a slight disadvantage for PE.

However, what actually matters is the flexibility of a theory, or the range of possible evidence that it is able to accommodate, with smaller ranges being better, provided that the actual evidence is accommodated. Simplicity is just a proxy for that, as more complex theories are typically capable of accommodating wider ranges of possible data. (See: "When Is Parsimony a Virtue?")

I think PE is an overly flexible theory. The notion of a “best explanation” in particular is vague and very open to interpretation. Also, intuitions about what counts as a “best explanation” are not themselves entirely independent of intuitions about which beliefs are justified. This makes it so that PE could accommodate many different sets of intuitions about which beliefs are justified, by adjusting what you consider “best”. That, in turn, means that PE is less strongly supported by its ability to accommodate epistemic intuitions than would be the case for a less flexible theory (such as PC).

4. Conclusion

PC > PE. The problems for PC have been greatly overstated, and PE faces more serious problems. There is no need to supplement PC with an appeal to the notion of explanation.