How to Wield Ockham's Razor (A Dialogue and Commentary)

Unius: ...another benefit of physicalism is that it is more parsimonious than dualism.

Dualis: So what if it is parsimonious? It denies something that is undeniable: the existence of conscious experience.

Unius: Seriously? How many times do I have to say that I am not denying that we have conscious experience? I am just saying that it does not require an autonomous realm of entities to be postulated in order to explain it. Our awareness of the world is physical—just like everything else!

Dualis: Yeah, okay, you say that, but...you know what, whatever, we've had this debate a thousand times. Let's talk about something else. What were you saying about parsimony?

Unius: Okay, sure. All I meant was: putting our explanatory differences aside, would you not admit, at least, that considerations of parsimony count in favour of physicalism?

Dualis: What do you mean? I don't care if something is parsimonious if it denies something fundamental.

Unius: I know, I know, but is it not at least regrettable that you must postulate two kinds of fundamental entity, rather than one? Should we not prefer, if we know absolutely nothing else about two theories, the simpler one?

Dualis: I suppose I kind of see what you mean. However, if I did admit that it was 'regrettable' that I cannot postulate one kind of entity instead of two, it would only be in some very abstract sense. 

Unius: What do you mean?

Dualis: Well, we both agree there must be physical entities, right?

Unius: Sure.

Dualis: Our difference is that I also think that conscious experience is a phenomenon that cannot be denied. 

Unius: I don't disagree with that, conscious experience is just not what you think it is.

Dualis: So our difference is that I think that the fact of conscious experience requires us to postulate a different kind of entity in order to explain it, and you don't? 

Unius: That seems right.

Dualis: But then, to go back to your question, it makes no sense to say that I 'regret' that I have to postulate another kind of fundamental entity—because I think it must be postulated to explain something, consciousness. It would be 'regrettable' if we were in a world that did not have conscious experience as part of it, but that is trivial because there would be nothing there for me to explain. And I would regret it not because it is less parsimonious but because in such a world I would see no need to postulate it, no reason. This is just as you, contra me, see no need to postulate it in this world.

Unius: But think of all the problems with postulating two kinds of fundamental entity, rather than one! How are these entities supposed to interact? What is the relationship between the 'mental' objects of our perception, and the physical objects that are related or correspond to them? And if the mental is not there in the most fundamental entities, but only emerges at a certain level of physical complexity, how is it supposed to 'emerge out of' the physical?

Dualis: Woah, woah, woah. Look, I understand that there are problems with dualism, but none of the points you raise have anything to do with parsimony. Sure, they have to do with postulating two specified kinds of entity, rather than one, and they pose a strong explanatory burden upon those who do so. But the problems are not there in virtue of the position's ontological profligacy, the number of kinds of entities it postulates. Rather, they are there by virtue of the nature of the kinds of entities postulated. No dualist will deny that there are notorious difficulties in reconciling the two.

Unius: I don't know, it sounds like you are admitting the point. One kind of entity faces less explanatory burden than the other. Hence we should prefer a theory that postulates less fundamental entities because of the problems with their multiplication.

Dulias: I understand what you just said, and I even agree with you. But that is a different point to the one you gave above. Earlier you said that parsimony by itself and independent of the content of the theory, independent of the perceived phenomena in need of explanation, and therefore independent of the difficulties of reconciling the co-existence of different kinds of fundamental entities, counts in favour of one theory. Now you are saying that there are some real difficulties in reconciling two specific kinds of fundamental entity with each other, given their nature. Once again, this is something no dualist would deny. 

Unius: And you don't regret these difficulties?

Dualis: Are you even listening to me? What I ought to regret is relative to what phenomena I think is in need of explanation, to my explanatory paradigm. Any regrets I should feel about dualism's difficulties are silenced by the fact that I think I must postulate mental entities. Any regret I feel would be irrational in that regard.

Unius: I see what you are getting at, but I am not convinced that there is not some intrinsic connection with a lower count for one's ontological kinds and true metaphysical system, even abstracted from a particular theory or explanatory paradigm.

Dualis: Maybe if I put it like this: Suppose I say to you that you have two theories to choose from. On the first theory, only the physical exists. On the second theory, nothing exists. And I don't mean (the) nothing, or 'the void', or anything like that. I mean literally that there is no existence whatsoever. What theory do you prefer?

Unius: This is getting a little facetious.

Dualis: Seriously. What theory would you prefer.

Unius: Obviously I prefer the first.

Dualis: Why?

Unius:...Because something obviously exists, whatever that existence happens to be. We cannot choose the second theory because it does not explain that fact.

Dualis: But would you agree that the second theory is more parsimonious than the first?

Unius: I guess so.

Dualis: But doesn't parsimony count for something? Is it not regrettable that you couldn't say there is nothing instead of something?

Unius: I see. I suppose what you are saying is that invoking parsimony here is a little ridiculous, given we could not be having this conversation if there were nothing at all.

Dualis: Indeed. But more precisely: explanatory necessity silences parsimony as an ontological consideration. It is not just another thing to 'weigh up' against things, it is utterly relative to one's explanatory paradigm. What it is like for parsimony to be invoked here, against the view that something exists rather than nothing, is what it feels for me when you invoke parsimony against dualism. It just does not make sense to wield Ockham's razor against those that reject your presuppositions about what counts as whether some entity is beyond (explanatory) necessity.

Unius: I think I might agree with you then. Though what do you mean by 'explanatory paradigm'? We've been saying it but I'm not really clear about what you mean.

Dualis: Just what the intuitive idea would suggest. From your explanatory perspective it makes sense for you to think that the mental can be shaved off of our ontology by Ockham's razor, considerations of parsimony, only because you do not think it is actually doing any explanatory work. But that is precisely our disagreement: whether it does explanatory work. I think it does. That is why it cannot be a consideration for me. In this sense, we have a different explanatory paradigm.

Unius: Right. Could you be a little more precise?

Dualis: Sure. Suppose that physicalism is one theory T1 and dualism is another T2. From your perspective T2 just is T1 plus another kind of entity, call it X. Therefore, T2 = T1 + X. Your view about X, as an entity, is that it is explanatorily superfluous. You see no reason to postulate it, and you therefore should not do so (given what you believe). Therefore, you have a straightforward application of Ockham's razor to the two views. My view about X, as an entity, is that it is explanatorily necessary. We must postulate it if we are to explain the phenomenon of the world. Therefore, I cannot apply Ockham's razor to the two theories. Here is the key though: how do you think we are to resolve this dispute?

Unius: Well, it can't be solved through Ockham's razor, by an appeal to parsimony,

Dualis: And why not? 

Unius: I follow you. Because we each have a different view about the dispensibility of entity X. Therefore, it is that we should be arguing about, the dispensibility of some entity and it's role in explaining some phenomena.

Dualis: Exactly.

Unius: I mean, I suppose the lesson is basically that we should just go back to arguing directly about the actual issue about what explanations are necessary and what are not. In fact, the way you just outlined your version of Ockham's razor is implcitly what we were already doing, if only implicitly: keeping things that are necessary and getting rid of things that are unnecessary, then going on to argue with others about the scope of that (explanatory) necessity. In fact, Ockham's razor seems already to be expressed this way by its most common rendering: "do not multiply entities beyond necessity", as long as "beyond necessity" is interpreted as "beyond explanatory necessity."

Dualis: Yeah, I think that's right. And I think that's all metaphysics really is, or ever was, or perhaps just should be.

Unius: So...I guess that leaves us back where we started.

Dualis: It does, but at least we know now we don't have to waste our time talking about parsimony and Ockham's razor again.

Unius: True that. Did you see Chalmers' new argument for panprotopsychism? He calls it the 'Twombie Schwampman.'

Dualis: Incredible.

 I. Ockham's razor is an oft-cited methodological principle that is supposed to prejudice our ontological theorising in favour of parsimony, rather than profligacy. The general idea is that we ought to cut superfluous entities out of our ontology if they are not needed. What is controversial about the idea is (1) what exactly the principle is really saying and (2) whether or when we are justified in using it. Largely putting aside others' views on the matter, I intend to put forward what I think the principle should say, and in what situations it is justified.

The most common formulation of the razor (if you have heard of the idea this is likely what you heard) is the following:

(OR1) Entities are not to be multiplied beyond necessity

The intuitive idea seems right: if it is not needed, get rid of it. However, by itself, it lacks real guidance. What speaks to whether something is necessary or not? More on this soon. Right now, we need something a little more specific. A more precise formulation of Ockham's razor, an attempt to answer (1), can be found in the Stanford Encyclopedia entry on simplicity in metaphysics. It goes as follows:

(OR2) Other things being equal, if T₁ is more ontologically parsimonious than T₂ then it is rational to prefer T₁ to T₂.   

Once again, the idea is simple, and this version offers more concrete guidance. Given the choice between two theories that you are deciding upon, you should prefer the theory with fewer entities. This is a good first pass at (1), but we now have two problems.

First, we need to specify what it means for two theories to be equal in all things other than the number of entities it postulates. Put simply: what does "other things being equal" actually mean? Second, it could rely, on some readings, on a dubious implicit premise. Namely, that some theory being simple is rationally preferable merely by virtue of it counting fewer entities. But why should we think that just because some theory has fewer entities it is rational to prefer it? Suppose we have two theories A and B. A postulates one type of fundamental entity and B postulates two. By itself, the number of entities they countenance tells us absolutely nothing about whether they are true. What is needed then, on top of this formulation of what the razor says, is an account of when (or whether) we are justified in appealing to it. Further, we must give an account without erroneously assuming that something is more likely to be true, just because it has fewer entities. While some philosophers think that this is actually what Ockham's razor is saying, that simpler theories have a higher intrinsic probability by virtue of their simplicity, this is something I reject.

II. My own view is that we can accept something like the formulation given by (OR2) only if we provide very specific meaning to "other things being equal." Done in the right way, we can also avoid assuming that simpler theories are more likely to be true merely by virtue of their simplicity. 

My proposal is the following. We should read "other things being equal" strictly, as indicating that both theories explain the exact same phenomena, in the exact same way, but that the latter theory posits further entities that do not explain any further phenomena. That is to say, in situations where T2 is just T1 + X, and X is some entity or type of entity playing no explanatory role, it is rational to prefer T1. Thus, "other things being equal" means "explanatorily equivalent." While this is a rather narrow scope for the principle, I think this is its proper place. This is because I do not think some theory is automatically superior to some other theory just because it has fewer entities. What matters is whether those entities are doing explanatory work. Ontological profligacy is entirely justified insofar as that profligacy is taken to be necessary to explain some feature of the world, some phenomenon.

Thus, a more literal rendering of what I think the razor should be is as follows:

(OR3) Entities are not to be postulated to exist (multiplied) if they are not necessary to explain anything (beyond necessity).

This simply means that, given the choice between two theories that are exactly the same, except that one posits some further entity that plays no explanatory role, we should accept the theory that does not posit these further entities. It says that we should not postulate entities unnecessarily, that is, for no reason. If an entity is postulated for no reason, we should not accept that it exists.

III. We can see the appeal of such a principle with an example. Suppose Lachlan smashes a window. This is an event. Now suppose I argued that the natural explanation we would give of the window being smashed was all correct. Lachlan kicked a ball and it hitting the window was sufficient to break it. Now suppose I argue that simultaneous with the football hitting the window, a God decided to smash the window using telekinesis. No explanatory demand is fulfilled by such an appeal because everything would have happened as it did without God, especially given I am not replacing the causal power of the ball as the thing smashing the window with God smashing the window but instead supplementing it with God. In this case, I would surely be wrong to postulate further entities (this God) because no phenomena are accounted for that were not accounted for before. In other words, I am postulating an entity for no explanatory reason. I am just arbitrarily supposing there is something else there. It is arbitrary because I could postulate any number of other entities or mechanisms to have caused this event, even though they would not further explain anything.

Arbitrariness is objectionable in metaphysics because it amounts to arguing that we should postulate entities for no reason, which would trivialise the whole endeavour. Suppose we were to ask someone who postulates such a God in the above situation, or any other entity doing no explanatory work, why they do so. Given they play no explanatory role, could a satisfactory reason be given? They cannot say “because it explains so and so phenomena”, because so and so phenomena have already been explained by their ontology. They cannot say “because it is useful to suppose that it exists” because just because something is useful does not mean it exists. And surely, if one cannot give a reason for why we should accept a posit, then we should not accept it.

If some stranger came up to you and asked if you could give them $100, and when you asked why, they reply “oh, for no reason” (and you somehow knew this was true), should the onus be on you to give a reason why you should not give them the money? No, the onus is on them. And if they proffer no reason (and you come up with no reason yourself), then you have no reason to give them money. The same burden holds for arbitrary metaphysics. If someone postulates some entity, the burden is on them to show why it is necessary to explain some phenomenon. Denying this principle would amount to arguing that we should postulate entities for no reason, which surely trivialises the whole endeavour because then, anything goes.

To summarise: we are justified in applying Ockham's razor when choosing between theories that are explanatorily equivalent but one postulates entities that explain nothing further. We are justified in eliminating entities that play no explanatory role, or, we are rational to choose the theory that lacks those entities, because we have no positive reason to accept them. If there is no reason to accept that some entity exists, then we should not do so. 

This is not to say I am not open to indispensability arguments, arguments that are trying to meet some explanatory burden, or transcendental arguments for necessary metaphysical structures. These are of a different structure to the kind of situations I am considering here. These kinds of arguments do give a positive reason why they are to be taken seriously. For example, Plato thought that his Forms made possible what he took to be the undeniable intelligibility of worldly entities. Heidegger thinks that a transcendental argument can be made from the revealing (individuation) of intelligible individual beings to the existence of the nothing. I think a transcendental argument can be made for the existence of a real duration to explain something I take to be undeniable: the perception of change. Thus, another way of thinking about what I am saying when I say that entities should not be multiplied beyond explanatory necessity is to think about its negation: entities should be postulated according to explanatory necessity.

IV. Why should this be the approach we take with regard to Ockham's razor, other than avoiding the kind of arbitrariness I point out above? Is it not kind of narrow? The main reason why we should endorse this kind of parsimony is that it pushes debates back to the substantive issue at hand, rather than to a kind of meta-ontological point-scoring, typical of analytic metaphysics. Indeed, the reason we should take this narrow approach is that virtually any other application of Ockham's razor will always be an objectionable kind of meta-ontological point scoring. What do I mean by this? I can show this with an example.

Take the debate between physicalists and dualists. It is sometimes said that one of the theoretical advantages of physicalism is that it is more parsimonious than dualism. This is because physicalists think that there is only one type of fundamental entity, the physical, and dualists think there are two types of fundamental entity, the physical and the mental. In other words, it is sometimes said that physicalism countenancing fewer types of entities than dualism is a reason, by itself, to prefer it to dualism. However, why should we think that the number of types of entities by itself tells us about whether some theory is true? The example in the above dialogue, of a theory that postulates nothing versus one that postulates something, is supposed to show this. If we are not given a further reason for why each type of entity is supposed to exist, then the number, by itself, provides nothing. Couched purely in terms of parsimony, the debate cannot be made sense of from this abstract position. Couched in terms of explanatory necessity, however, we can make perfect sense of it.

The reason dualists postulate two types of entity is to explain two types of phenomena that we observe and which are in need of explanation: the 'outside' world, which seems to have shape, structure, and dimensionality, and our experience of that world, which seems to be of a different kind, something to which it appears to be a category error to ask: "what is the width and depth of your perception right now?" Not only this, the dualist thinks it is virtually impossible to deny that there exists this thing, consciousness, that is something above and beyond the physical world. This means that, for a dualist, the fact that physicalists do not accept the mental as something existing is a decisive reason against preferring physicalism. And so it should be if you accept their views about what is in need of explanation. The relative parsimony of the alternative is utterly silent in the face of the fact that, according to dualists, physicalism fails to explain something we cannot deny, the reality of the mental. In this situation, it is not merely that the dualist 'weighed up' considerations of explanatory adequacy and parsimony, finding the former to outweigh the latter. Rather, the dualist thinks that the more parsimonious theory fails to live up to an explanatory task, which ultimately silences its parsimony as a real consideration for it. That is to say, it does not make sense to apply Ockham's razor against the mental, for the dualist, because its multiplication of entities is not beyond necessity.

Of course, the physicalist will disagree that there are both types of entity. They will say that the seeming discrepancy between these phenomena, the external world and our experience of it, is illusory. Both the outside world and our experience of it are fundamentally the same kind of thing. They are both physical. However, they will not say this because their view is more parsimonious, but because they think that postulating an extra class of entities to explain the mental is explanatorily superfluous. The mental can be fully explained by recourse to the physical. They will attempt to show this with arguments independent of parsimony because parsimony by itself cannot show this. This means the choice between theories for a physicalist is then straightforwardly a case where Ockham's razor, as formulated above, applies. Dualism is just physicalism plus X, where X is doing no explanatory work. However, its use in no way rests on the relative parsimony between theories, but on the explanatory necessity of postulating them.

We can now see how this pushes us back to the actual debate, rather than towards meta-ontological point scoring, because at no point does either party appeal to parsimony to argue for their theory. We can see this by paying careful attention to where the disagreement actually lies for each side of the debate. What they disagree about is whether some phenomenon must be explained by the postulation of another kind of entity. In this case, whether conscious experience is something that must be explained by the postulation of non-physical entities. Once this is settled (if it ever were), then we know whether we can apply Ockham's razor or not (where the physicalist thinks we can and the dualist thinks we cannot). But for this to be settled, then arguments must be made for or against the explanatory necessity of postulating it, not at the abstract level of parsimony. In other words, to know whether Ockham's razor applies to a choice between theories, one must simply argue about the tenability of the theories themselves, not the meta-ontological 'virtue' of parsimony. In this case, the fact that physicalism is more parsimonious than dualism is completely silent to the dualist because it refuses to explain something that requires it.

It might be objected that counting 2 types of entity is somehow 'more arbitrary' than 1. If we can accept 2 types of entity, then why are there not 3 or 4 types? And so on. But this will not work. First, such an argument no longer has to do with parsimony, it has to do with arbitrariness. Second, if we are evaluating a theory purely on the number of types of fundamental entities it accepts, 1 type of entity is exactly as arbitrary as any other number. What makes a certain number of types of entity non-arbitrary is the explanatory role they play. Peirce postulated 3 fundamental categories of being and he had an explanation for why each was necessary. Just because there are 3 categories rather than 1 or 2 does not license the use of Ockham's razor against his view, nor make it arbitrary. Rather, what matters is whether or not he was right about their necessity, whether or not his arguments for it are sound. The same is true for Plotinus' 4 ascending categories of being. As arbitrary as 4 may seem, and as wrong as he may be about them, each layer explains some phenomena and is argued for in that regard. Ockham's razor does not automatically cut against the system absent a response to those arguments. Nor, in fact, does arbitrariness, because the whole point of any system is that its constituents are non-arbitrary. 

We can also see how virtually any other application of the razor is the objectionable kind of meta-ontological point scoring I complain about above. This is because there are only two options for some proposed entity: either it is deemed explanatorily necessary, or it is not. If it is, then it cannot be eliminated by considerations of parsimony, and if it is not, then there is no reason to postulate it. In the former case, Ockham's razor does not apply. In the latter, it is the fact that there is no reason to postulate the entity in one's theorising, not that not doing so would make it more parsimonious, that makes the theory without it preferable. In both cases mere simplicity is irrelevant.

V. What this points to is that there is a certain relativity to the use of Ockham's razor, as I have proposed it. Whether it applies in a particular case for a particular philosopher will be relative to what phenomena one thinks is in need of some special explanatory posit. Suppose we call this one's 'explanatory paradigm', which describes (1) those things one thinks are in need of explanation, and of those things, (2) which of them require postulating entities to explain them. Ockham's razor is therefore the principle that says you should not postulate entities beyond what your explanatory paradigm circumscribes. Disagreements about whether something should fall within our explanatory paradigm cannot be resolved at the level of the count of those things, but only at the level of settling the explanatory paradigm itself, at the level of first deciding on, or arguing about, (1) and (2). We discovered this through our analysis of the debate between physicalists and dualists, and I would argue it applies to all other metaphysical debates. It is here we encounter what is the ultimate task of metaphysics: to determine what needs to be explained, how to explain it, and what counts as an explanation for it. A proper appreciation of this version of Ockham's razor orients us towards this ultimate task.