On the Use and Abuse of Occam's Razor

I am not historian enough to pronounce upon the relation of what is standardly called Occam's Razor to the writings of the 14th century William of Ockham. The different spellings of his name will serve as a reminder to be careful about reading contemporary concerns into the works of philosophers long dead. Setting aside historical concerns, Occam's Razor is standardly taken to be a principle of theoretical economy or  parsimony that states:

   OR. Do not multiply entities beyond necessity.

It is sometimes formulated in Latin: Entia non sunt multiplicanda praeter necessitatem. The principle is presumably to be interpreted qualitatively rather than quantitatively, thus:

   OR*. Do not multiply TYPES of entity beyond necessity.

It is not individual entities that are not to be multiplied, but types or kinds or categories of entity.  To illustrate.  Some criticized David Lewis' extreme modal realism on the ground that it proliferates concreta: there are not only all the actual  concreta , there are all those merely possible ones as well.  He responded quite plausibly to the proliferation charge by pointing out that the Razor applies to categories of entity, not individual entities, and that category-wise his ontology is sparse indeed.

'Multiply' is a picturesque way of saying posit. (Obviously, there are as many categories of entity as there are, and one cannot cause them to 'multiply.')  And let's not forget the crucial qualification: beyond necessity.  That means: beyond what is needed for purposes of adequate explanation of the data that are to be explained.  Hence:

OR**  Do not posit types of entity in excess of what is needed for purposes of explanation.

So the principle enjoins us to refrain from positing more types of entity than we need to explain the phenomena that need to be explained. It is obvious that (OR**) does not tell us to prefer theory T1 over theory T2 if T1 posits fewer types of entity than T2. What it tells us is to prefer T1 over T2 if T1 posits fewer types of entity AND accounts adequately for all  the data. So there is a trade-off between positing and accounting.

Our old pal London Ed sometimes seems to be unaware of this.  He seems to think that simply brandishing the Razor suffices to refute a theory.  Together with this he sometimes displays a tendency to think that whole categories of entity can be as it were  shamed out of existence by labeling them 'queer.'  I picked up that word from him.  A nice, arch, donnish epithet.  But that is just name-calling, a tactic best left to ideologues. 

What is offensive about Razor brandishing is the apparent ignorance on the part of some brandishers of the fact that we all agree that one ought not posit types of entity in excess of the needs of explanation. What we don't agree on, however, is whether or not a given class of entities is needed for explanatory purposes.  That is where the interesting questions and the real disagreements lie. 

The Razor is a purely methodological principle.  It does not dictate any particular ontology.  Taken as such, and apart from its association with the nominalist Ockham, it does not favor nominalism (the view that everything is a particular) over realism (the view that there are both particulars and universals).  It does not favor any ontology over any other. 

Nor does it rule out so-called 'abstract objects' such as Fregean propositions.  I gave an argument a while back (1 August 2010 to be precise) to the conclusion that there cannot, as a matter of metaphysical necessity, be nothing at all, that there must be at least one abstract object, a proposition.  A fellow philosopher commented on that post, Thinking about Nothing, and made the objection that I was multipying entities.  But again, the salient question is whether the entity-positing is necessary for explanatory purposes.  If my argument was a good one, then it was.  One cannot refute such an argument simply by claiming that it introduces a type of entity that is less familiar than one's favorite types.

To sum up.  Philosophy is in large part, though not entirely, an explanatory enterprise.  As such it ought to proceed according to the methodological principle formulated above as (OR**).  This principle is not controversial.  Hence it should not be presented to one's opponents as if it were controversial and denied by them.  Nor is it a principle that takes sides on the substantive questions of ontology. 

What I am objecting to is the idea is that by earnest asseverations of a wholly uncontroversial methodological principle one actually  advances the substantive debate.  After all, no one enjoins that we multiply entities beyond necessity.

Comments

[london ed]

>>He seems to think that simply brandishing the Razor suffices to refute a theory.

Why do you think this? In the post I think you are referring to I said "If an inferential semantics is sufficient, then the Razor tells us it is necessary".

I think you took the assertion of the conditional 'if A then B' as an assertion of the antecedent A. It's a very common mistake :)

On the history, the Razor predates Ockham, and was a common methodological principle adopted by scholastic philosophers, including Scotus (who was hardly a nominalist).

On the connection with nominalism, Ockham says that we must not multiply entities (i.e. types of entity) in accordance with the multiplicity of terms. I.e. we shouldn't infer that because a certain term looks like a name for a sort of entity, that there really is such a type of entity. His methodology is to eliminate propositions containing such terms and replace them with others which (according to him) mean the same, but which don't tempt us into supposing the existence of real entities. "Nor do we have to multiply things in such locutions as "a column is to the right by to-the-rightness", "God is creating by creation, is good by goodness, is just by justice, is powerful by power", "an accident inheres by inherence", "a subject is subjected by subjection", "a suitable thing is suitable by suitability", "a chimera is nothing by nothingness", "a blind thing is blind by blindness", " a body is mobile by mobility" and such other innumerable things."

Obviously this is a specific application of the Razor. Show that we can fully explain the meaning of a proposition by positing fewer entities than other explanations. Thus the explanation which posits fewer entities (or types of entities) is to be preferred.