Monday, December 23, 2013

RIP Peter Geach

His daughter Tamsin Geach has reported that Peter Geach has passed away Saturday. He was 91 years old. In his long and fruitful career Geach had made a number of important contributions to the philosophies of logic, language, and mind, and to the interpretation of Frege, Wittgenstein and Aquinas.

Geach is now probably best known for providing the canonical statement of the embedding problem: for theories that deny that certain classes of statements are factual (capable of being true or false), like expressivism says about moral claims, it is a problem that these statements can be embedded within logical inferences. The expressivist understands 'don't torture the cat' as something like 'boo to torturing the cat!'. But you can make an argument like 'if it is wrong to torture the cat, it is wrong to get someone else to torture the cat; since it's wrong to torture the cat, it's wrong to get someone else to torture the cat'. This looks for all the world like a straightforward inference, the but expressivist can't say this, because you need truth-apt statements to make an argument, and they deny that claims like 'it is wrong to torture the cat' could be true or false. So, the expressivist can't explain a large amount of things that we normally say without any problem.

Geach was married to the late Elizabeth Anscombe, and together the two of them did as much as anybody else to improve our understanding of propositions and propositional attitudes, and to flesh out our understanding of intelligent actions. They both were converts to Catholicism and tirelessly worked within its intellectual tradition, trying to bring the work of especially Thomas Aquinas to modern attention and prominence. Geach was honoured by the Pope for his work.

I'd like to suggest to you a characteristically short and incisive paper of his, 'Good and Evil', wherein Geach introduced a distinction which still continues to bear fruit in meta-ethics and value theory. Geach asks us to distinguish *attributive* from *predicative* adjectives, with predicative adjectives like 'yellow' and 'round' being ones where 'something is an x y' means it is an x and it is a y (a yellow apple is both an apple and yellow), whereas this is not true for attributive predicates like 'large' and 'tall'--a large flea is not large. Geach suggests that 'good' and similar evaluative adjectives are attributive and not predicative. This would mean that analyses like those of GE Moore miss their target ('yellow' was the adjective Moore compared 'good' to), and that to make an evaluation of something means that you are comparing it with some appropriate class of things (a large flea is large compared to other fleas, but not as compared to chairs, tables, dogs, humans, boats, houses, and so on).

Tuesday, November 26, 2013

Why you shouldn't read usage advice (with examples)

Helen Sword is someone at my university who coaches academic writing, and has written a large number of articles and a couple of books on the subject. She has recently released another book, which has prompted a few discussions I've had with colleagues where I warn them off work like hers. I have not read her latest book, and will only do so if compelled to. I have read some of her earlier work and dealt with her in person. I can only report that much of her recommendations are badly informed and so lacking in merit that she herself fails to follow her own advice, even as she is giving it. Sword is in no way alone in this--she is part of an industry of people giving usage advice which nobody should pay attention to. This includes the two most widely-cited examples of the form: George Orwell in 'Politics and the English Language' (the most famous, but not earliest, piece telling you to avoid writing in the passive voice) and William Strunk Jr. and E.B. White in the most famous usage manual of them all, The Elements of Style. Despite the respect these works garner, the characterisations they give of good or bad writing are often demonstrably false, and often enough these characterisations are false even of their own writing. The linguist Geoffrey Pullum is one person who has embarked upon a crusade on this kind of bad usage advice in various venues, a crusade of which this post should be seen as a small skirmish. It is an extremely interesting fact that people can be so radically wrong about the rules they follow, even when they correctly follow those rules (something I call second-order error). This fact forms part of my own research and which is part of why I have so much to say about it--our ignorance about our own norms of language use is a central example for me. But for now I'll only talk about how a lot of this advice, including Sword's, is plainly wrong in their descriptions of good and bad writing.

Sword made her name in the wider consciousness with a piece in the New York Times on why nominalisations (making nouns out of verbs) make for bad writing. Her analysis on this point is not just wrong, but disastrously wrong. Let me explain my very harsh judgement. As computational linguist Mark Liberman points out, she is not following her own advice: 80% of the nouns she uses in the final paragraph of the piece, where she offers a way for you to strip your writing of nominalisations, are themselves nominalisations. She can't claim to avoid recent nominalisations, and has the good sense not to do so: firstly, because such advice would be too daft for anybody to take seriously (it would be something like 'nominalisations are fine except if you know their etymology'); and secondly, because she used a number that are transparently nominalisations even to modern readers ('test', 'attack'). So, her advice is so bad that she doesn't herself follow it. Further, her writing is good enough to pass under her eyes and those of her editor and still not be rejected (and mine--she writes perfectly well). So, her analysis is descriptively inadequate--her writing is full of nominalisations, and is no worse off for it. So, her description of good writing as avoiding nominalisations is simply false (it turns out that nobody avoids nominalisations), and even if we take what she says as a revisionist prescription (something like 'never mind that everybody uses nominalisations, you shouldn't'), she herself doesn't seem to take it seriously, because she is doing what she tells us not to even as she is telling us. I don't know how to describe this except as 'disastrously wrong'. Other ventures of hers, like her 'Writer's Diet Test', fare no better when given serious scrutiny. Though she will of course also be able to give you good advice, as any writer of her level of competence would be able to, she wouldn't be able to distinguish her good advice from her bad advice. Thus, like almost every example of the form, her usage manuals are not to be taken seriously.

Now, onto the more general reasons. Nominalisations, like almost every other bugbear of usage manuals, are a perfectly normal feature of English. In English nominalisation is a productive way to say things: clauses with nouns put into verb-forms in the verb's place are something every competent speaker of the language understands. Trying to stigmatise a perfectly normal part of the language is at best going to be deeply revisionist, if not entirely wrong-headed, and is likely to be unmotivated. If what we are doing are looking to examples of writing to imitate, then you need to respect the fact that the instances of actual writing you identify will also be using the actual rules of the language they are written in. For English this means that they will use nominalisations, the passive voice, split infinitives, etc., since these are all features of English. And they are perfectly fine features! Everybody uses them, even in the most widely-admired prose. To use a different example, take the passive voice construction. A string of people, most prominently George Orwell in 'Politics and the English Language' and Strunk and White in their The Elements of Style, tell you to avoid the passive voice in writing. Sword does so as well, in her work and in person. But they are in a similar position as Sword on nominalisations: the advice is nonsense, and they don't follow it themselves. In a revealing irony, Orwell, Strunk and White all use the passive more than the average English writer. The average rate of passive voice constructions in written English tops out at around 13% (depending on which corpus you check), whereas in the same essay Orwell tells you to avoid it he uses it in about 20% of possible cases, and E.B. White consistently uses it in between 20% and 30% of cases. So much for their advice--nobody should take it seriously. Orwell and White are, of course, two of the finest prose writers of their time, and we would do well to imitate them. But they are mistaken about why the style they admired was good, and other examples bad.

I have not yet said a word about the tone in which Sword delivers her message. I'll restrict myself to saying that it is obnoxious to attempt to make people anxious about a perfectly normal part of their language use, just as it is obnoxious to try and make people anxious about perfectly normal and harmless things they would normally do. In this respect, Sword is in the same boat as Orwell, Strunk and White, and a whole industry of people who act from an assumed authority that they can't but fail to undermine.

Of course, a lot of writing is bad, and people like Sword and the stream of other advice-givers she is part of are trying to improve it. They may be trying to help in the best way they know how (but, given how easy it is to show how bad their descriptions of good and bad writing is, they show an unseemly unwillingness to look at their own claims critically). But the fact that there is a problem to be addressed doesn't mean that their solution is correct. Their solutions are often radically mistaken. The examples I've discussed here are demonstrably mistaken.

I have a lot to say on this topic, because the phenomenon of mistakenly identifying the rules you are successfully acting from is of great interest for my own research. Orwell and White are, as I've said, excellent writers. They are able to reliably and regularly write in a clear and engaging style. Writing is a rule-governed activity. So they are good at this rule-governed activity. But they are also radically mistaken about what rules they are following, even when they are following them. This is very surprising, and deserves further investigation.

It is also worth noting that I link a lot to Language Log, because as a popular blog by top-class linguists it is an extremely valuable resource for the type of thing I'm doing here. If you have any interest in linguistics and language-use, I recommend it unreservedly.

Thursday, October 24, 2013

Why Kant's moral theory takes account of the consequences of actions

This is a sketch of a response to a common misconception about Kant, which I wrote for a discussion elsewhere.
I think it's false that Kant's theory doesn't consider the consequences of an action.What Kant's view doesn't do is make the rightness or wrongness of any act depend upon what consequences it brings about. To make sense of this, we need to ask a series of questions.
  1. Is the only way to take account of consequences be to make the rightness or not of an act depend directly on what its consequences are? Probably not. You can take account of something without letting it have a veto (as it were) over the rightness of an action.
  2. What place could consequences have in Kant's theory? The consequences will come from, amongst other things, the generalisable maxims Kant wants you to evaluate using the categorical imperative. A certain agent adopting a certain maxim will have certain consequences.
  3. Since the maxims are only part of the causes of consequences, what of the other parts? Kant takes pains at various points to argue that there are limits to the things people are responsible for--in particular, they aren't responsible for accidents (neither accidentally good or accidentally bad things). This means that some consequences would be salient for evaluating the maxims (on Kant's view) only if they are non-accidental in some appropriate sense. At the very least, this means the consequences of a maxim has to be regular and predictable, otherwise it wouldn't even be intelligible for someone to accept or discard that maxim with those consequences in mind.
  4. Would any maxim which, if generally accepted, lead to predominantly bad consequences be the maxim agents possessing good will are going to assent to? Probably not, or not often, because good will is supposed to (at least in part) consist of treating individuals as ends, and predominantly bad consequences are predominantly bad for someone, so the maxims leading to those bad consequences would be pretty unlikely to be treating everybody with the appropriate respect.
So, by the time we've answered 4, we see that any maxim which would non-accidentally lead to predominantly bad consequences to some people are unlikely to be adopted by the categorical imperative. There's a lot here to be made precise: what the threshold is where consequences are too bad too endorse the maxim that leads to them; how to measure that threshold within Kant's theory; what exactly we mean with 'would non-accidentally lead'; and so on. But, however you answer those questions, there is going to be some way in which the consequences are considered by the categorical imperative.
This is rough and ready reasoning at the moment, and making it precise is difficult (not least of all because how to understand the categorical imperative is contested), but there's something to it.