Reviewing Non-Fiction Is Hard

Sometimes I read non-fiction books, and really enjoy them. “What an awesome book,” I’ll think. But it’s actually quite hard to tell if non-fiction books are any good or not. At least, it’s hard to tell just by reading them. I guess you could read a review. But someone has to write the reviews.

The reason it’s hard is that you’ll usually be in one of two situations. Either you’ll be an expert in the topic the book is about, or you won’t be. Suppose you’re not, and so a lot of the stuff in the book is new to you. You don’t know if the book is any good or not, because you don’t know whether the stuff in the book is right or not. You can try factchecking it, but even if you can track down the sources they’ll often be buried in difficult academic writing of a sort you’re not really competent to understand. And if the book you’re reading is any good, a lot of what it’s telling you won’t be checkable facts, but rather a kind of expert insight and analysis that you wouldn’t be able to reconstruct yourself. And of course some books contain original research, in which case they can’t really be checked because in a way they are the source. These three all blur into each other, but they all make it very hard for a non-expert to tell if a non-fiction book is any good just by reading it. (And if we’re being strict about “just by reading it”, you’re not allowed to factcheck it anyway! But let’s not be strict. It's hard in any case.)

A good example of this danger is Daniel Kahneman’s Thinking, Fast and Slow. Kahneman is one of the world’s top psychologists, he did pioneering work on cognitive biases with Amos Tversky, and he won the economics Nobel "for having integrated insights from psychological research into economic science, especially concerning human judgment and decision-making under uncertainty". The book is about cognitive biases and decision-making under uncertainty, and the title refers to two styles of thought, one automatic and not very conscious and the other fairly carefully thought through and more conscious. It tells you about lots of cool little findings from psychological research, like how people are more likely to believe something they read if it’s written more legibly (pp62-3), and it integrates the anecdotes into a general narrative about the different ways people think.

When that book came out, it was very well received by people who weren’t already experts on the topic. I vaguely remember experts being more divided on it, but I’m not an expert and I thought it was awesome. I was entertained by the anecdotes, and I really felt like I was learning something. I told my friends the anecdotes as if we could be confident that they were true, and I recommended the book to them. But since it came out, psychology has had a bit of an existential crisis based on the fact that lots of its little findings don’t replicate. A lot of effects people found may well have been flukes that only seemed representative of how people behave because when psychologists have done experiments and not found anything cool they haven’t told anyone about them. Or they have but nobody has listened, which makes them less likely to bother telling people next time. Kahneman himself is very concerned about the whole thing, and he thinks he was a bit too credulous about some of the stuff in the book.

Now, I don’t want to set Kahneman up as some kind of fall-guy here. It may still be a good book, and the issues with some of the anecdotes may be kind of minor. He’s still a great psychologist and communicator, he was writing in good faith, and his willingness to publicly address problems with his own work is impressive and an example to his colleagues. The point is that I wasn’t competent to judge how good his book was. And to be honest, I still couldn’t tell you.

So, it’s hard to tell if a non-fiction book is good if you’re not already very familiar with the subject. Well, duh! But what if you are an expert? What if you could have written the book yourself? In that case you have a different problem: The Curse Of Knowledge. In his book The Sense of Style, Steven Pinker argues that the reason people often write badly is they struggle to imagine what it’s like not to know things that they in fact do know. You know what you want to say, but your reader has to try to work it out from what you’ve written, and it’s hard to put yourself into their mind and work out whether what you’ve written would still be clear. And if you’re trying to explain something you already understand but they don’t, you have to put yourself into their mind and work out whether they can understand the thing you’re trying to explain on the basis of what you’ve written. That’s hard too. Now, imagine you’re an expert reading a book by another expert on the same thing, and you’re trying to work out whether a non-expert will be able to understand the thing the book is about on the basis of what the book says. It’s not easy to do.

So, who should be reviewing books, if both experts and novices face systematic obstacles? Three suggestions come to mind.

• Someone who is neither an expert nor a novice. What you want is someone who doesn’t already know what the book says, and so doesn’t have the curse of knowledge, but who’s competent enough in the sort of thing the book is about to be able to check if it’s right, once they’ve been told the things the book says. In general it’s often easier to check an answer to a question than find the answer. While I wasn’t competent to check if Kahneman’s book was any good, maybe a psychologist in a different field could have done it.
• A great teacher. The curse of knowledge essentially arises because a certain kind of imaginative exercise is difficult. But some people seem to be quite good at it. Being good at it is part of the skillset of a great teacher: they need to be able to get into the minds of the students and tell whether what they’re saying would communicate the material to someone who wasn’t already familiar with it.
• A novice and an expert working together. The two problems are pretty separate, so in theory the expert ought to be able to read the book to check that what it says is sound, while the novice reads it to see if they feel like they’re learning something. And after they’ve both read it, the novice and the expert can talk to each other so the expert can check that the novice really did learn the things they felt like they were learning.

My favourite one is the last one. While an expert in an adjacent field might be able to do a decent job of factchecking, they won’t do as good a job as an expert, and it’ll be harder for them. Their transferable research skills will also mean they still have to do a bit of difficult imagining to get into the minds of the intended audience. A great teacher might be able to do the job, but we don’t even have enough great teachers to fill all the teaching jobs, let alone all the reviewing jobs as well. This leaves the last option. Unlike great teachers, experts are ten a penny, and novices are twenty a penny. Of course, you do need two people. But it’s still my favourite option, and it’s kind of odd that it practically never happens.

Avicenna vs Turtles

Anglophone metaphysicians, and perhaps some other metaphysicians too, fairly recently started thinking about things in terms of ontological dependence. The idea is that some things depend on other things, or are grounded in those things; some facts are true in virtue of other facts, some things are fundamental while some things depend on the fundamental things, and so on. There’s a whole mess of concepts in the vicinity and we still haven’t sorted it all out, but it seemed like a useful way to think. It still does. I still think like that myself, sometimes.

One question you might ask about this framework is this:

• Must there be a funamental level? Couldn’t it just be turtles all the way down?

Ross Cameron addressed this question in a paper a few years ago (Cameron 2008). He basically came to the conclusion that he couldn’t find anything incoherent about everything being grounded in something further down, but it’s an inelegant set-up and we should expect the world not to be like that. When I’ve come across people citing his paper it has mostly been to agree with his conclusion, although this impression may be unrepresentative or at least out of date. I like the paper too, and I’m pretty sympathetic to his take on the issue: metaphysical systems don’t have to be incoherent to be implausible.

Now, Cameron wasn’t the first person to think about this sort of thing. He mentions some predecessors in the paper, but today I’m going to talk about one he doesn’t mention. About a thousand years ago Avicenna was coming up with a proof of the existence of God, like you do, and his proof involved thinking about something structurally quite similar to the turtles question and coming to the opposite conclusion over whether there has to be something at the bottom. He thought the infinite descent scenario really was incoherent. Since one can count the people cleverer than Avicenna on the fingers of one hand, if he’s challenging our contemporary consensus we should probably take a look at what he had to say.

I’m not an Avicenna scholar, unfortunately. My knowledge of Avicenna comes mostly from a podcast by one Avicenna scholar (Adamson 2010-, especially #138-142) and a book by another (McGinnis 2010). I’ve also read a few passages from Avicenna’s own work in translation, including one on the argument I’m talking about here (McGinnis and Reisman 2007, especially pp214-5). It’s not nothing, but you probably won’t find me trying to turn this blogpost into a paper. (But if you’re an Avicenna scholar and think there’s something here worth tackling properly together, I’m not busy.) You might wonder why I’m writing about it at all, given my incompetence to do the topic justice; it’s basically because it seemed like there was something there and nobody else seemed to be writing about it. Maybe that’s because the medieval-philosophy-in-Arabic crowd and the contemporary-analytic-metaphysics crowd don’t overlap much. Anyway, if you think I’m talking rubbish but not such irredeemable rubbish as not to be worth engaging with, we can have a discussion in the comments or wherever and try to understand the issue better.

So, here’s Avicenna’s argument as I understand it.

1. He’s got a distinction between things which are necessary through (or in) themselves and things which are necessary through another. He’s not understanding necessity the same way philosophers do nowadays - possible worlds and all that - but I wouldn’t like to try to explain exactly how he is understanding it. It seems at least to be structurally a bit like something related to our contemporary concept of ontological dependence, though.
2. He thinks that everything is either one or the other, and nothing is both. In terms of ontological dependence, you can think about things that depend on something else, and things that don’t.
3. He wants to show that at least one thing must be necessary in itself. He’s going to go on to argue that various things follow from something being necessary in itself, and that there can only be one such thing, and he’s going to say that this necessary existent is God. But for the moment he just wants to show that there’s at least one.
4. We know that something exists. Look around yourself, look within yourself, whatever. Something exists.
5. Now gather everything together that is necessary through another. If there aren’t any such things, then since something exists, something must be necessary in itself. But if there are, then gather them together into one big object. (We should resist the temptation to call this object The Great Mumkin.)
6. Is this object necessary in itself, or through another? If it’s necessary in itself, then we’re done. If not, then what is it necessary through? (I’ve seen this step presented in different ways, and Avicenna may have presented it himself in different ways. He actually thinks it’d be absurd for the sum of all things necessary through another to be necessary in itself, but it’s worth noting that we don’t have to agree about that for his argument to work.)
7. If it’s necessary through a part of itself, that’s either absurd or counts as the part being necessary through itself. (I’m not really sure how this step goes, and it seems to me that it’s where most of the action is.)
8. If it’s necessary through something that isn’t part of it, then that thing must be necessary in itself, because everything else is a part of it.
9. So whichever option we go for, something is necessary in itself.

Now, I expect I’ve garbled the argument somewhere. The bit where we say that things can’t be necessary through a part of themselves seems especially dodgy. Here’s a worry: take the sum of everything that exists. Is that necessary in itself, or through another? If anything besides God exists, then it can’t be necessary in itself, at least according to Avicenna, because he` thinks only one thing is, and that thing is God and doesn’t have proper parts. But if it’s necessary through another, then that other thing must be a part of it, because everything is. But that’s not supposed to be allowed. So I don’t really get what’s going on there. (I guess he could say that in this case we choose the option at step 7 of the part being necessary in itself. But I do think I’m missing something here.)

One way of patching this is to say that everything is either necessary in itself, or is necessary through another, or is the sum of something necessary in itself and something necessary through another. (I’m using a pretty classical-mereology framework, because Avicenna seems to be. To call Avicenna a classical mereologist would be anachronistic by 900 years or so, but the argument helps itself to principles that are accepted by classical mereology but rejected in some non-classical mereologies. If classical mereology rules out turtles all the way down, that’s interesting in itself. Investigating which mereological principles are essential to the argument and which aren’t would be interesting too, and if the argument has something in it then it’s something we should do.) If we make this assumption, the argument still goes through much the same. You just include the mixed option in step six, and note that the mixed option also entails that something is necessary in itself. The sum of all things would be the sum of the Necessary Existent, which is necessary in itself, and Creation, which is necessary through another, viz. the Necessary Existent.

With different mereological assumptions you might also try constructing the sum-of-all-dependents object by removing everything necessary in itself and taking what’s left. That relies on a complementation principle you might want to reject (but which classical mereology accepts), but it’s worth noting the option. That object might not contain all dependents - if there’s a dependent object that is part of a necessary object, for example - but if it depends on something outside itself then something is outside it, and so something is necessary in itself because everything outside it is part of the sum of things necessary in themselves.

Anyway, let’s adapt Avicenna’s argument to the question of whether everything might be dependent on something else. We’ll call things that are dependent on something else dependent, and other things independent. We’ll allow mixed cases, and assume classical mereology (though not atomicity - there could be gunky things all of whose parts have proper parts).

1. Something exists.
2. If there are no dependent things, something is independent and we’re done. Otherwise take the sum of all dependent things.
3. If it’s independent or mixed, we’re done. So assume it’s dependent. What’s it dependent on?
4. If it’s dependent on something that isn’t part of it, that thing must be independent or mixed, and we’re done.
5. If it’s dependent on something that is part of it, that’s either absurd or counts as it being independent or mixed. (This is the dodgy step.)

In Cameron’s paper, he does sort of address a version of the summing-the-dependents objection, although not with particular reference to Avicenna. Let’s look at what he says:

Another potential justification for the intuition is familiar from the debate between Leibniz and Hume. Here, the thought is that if there could be an infinite chain of entities e1, e2, e3, ... such that e1 is ontologically dependent on e2, and e2 ontologically dependent on e3 etc, then, while every entity in the chain is grounded, nothing grounds the chain itself. Even if there needn’t be a first member of the chain – an independent entity that provides the ultimate ontological grounding for every member of the chain - there must be an ontologically independent entity to ground the existence of the chain itself.

But that’s unconvincing. Grant for the sake of argument that not only must every being on the chain have an ontological grounding but the chain itself must have an ontological grounding. This doesn’t entail that anything is an independent entity. Perhaps the chain of entities e1, e2, e3, ... depends on a further entity ea1 which depends on ea2, which depends on ea3 etc? And if someone asks “but what about the chain ea1, ea2, ea3...?” we can appeal to a new entity eb1 which is the ontological ground of this new chain, and which depends on eb2 which depends on eb3 etc. And so on. In each case, the infinite chain of entities is dependent on an entity which is itself the first member of another infinite chain. Provided we’re prepared to postulate more and more entities, one for every cardinal number, then nothing will go ungrounded. (Cameron 2008: 11)

I don’t think this works against Avicenna’s version of the objection. The problem is that mereology isn’t like set theory. (The reference to “one for every cardinal number” is talking about the way set theory deals with this kind of thing.) In set theory you can’t just gather all the things of a certain kind into a set and ask questions about it. You get Russell’s paradox and others if you allow that. But in mereology you can gather all the dependents together into a sum and ask questions about it. You don’t get the paradoxes, and in fact accepting unrestricted mereological composition is fairly popular among people who work on the topic. (The difference basically arises because the set of all Xs can’t have members that aren’t Xs, but the sum of all Xs can have parts that aren’t Xs. For example, the sum of all cars has parts that are wheels, not cars.) Since mereology allows this kind of comprehension principle, it doesn’t help to postulate more and more entities as Cameron suggests. Anything you postulate will either be part of the sum-of-all-dependents already or won’t be a dependent. There are issues about infinite extensibility and unrestricted quantification which might apply here, but our understanding of that is much less settled than our understanding of set theory, and moving from the difficulties of unrestricted quantification to the impossibility of unrestricted composition is a leap that would need some heavy-duty justification. I think it’s fair to say that in light of Avicenna’s version of the argument, Cameron hasn’t really said enough here to fend off the objection.

So I guess bringing Avicenna into the debate wasn’t a complete waste of time. But might there be something else wrong with Avicenna’s argument? The argument looks pretty solid, at least given the assumptions about mereology, except for the step at the end. Why shouldn’t an object be dependent on something that’s part of it? There are a couple of structures we can imagine as challenges to Avicenna’s picture.

• Gunk: everything is dependent on its proper parts taken together, and everything has proper parts. So everything is dependent on something.
• Turtles: reality is made of the earth sitting on top of an infinite descending series of turtles. Everything is dependent on the sum of the things its parts depend on, and everything is dependent on any segment of the series unbounded below and wholly strictly lower than it, if any. So the earth is dependent on the sum of the turtles, the top turtle is dependent on the sum of the other turtles, and the whole of reality is dependent on the sum of the turtles but not on itself (since nothing depends on the earth), and the sum of the turtles is dependent on the sum of the turtles other than the top one. (There are some issues to go into about overdetermination, joint dependence and so on, but I think it should be possible to fill in these details in this general picture.)
• If you prefer, you can work with Simple Turtles: reality is a mereologically simple earth above a (not densely ordered) infinite descending series of mereologically simple turtles. Everything depends on the sum of everything strictly below its top part.

That wasn’t so hard. What was Avicenna thinking? Three possibilities come to mind:

• He had a notion of necessity-in-itself which rules out these structures.
• He had a notion which doesn’t rule out these structures but he didn’t think of them.
• He had a notion which doesn’t rule out these structures but had other substantive commitments that do.

All are prima facie plausible, though the second is uncharitable. Avicenna scholars will have views, but I’m not in a position to say what those views would be. But we can still think about how we should respond to these cases.

It’s worth taking the gunk and turtle cases separately. With gunk I think the best thing to do is just admit defeat if we’re allowing that things can depend on their proper parts. I think composition is identity, and things don’t depend on their parts: things are their parts. (Gunk is a bit weird if composition is identity, but I don’t see that it’s incoherent. Indeed, I don’t see that you couldn’t have gunky pluralities even if composition isn’t identity. Regular readers will be familiar with gunky pluralities from the previous post.) If you don’t think composition is identity - and if you’ve worked in the area then you probably don’t - then you might adopt some other substantive commitment linking mereology and dependence, like saying if an object depends on a part of x it depends on x, and so if things depended on their parts they would depend on themselves, which tends not to be allowed. But if you think things can depend on their parts, I guess you’re welcome to think the gunk example undermines the Avicennan argument. (And Avicenna does seem to think things depend in some way on their proper parts, which is part of why the necessary existent can’t have proper parts. I definitely feel like I’m missing something.)

With Turtles and Simple Turtles there isn’t really a ready-made metaphysical thesis like composition as identity that I can wheel out to undermine the argument. So what can we say? Well, I don’t have a great answer, but I do think that even being able to ask the question moves the debate forward a bit in terms of where you can apply pressure. The problem Cameron had with turtles all the way down was that it’s theoretically uneconomical: you can’t have one base explaining everything, because the base needs a separate explanation, and so on forever. It’s inelegant. But now we have a different objection: Turtles isn’t just inelegant; it’s weird. And where there’s weirdness, there might be Rationally Compelling Metaphysically Necessary Principles to rule it out. If you can think of one, feel free to put it in the comments.

References:

• Adamson, Peter (2010-present): The History of Philosophy Without Any Gaps. Podcast. Historyofphilosophy.net.
• Cameron, Ross P. (2008). Turtles all the way down: Regress, priority and fundamentality. Philosophical Quarterly 58 (230):1-14.
• McGinnis, Jon (2010). Avicenna. Oxford University Press.
• McGinnis, Jon & Reisman, David C. (eds.) (2007). Classical Arabic Philosophy: An Anthology of Sources. Hackett.

Gunky Pluralities

Philosophers have been squeezed out of most traditional areas of physics by fancy scientists with their “experiments” and their “research budgets”, but one thing they’ve managed to cling on to is mereology. We’ve even reclassified it as metaphysics so the physicists won’t accidentally stumble across it when looking for a book on something more important. Mereology is the theory of parts and wholes. Mereologists try to answer questions like these:

• Is there a great big object which all the other objects are parts of?
• When x is a proper part of y, is there always another object z which is the rest of y?
• Does anything even have proper parts? (A proper part of x is a part of x that isn’t x itself.)
• Could there be two different objects made out of the same parts at the same time?
• Could there be something all of whose parts had multiple parts?

The last question is about gunk. I wrote here once before about gunk, when describing an argument by David Hume against the possibility of gunk. You have to be a little bit careful when defining gunk, because what seem like equivalent definitions might not be equivalent if you accept some other exotic principles about parthood, or at least reject some mundane ones. One fairly careful definition of gunk would be “something all of whose parts have proper parts”, but that doesn’t really get at what we had in mind if we allow that things can be proper parts of each other.

Now, parts and wholes aren’t the only game in town when it comes to gathering objects together. Somebody once told me that Bolzano identified like fifteen different ways of gathering objects together. Maybe it wasn’t fifteen, but I’m pretty sure it was a lot. Although perhaps this person was pulling my leg.

Anyway, one alternative to making a whole out of some objects is to make a plurality out of them. A broom is an object made of a brush and a handle. A brush and a handle are a plurality whose members are a brush and a handle. (Sets are different again; two things form a set, but they are a plurality.) Now, you might think that there’s no distinction here: a broom just is a brush and a handle. If that’s what you think then I’m actually on your side, but most people who work on this stuff don’t think that’s right. (One person who works on this stuff and thinks it is right is Meg Wallace, of whose work on both this and other things I am a fan.) The mainstream view is that a broom can’t be a brush and a handle, because (apart from anything else) a broom is one thing and a brush and a handle are two things. Anyway, let’s make the distinction.

It turns out that a lot of the candidate principles governing wholes and their parts are analogous to candidate principles governing pluralities and their subpluralities. Let’s say that Tom and Harry are among Tom, Dick and Harry. Let’s also say that Tom is among Tom, Dick and Harry, and that Tom, Dick and Harry are among Tom, Dick and Harry, but not properly among them. And let's allow that "some things, the xs, are F" can be true even if only one thing is F, so some things are (each!) Buzz Aldrin. Now we have a language in which to ask similar questions about pluralities and subpluralities to the ones we asked about parts and wholes.

• Are there some things, the xs, such that whenever there are some things they’re among the xs?
• When the xs are properly among the ys, are there always some things, the zs, that are the rest of the ys?
• Are any things properly among any other things?
• Could there be some things, the xs and the ys, such that any things among the xs were also among the ys and vice versa, but the xs weren’t the ys?
• Could there be some things, the xs, such that whenever some things the ys were among the xs there were some things, the zs, properly among the ys?

Call some xs that fit the definition in the last question a gunky plurality. Could you have gunky pluralities? Are they ridiculous? I asked Twitter if they were ridiculous, and the eight respondents were evenly split on the matter.

Am I an experimental philosopher yet?

I was a little bit surprised. I think gunky pluralities are coherent, but in the past I’ve never detected much enthusiasm for them. While a Twitter poll with eight respondents doesn’t give much of an indication of the frequency of a position among any population apart from the people who responded to it, I was quite surprised to see that four people, not including me, saw the tweet who don’t think gunky pluralities are ridiculous. Maybe they didn’t understand the question. I did phrase it in terms of membership instead of in terms of amongness, but if anything I’d expect that to make the position seem more ridiculous, not less.

Now, I’m invested in gunky pluralities being coherent because I think composition’s identity and amongness is parthood, and so if gunky pluralities are incoherent then gunk is incoherent, and nobody wants to be committed to that. (Someone with a fancy research budget might come along and make a fool of you.) But even if you don’t think composition’s identity, and I suppose even if you don’t think merelogical gunk is possible, you can still make sense of the question about gunky pluralities. Are they ridiculous or aren’t they?

I think that Hume and Leibniz probably took the view that they were ridiculous. Hume may well have been thinking only about mereological gunk, and I’m pretty sure Leibniz was, but it would have been kind of weird to endorse their arguments for the mereological case and not the analogous arguments for the pluralities case. It’s possible of course that they didn’t really see a distinction, Bolzano not having arrived on the scene until the following century. Hume credited his argument to a Monsieur Malezieu, who I guess is probably this guy, although his English Wikipedia article could use some work and his French one doesn't mention Hume. Here are some quotes for you:

It is evident, that existence in itself belongs only to unity, and is never applicable to number, but on account of the unites, of which the number is composed. Twenty men may be said to exist; but it is only because one, two, three, four, &c. are existent, and if you deny the existence of the latter, that of the former falls of course. It is therefore utterly absurd to suppose any number to exist, and yet deny the existence of unites; [...]

But the unity, which can exist alone, and whose existence is necessary to that of all number, is of another kind, and must be perfectly indivisible, and incapable of being resolved into any lesser unity. (Hume, Treatise on Human Nature, 1.2.2)

And there must be simple substances, since there are compounds; for a compound is nothing but a collection or aggregatum of simple things. (Leibniz, Monadology, translated by Robert Latta, proposition 2)

It’s also true that our language gets a bit strained when we try to talk in a way that never presupposes that the thing we’re referring to is just one thing. You’ll have noticed that when I was trying to do it earlier. On the other hand, there doesn’t seem to be any reason you can’t construct a logic that can accommodate gunky pluralities. You might struggle to get a model theory in terms of sets that didn’t make a certain kind of person a bit grumpy, but this kind of person is already grumpy about variable-domain model theory for modal logic, so you’ll be in good company. (It may be that their grumpiness is warranted, but my impression is that even if the objection in the case of modal logic succeeds, the analogous objection would be question-begging in the case of gunky pluralities. But I’m pretty open to being wrong about that.) One possibility is that the idea of gunky pluralities is one of those things that’s ridiculous without being incoherent. I don’t really get what the problem is supposed to be, though. If you think they’re ridiculous, and it seems at least four of you do, let me know why in the comments!