## April 30, 2012

### Books to Read While the Algae Grow in Your Fur, April 2012

Attention conservation notice: I have no taste.

Susan Whitfield, Life along the Silk Road
Not-quite-historical fiction: life stories of sundry Silk Road characters — merchants, monks, soldiers, artists, ordinary widows — distributed from Samarkand to Chang-an, and from 700 to 900 AD. These are all more or less composites of actual people, glimpsed from the archaeological record, and especially through the manuscripts preserved at Dunhuang and saved/stolen by Aurel Stein. (In fact the whole book owes a great deal to Stein, with a lot of input from Beckwith's The Tibetan Empire in Central Asia.) The lack of references makes it hard to know how much is stitched together from sources and how much is Whitfield's invention, but at the very least it's well-told.
Nathan Long, Jane Carver of Waar
Mind candy. This is at once a parody of, and homage to, Barsoom. Unlike Burroughs, Long's book can be enjoyed after the Golden Age of Science Fiction (i.e., by those over the age of sixteen): his characters are all at least two-dimensional (Jane herself is an engaging narrator, though definitely at the Hill end of the Moby-Hill spectrum), his style is decent, and the plot is actually interesting. I think it would be enjoyable even if you hadn't dosed up on planetary romances as a kid.
Sequel.
James S. A. Corey (i.e., Daniel Abraham and Ty Franck), Leviathan Wakes
Mind candy. Space opera, confined to the solar system a few centuries hence. This has gotten a lot of favorable attention, but I found it merely OK; perhaps I'd have enjoyed it more if my expectations had been lower. It's split between two plot lines, with two point-of-view characters; I enjoyed (but wasn't blown away by) one of them, but found the other both over-predictable and irritating. It does some things well (a reasonably-sized solar system! minimal handwavium! a non-grim-meathook-future future! some decent characterization!), but it never really managed to grab me. It's definitely nowhere near as good as say, McAuley's The Quiet War, to name a recent and thematically-similar book. The sequel will be out soon, and seems like it will be continuing along the better of the two narrative threads here, so I might pick it up, but I won't rush to do so.
Spoiler-laden griping: Bar bs gur gjb cybg yvarf vf n uneq-obvyrq vairfgvtngvba, pbzcyrgr jvgu na nypbubyvp zvqqyr-ntrq qrgrpgvir, pbeehcg vagevthrf, naq n zlfgrevbhf qnzr jub gur qrgrpgvir snyyf va ybir jvgu. V qba'g yvxr gur uneq-obvyrq traer, orpnhfr, juvyr V nz irel fragvzragny, vgf cnegvphyne pbzovangvba bs fragvzragnyvgl naq plavpvfz vf bss-chggvat. Fb onfvpnyyl V jnagrq gb fxvc nyy gur puncgref sebz Zvyyre'f cbvag bs ivrj, naq whfg sbyybj gubfr jvgu Ubyqra naq uvf perj. Yrff crefbanyyl (v.r., nf n engvbanyvmngvba), abve cerfhccbfrf fhpu n irel cnegvphyne, uvfgbevpnyyl-yvzvgrq phygheny frggvat gung frrvat vg fvzcyl qhzcrq vagb jung fubhyq or n enqvpnyyl arj xvaq bs fbpvrgl jnf wneevat. (Rirelguvat ba Prerf jbexf yvxr Puvpntb pvepn 1940 orpnhfr ubj ryfr?).
Ba n qvssrerag cynar nygbtrgure, Cebgbtra'f ernfbaf sbe jnagvat gb gel bhg gur nyvra ivehf/znpuvar ba gur jubyr cbchyngvba bs Rebf ner jrnx. Vs gur cbvag bs gur znpuvar vf gb gnxr bire rkvfgvat ovbznff naq erfuncr vg nppbeqvat gb fbzr cebtenz, vg jbhyq frrz vasvavgryl rnfvre gb tvir vg hzcgrra gbaf bs lrnfg gb cynl jvgu, guna gb fcraq lrnef bepurfgengvat gur gnxr-bire bs n pbybal jvgu bire n zvyyvba crbcyr, gb fnl abguvat bs gur erqhprq cbffvovyvgl sbe oybj-onpx, frphevgl oernpurf, rgp. Ab qbhog gurl'q jnag gb gel vg ba crbcyr riraghnyyl, ohg fgnegvat gurer, jvgu ab pbageby bire rssrpgf, vf whfg onq rkcrevzragny qrfvta. Cyhf "tvir gur napvrag fhcre-nqinaprq nyvra jne znpuvar pbageby bire na nfgrebvq" qbrf abg fbhaq yvxr n cyna juvpu jbhyq qrirybc gb n fbpvbcngu'f nqinagntr. (Gurl jbhyqa'g pner nobhg gur qnzntr gb bguref, ohg gurzfryirf?)
In conclusion, bring me back my cane and then get off my lawn, you're trampling the lilies.
Matthew Johnson, Fall from Earth [buying: publisher, audio]
Mind candy. Scheme-laden first-contact space opera with a social setting I can only call "The Ming Dynasty IN SPAAAAACE". Good enough that I will keep a look out for more from Johnson.
It's a small thing, but Johnson shows no appreciation of the energy required to move food from planet to planet, which makes his "equitable marketing system" a complete non-starter. (But he shares this flaw with Cherryh's deservedly-admired Downbelow Station.) If, however, the magistracy wants to make sure that no world can become self-sufficient, the way to do it would be to restrict their manufacturing, since any colony would be dependent for survival on a complex industrial infrastructure.
Bernard Williams, Truth and Truthfulness: An Essay in Genealogy
Shorter Williams: "Say what you mean. Bear witness. Iterate." (The late John M. Ford, in a different context.)
Slightly longer: You can get a decent sense of what the book is about from the publishers, so I'll comment without much exposition.
When Williams talks about a "genealogy" of some idea or practice, he means an account of why, if it did not exist, we would have to invent it. Specifically, he spins a state-of-nature story about how if, in the state of nature, human beings did not have an idea of truth, but nonetheless were social and rational animals, and so dependent on a division of epistemic labor, they would have to form one, and two "virtues of truthfulness", namely "sincerity" (Ford's "say what you mean") and "accuracy" (Ford's "bear witness") to make it effective. This is not intended as history or pre-history (Williams: "the state of nature is not the Pleistocene"), but it is a bit mysterious to me how then it is supposed to explain our notions of truth, truthfulness, sincerity, accuracy, etc., much less explain them "non-reductively". Perhaps — this is suggested by his section on "Shameful Origins" — it is just supposed to make us feel better about having them, by convincing us that we could have acquired such ideas in a way which doesn't discredit them. (We are not suckers.)
It may sound odd to describe "accuracy" as a virtue, but being accurate --- bearing good witness --- means things like check tendencies to leap to conclusion, choosing appropriate methods of inquiry, taking pains to secure all the relevant facts (Williams is especially good on the notion of "facts"), etc. Williams is indeed eloquent on how the virtues of accuracy are one of the things which have made the pursuit of science a source of human values, especially in circumstances where honesty otherwise was hard.
As this last suggests, culture lets us articulate the raw virtues of sincerity and accuracy into incredibly elaborate and interlocking complexes of attitudes and practices (Ford's "iterate"). From the inside, these have, or at least seem to have, intrinsic as well as instrumental value, and indeed they would not work at if their value was just instrumental. I confess that I do not fully follow Williams's attempt to try to explain when or why or how the virtues of truth become "intrinsic values". It seems to be something like: people find these values compelling, in a way which they would not if they saw them just as handy tools for achieving selfish ends; this in turn makes these values successful commitment devices [1]. Williams seems to me to equivocate as to whether these virtues really do have such intrinsic value, but on balance I am just as happy that he strayed no deeper into the swamp of meta-ethics, and wisely turned back to the sounder terrain of looking at certain episodes in the articulation of these virtues. The two main case-studies he gives are contrasts of Thucydides and Herodotus on history, and of Rousseau and Diderot on authenticity and the self. Both of these really have a wider, philosophical import, and as such they would both have been stronger for a more comparative, cross-cultural perspective — not in the service of the small virtue of courtesy (Williams has mercifully few "what you mean 'we', white man?" moments), but rather in the service of the great virtue of accuracy [2].
But I see that I am descending into my usual quibbling. This is a profoundly thoughtful and profoundly learned book, which says interesting things about some of the deepest and most humanly-important problems in philosophy, and says them elegantly. Go read.
[1] I cannot help but be reminded of William James:
Now, why do the various animals do what seem to us such strange things, in the presence of such outlandish stimuli? Why does the hen, for example, submit herself to the tedium of incubating such a fearfully uninteresting set of objects as a nestful of eggs, unless she have some sort of a prophetic inkling of the result? The only answer is ad hominem. We can only interpret the instincts of brutes by what we know of instincts in ourselves. Why do men always lie down, when they can, on soft beds rather than on hard floors? Why do they sit round the stove on a cold day? Why, in a, room, do they place themselves, ninety-nine times out of a hundred, with their faces towards its middle rather than to the wall? Why do they prefer saddle of mutton and champagne to hard-tack and ditch-water? Why does the maiden interest the youth so that everything about her seems more important and significant than anything else in the world? Nothing more can be said than that these are human ways, and that every creature likes its own ways, and takes to the following them as a, matter of course. Science may come and consider these ways, and find that most of them are useful. But it is not for the sake of their utility that they are followed, but because at the moment of following them we feel that that is the only appropriate and natural thing to do. Not one man in a billion, when taking his dinner, ever thinks of utility. He eats because the food tastes good and makes him want more. If you ask him why he should want to eat more of what tastes like that, instead of revering you as a philosopher he will probably laugh at you for a fool. The connection between the savory sensation and the act it awakens is for him absolute and selbstverständlich, an "a priori synthesis" of the most perfect sort, needing no proof but its own evidence. It takes, in short, what Berkeley calls a mind debauched by learning to carry the process of making the natural seem strange, so far as to ask for the why of any instinctive human act. To the metaphysician alone can such questions occur as: Why do we smile, when pleased, and not scowl? Why are we unable to talk to a crowd as we talk to a single friend? Why does a particular maiden turn our wits so upside-down? The common man can only say, "Of course we smile, of course our heart palpitates at the sight of the crowd, of course we love the maiden, that beautiful soul clad in that perfect form, so palpably and flagrantly made from all eternity to be loved!"
And so, probably, does each animal feel about the particular things it tends to do in presence of particular objects. They, too, are a priori syntheses. To the lion it is the lioness which is made to be loved; to the bear, the she-bear. To the broody hen the notion would probably seem monstrous that there should be a creature in the world to whom a nestful of eggs was not the utterly fascinating and precious and never-to-be-too-much-sat-upon object which it is to her.
Thus we may be sure that, however mysterious some animals' instincts may appear to us, our instincts will appear no less mysterious to them. And we may conclude that, to the animal which obeys it, every impulse and every step of every instinct shines with its own sufficient light, end seems at the moment the only eternally right and proper thing to do. It is done for its own sake exclusively. What voluptuous thrill may not shake a fly, when she at last discovers the one particular leaf, or carrion, or bit of dung, that out of all the world can stimulate her ovipositor to its discharge? Does not the discharge then seem to her the only fitting thing? And need she care or know anything about the future maggot and its food?
More soberly, or at least with fewer hens and maggots, this is highly reminiscent of Robert Frank's Passion within Reason, which I do not believe Williams mentions. ^
[2] Williams claims, quite plausibly, that Thucydides had different ideas about historical explanation and historical evidence than did Herodotus — ones which are both stricter about what counts as acceptable history, and which are supported by compelling rationales even within the older framework. He also claims, more sketchily, that Herodotus was immersed in a culture which was still partly oral and partly literature, while Thucydides was not. If all this was right, should not the same contrast show up in the historical traditions of China, the Islamic world, etc.? Why does such a tradition not seem to be indigenous to India? (Cf., on all this, Brown's History, Hierarchy, and Human Nature.) Western Europe, after the fall of the western Roman Empire, never lost literacy, but it certainly didn't produce histories like Thucydides's for many centuries: why, on Williams's account, not? (Actually, outside of Italy, did western Europe ever produce such histories before the fall of the empire?) If there are important distinctions between these cases, such that Williams's account applies only in the special circumstances of the Aegean around 500--300 BC, what are those circumstances? — Let me add that it was Williams who made all these considerations relevant, not me. ^
J. C. W. Rayner and D. J. Best, Smooth Tests of Goodness of Fit
Suppose a random variable $Y$ is confined to the unit interval $[0,1]$, and we want to test whether it is uniformly distributed. One way to do this would be to construct alternative distributions which are in some sense smooth departures from uniformity, with densities $g(y;\theta) = e^{\sum_{j=1}^{d}{\theta_j h_j(y)}}/z(\theta) ~ ,$ where it is convenient to chose the $h_j$ functions to be an orthonormal basis --- the cosine basis, say, or the Legendre polynomials. (That is, they are orthonormal in $L_2$, the space of square-integrable functions on the unit interval.) Uniformity is then the special case $\theta = 0$, and we can test it against the alternative that $\theta \neq 0$ by the usual devices of a likelihood-ratio test, a score test, etc., which will all, under the null hypothesis, have an asymptotic $\chi^2_d$ distribution. This is Neyman's original smooth test, which seems to have originated from the problem of how to combine p-values from independent experiments, which should all be uniformly distributed under the null hypothesis. One nice feature of this test is that if we reject the null, we immediately have an alternative, namely our maximum likelihood estimate of $\theta$, for what the actual distribution is --- it tells us not just that the null model is wrong, but how, and what a better one would be like.
The real power of this comes from the following observation. If $X$ is distributed according to some continuous CDF $F$, then $Y=F(X)$ is uniformly distributed on $[0,1]$. The smooth alternatives for $Y$ translate into smooth alternatives for $X$, with densities $g_X(x,\theta) = f(x) e^{\sum_{j=1}^{d}{\theta_j h_j(F(x))}}/z(\theta) ~ .$ We can test whether $X \sim F$ by, once again, testing with $\theta = 0$, and the theory works just as before. If $F$ is not fixed but involves some parameters $\beta$, then we consider the smooth alternative densities $g_{X}(x;\beta,\theta) = f(x;\beta) e^{\sum_{j=1}^{d}{\theta_j h_j(F(x;\beta))}}/z(\theta) ~,$ and again we test the specification by testing $\theta = 0$. Since this always involves fixing $d$ parameters, we always get a $\chi^2_d$ asymptotic distribution under the null.
Rayner and Best's monograph is a clear, if now somewhat old-fashioned, exposition of Neyman's smooth test and its relatives and extensions. They actually begin with Pearson's $X^2$ or $\chi^2$ test, which can be seen as a smooth test for multinomial (rather than continuous) data, before going on to consider the general theory of likelihood ratio and score tests, and Neyman's smooth tests. Much of the book is taken up with various permutations of discretizing continuous variables and/or allowing estimation of the parameters I have written $\beta$; the latter concern seems less important these days.
An important set of developments which does not get as much attention here as a more recent treatment would give is that of picking the order of the alternatives $d$. Neyman suggested $d = 4$ but emphasized it was guess; some later workers guessed $d = 2$ should be enough. Really, however, this is a problem of model selection or capacity control, and so all the usual tools, like cross-validation or information criteria, can be applied. This is one place where BIC has proved particularly useful, leading to "data-driven" smooth tests. These no longer have nice $\chi^2$ asymptotics, but it's pretty easy to get their sampling distributions from simulation.
Despite these limits, this is still a useful reference for people interested in specification checking.
Aliette De Bodard, Servant of the Underworld
Mind candy: historical fantasy/mystery set in Tenochtitlan (a few generations before what would be the Conquest), only with the mythology of the Aztecs being literally true and magic very much a part of actual life. It had some typical first-novel flaws (too much exposition, the plot drags in places), but overall decent.

Posted at April 30, 2012 23:59 | permanent link