Translating from BS to plain English, David Brooks believes that empirical Renaissance values should guide our thinking, except when they disagree with David Brooks. Yes this makes Brooks a hypocritical tool, but the statement is also a tautology. Everyone thinks like that. No healthy person ever lets empirical thinking or their inbuilt emotional prejudices guide every single thing they do. The great rationalist Emmanuel Kant made the same general point. We will not leave the elderly on an iceberg or legalize strangling puppies even if some (hypothetical) studies said that we would be somehow better off, nor (I hope) just because we feel like it.
Brooks is thus not an intolerably smug ass just because he does what everyone but the insane and the tea party already do. Rather, in a fair world Brooks would get his underwear pulled over his head for acting like he deserves a Nobel prize every time he cites a study and another Nobel each time he criticizes one. If you want to be the better man, dude, take some of the millions you paid for that Cleveland Park mansion “with vast spaces for entertaining” and give it to the homeless.
Brooks is only reminding everyone that Big Data is meaningless without consulting Brooks to find out what it all means.
Your criticism of Brooks’s massive sense of entitlement and self-regard must be incorrect, as Brooks is well known to be – indeed, has appointed himself as – an expert on humility.
But Sally Quinn would not approve. And BoBo is all about defending his sinecure on the cocktail weenie circuit.
If your math tells you that you should eat babies, it’s much more likely that you messed up somewhere, rather than that you should actually eat babies. I don’t see what’s so irrational about that. Of all the times that people thought they should eat babies, how often were they actually right? So, yeah, just make a note of that.
I disagree. Yes, this process exists in all humans. Water is also able to poison you. Almost all comparisons are a matter of scale, not kind. Viewed as a difference in scale, Brooks is a champion rationalizer, throwing on random facts and big words, claiming whatever values he hears are popular, and then ALWAYS circling around to whatever conservative point he thought was true to start with. The vagueness of his logic chains and his ability to pretend he’s following the facts (apparently believing it) to conclude their exact opposite is far beyond regular humans. I say that despite rationalizing being one of humanity’s major thought processes. Brooks vastly exceeds the norm.
@Warren Terra: He does have much to be humble about.
Shockingly, I bet the David Brooks doesn’t worry so much about the “big data” aspects of the study of, say, Economics (just to pull a random example from my ass) that miss out on the morality and ethics of decision making when those Economic studies promote more money in David Brooks’s pockets.
(What Brooks should be concerned about is people mistaking the model for reality – because that’s where immorality and unethical consequences tend to arise. As grendelkhan points out, if your model leads you to believe that “eating babies” is the right answer, your model is clearly divorced from reality. This kind of problem is also the foundation of “libertarian” thought, such as it is…)
We have a running joke in my house that David Brooks’ idiocy has devalued my daughter’s U of Chicago degree.
Much more of this nonsense from him and it won’t be a joke.
Don’t forget that this is of a parcel with his inability to cope with Nate Silver being right when Brooks was wrong.
Huh… what is Brooks talking about? I guess I have to get this ‘Big Data’ book. From the description of the book, seems to be about predictive numeric and text data mining. OK, I plead guilty to committing ‘Big Data’ analysis from time to time.
But I can’t figure out what Brooks is talking about.
Conceptually, statistics is usually pretty simple minded stuff, even if the details get pesky.
OK, if you have big data you can do two things. One, you have enough data to use cool whizz bang techniques to figure out whether the correlations you find are stable across new samples, or just due to chance.
Two, since you can never observe causal relationships directly, but only through correlations you see, having more data allows you to better figure out which causal model is most consistent with the observed correlations.
Having enough data, and being able to acquire big chunks of new data on a regular basis allows you to try out, and evaluate, how different models predict out of the estimation sample. (Edit: evaluate both stability of correlations and consistency with competing causal models).
Brooks should read him some Judea Pearl and learn about directed acyclic graphs (DAGs). Most of the approaches to choosing which causal model is most consistent with observed correlations reduce to DAGs, and you can draw cute diagrams rather than wade through mysterious algebra.
But with Brooks, one suspects there is a subtext crawling through the backdoor (har har, backdoor, inside joke for stats people out there).
I short, I can’t work up any moral or existential dilemmas when I work with Bid Data (whatever it is, or they are, if you want to be pedantic). I guess I am either a liberal so debauched my soul is gone, or I am shallow.
Edit: also, Brooks seems to confuse predicting individual responses and averages repsonses, which are two different things.
Most of the advocates understand data is a tool, not a worldview. My worries mostly concentrate on the cultural impact of the big data vogue. If you adopt a mind-set that replaces the narrative with the empirical, you have problems thinking about personal responsibility and morality, which are based on causation. You wind up with a demoralized society. But that’s a subject for another day.
Brooks can be easily translated using the Lewis Carrol lexicon.
TWEEDLEDUM: I know what you’re thinking about . . . but it isn’t so, nohow.
TWEEDLEDEE: Contrariwise . . . if it was so, it might be . . . That’s logic.
Villago Delenda Est
The comedy continues to flow like water here at BJ this morning. First DougJ and “the right thinks” and now this.
@Ben Franklin: I think this means he has mistaken number crunching for utilitarianism. A rather typically brookdian confusion between is and ought.
Bighorn Ordovian Dolomite
Well, I (sorta) took Charlie up on his suggestion to translate this. Here’s what you get with a google translate to Japanes and then back to English:
I honestly don’t think the readability suffered all that much.
Belafon (formerly anonevent)
@jl: All of Brook’s graphs would be fully connected, because every graph with a good outcome would have him as the cause, and every graph with a bad outcome would have not listening to him as the cause.
@jl: Hmm, I have a feeling we may have met before.
Brooks should read him some Judea Pearl
A man who not only knows more about data and information than Brooks (no shame on Brooks for that), but also vastly more about morality, humility and sacrifice than Brooks would ever care to contemplate. I’m almost ticked at you for linking their names in any context.
I am just trying to imagine the sort of entertainment that one could expect at Festung Brooks. So far, I haven’t been able to think of a single thing Brooks could do or say that would be enjoyable or interesting.
mike with a mic
I don’t think it’s exactly that. Most conservatives hate the issue that numbers prove deficit spending is good for the economy and and social security is good for the economy. There’s no way to argue the data there. So the only counter point they can make is “yes but, deficit spending is immoral and social security creates people dependent on the government and a lack of personal agency. Thus even if these issues are good according to the data, they are bad for moral issues that data is incapable of quantifying, and so the conservative view is correct”.
Brooks knows that the economic policies and domestic role of government that he and his backers support can never be justified on the data. It’s a joke. All he’s got is to warn over risks in the unquantifiable areas of societal morality and behavior and to insist that’s what matters.
Won’t someone please think of Burkean morality!
Oh, I don’t have anything to add — I just wanted to make sure my copy/paste function was working. Having this line handy should save me a lot time.
Somewhat OT, Chomsky nails it on Adam Smith.
(Not completely unrelated; replace Renaissance by Enlightenment.)
Guy is a complete pseudointellectual.
@mike with a mic:
Well, that last part is pretty much all they care about. The numbers are there to be trumpeted loudly when they prove what conservatives want to believe, and ignored when they don’t. For them, it’s all about How The World Ought To Be.
Bobo is only advocating that we be suspicious of Big Data because it keeps rudely disproving every fucking thing he and his political team believe in and he’s quite tired of it, thankyouverymuch.
@mike with a mic:
…yes, this. Bobo and those he fellates need to turn politics and public policy into a morality play, because the data demonstrates that they’re full of shit on the merits.
I wonder if the Brooks column is related to the Reinhart-Rogoff blow up on data processing errors, oddly selective use of available data, and funny rules for weighting observations.
Odd reactionary doctrines with no (out of sample) predictve value cannot be failed by numbers. Rather, numbers can only fail odd reactionary doctrines. Or something like that.
Let’s see if more of the VSP pundits come out with similar columns and posts in the next week or so.
Cynical me figures that orders go out for these things, via fax or text msg.
That op-ed is pure gibberish. He says things like
What does this even mean? This makes no sense in English or any sense mathematically. Student run newspapers have better op-ed writers than Brooks.
ETA: Continuous functions have a very specific meaning in calculus. When something is continuous you can differentiate it, integrate it etc. Calculus 101 fail.
Not that I actually work with big data per se, but it’s amusing how (at least according to circulating stories) R-R made their boo-boos in Excel, not (say) Hadoop.
I guess you should always check your work, whatever you’re working with, but given the scale of the dataset they had, you’d think it would have been pretty manageable.
Wonder what Brooks will say about this paper http://www.peri.umass.edu/236/hash/31e2ff374b6377b2ddec04deaa6388b1/publication/566/ that finds ‘coding errors’ and selective data inclusion in the primary study used as the foundation to support global austerity movements (and Paul Ryan’s budget). Don’t know how it works in economic journals, but in medical and scientific journals, these issues would be grounds for retraction. Given that, I’m sure Brooks will rush to correct his position.
Summary here: http://www.nextnewdeal.net/rortybomb/researchers-finally-replicated-reinhart-rogoff-and-there-are-serious-problems
It’s more pseudointellectual word salad.
Though you can’t necessarily differentiate a continuous function. (Ex: y = abs(x) at x=0.)
@liberal: OK let me put it this way you cannot differentiate a function unless it is continuous.
I will admit that Brooks is right about one thing. Causality is very distinct and different from correlation. But he writes as if researchers and statisticians do not understand this.
But they seem to understand the issue (Edit: or at least are aware of it), otherwise how did mountains of books and articles on the topic come be published?
You have to tackle this issue head on as soon as you teach multiple regression (that is regressions with more than on ‘explanatory’ variable) in a beginning or intermediate stats class. You should probably talk about it earlier, but by then you MUST talk about it.
Either this big data book does not mention that issue (which means it is a bad stats book), or Brooks did not read it, or Brooks read it and did not understand it.
I am betting it is the latter, our punditubbies is usually innumerate.
mike with a mic
Every Brooks diatribe is a long winded way of saying “look, the world might be one way, and the data might say one thing, but human beings are different because of human nature. So we must enforce a center right Burkean morality on humanity or mankind will turn into the barbarian horde and civilization will collapse, and don’t say I didn’t warn you” with a side shot of “this obvious does not apply to those who went to the Ivy Leagues because their education by nature proves they are properly breed and will act right”.
He screws up his analogies and metaphors left and right, but that’s always where he is going.
Causality is very distinct and different from correlation.
We could quibble and muse here about the meaning of ‘distinct.’
But the larger point is that quantitative scientists use correlation in service of causal in precisely the way Brooks claims to have discovered – i.e. there is a model (‘narrative’ in his terms) evaluated, and potentially updated, against observed associations. As always with Brooks, its his ignorance about his own ignorance that makes him look the most foolish.
It sure does keep the $ rolling in, though. Perhaps we’re the greater fools, in utilitarian terms.
Not that hard to check calcs on much bigger data sets on Excel. Weird.
And, you know, data sets and calculations in Excel are kinda important to double and triple check. At least if you are mere mortal person without perfect recall of every button you push.
Short Bus Bully
It’s an oldie but a still a goodie: “[David Brooks] is what stupid people think an intellectual sounds like.”
And he tells them what they already believe, so there’s that.
@jl: You have to tackle this issue head on as soon as you teach multiple regression
This should be dealt with on day one in any course dealing with observations subject to unobserved causal influences (in any discipline), no matter how elementary the course. The methods can get tricky but the basic issue is very easy to understand. Waiting until you’re knee-deep in methods makes things needlessly confusing.
Cheryl from Maryland
More Zaum from David Brooks. That man is full of Zaum.
OK. you have a point. Many specifications of different causal models with different predictions about what will happen in new situations can produce the same or very similar correlations in one set of sample data.
Is that better?
I know that there is a lot of musing about fundamental relationship between deep causal relationships and observed correlations in philosophy and statistics. Gets weirder and more confusing in nonstationary systems.
My main beef with Excel these day is that it really gives naive people who have data that should be in an RDMS lots of room to screw everything up.
Though I guess that’s not really Excel’s fault.
I think what Bobo is trying to say is that there’s no point trying to precisely measure and quantify human behavior because humans are unpredictable. And that’s not complete bullshit, but the whole point of statistics is to try to make sense of that sort of thing, and the good statisticians do a decent job of it.
What he is saying actually can be applied to disprove pretty much the entire Chicago school of economics. Their fancy models, which were then exported to the real world with mostly brutal results for actual people and economies, assumed that people were rational actors who optimized their choices in the marketplace. A thirty-second trip to the local 7-11 is enough to disabuse most people of that notion permanently, yet it was one of the foundational pieces of their economic theories. So their entire belief system is based on an obvious and deliberate misunderstanding of the population they were trying to study and predict. To paraphrase ol’ Rummy, they went to war with the human beings they wanted to have, not the ones they had.
But that doesn’t, as Brooks might want us to think, undermine the value of empiricism. Quite the opposite – economists who don’t assume people and/or macro markets are always rational and always optimize, and instead study their actual behavior in the real world and do honest statistical analysis and modeling of what they find, tend to do much better when attempting to explain what is/was happening or forecast what will happen.
in short, Bob is full of shit. Suns rises in east also too.
@jl: Hey, I wasn’t picking a fight – hence “quibble” – just point out that it’s not even clear Brooks gets that “one thing” right :)
It’s actually a fascinating conversation. You’d think at this late date we’d have a pretty solid grasp on what we mean by ‘evidence’ and ’cause;’ turns out things get pretty weird in a hurry.
@Turgidson: Though, Brooks does seem to confuse the use of statistics in predicting an individual observation, and predicting average of a large number of observations. That is, confusing predicting individual response with predicting average response. And that leads to confusion of predicting influence of an explanatory factor for an individual, with influence of an explanatory factor in predicting average response in a large sample. (Edit: and note that distinct causal models may be appropriate for each case)
Again, that is stuff introduced beginning stats courses, and something that absolutely must be covered in detail in an intermediate course.
So, to repeat, either that Big Data book is very superficial, or Brooks never read it, or he did not understand much of it.
@jl: Brooks does seem to confuse the use of statistics in predicting an individual observation
Well yeah, even the title of the column is a pretty dumb error in this vein, though he may not be responsible for that.
Who says we would not strand the elderly on an ice floe(if we could find one) or strangle puppies. As long as we could find a guy at Justice to write the enabling memo, we’d be fine with it.
Of course, if these hyper-rational Chicago school professors actually spent thirty seconds talking to a marketing professor in the building next door they’d have been disabused of these ideas as well.
But even if they had tried to obtain all relevant information (like rational actors), they’d have had to decide and admit that they were wrong and others were right (like rational actors), and then instantly change their lifelong behavior to compensate (like rational actors). Not bloody likely.
I’m not very familiar with Pearl’s work, but he won my loyalty with a statement in a recent ACM Communications interview, “Economists have betrayed causality.”
You’re a scientist, Doug, so I won’t belabor this. But as a practical matter, no one can use empirical reasoning to govern all their actions because empiricism has no foundation without theory.
To put it bluntly, what you observe is a function of what you theorize. If your theoretical framework is the Ptolemaic system of epicycles, then errors in observations of the planets’ orbits lead you to add more epicycles. If your theoretical framework is the Copernicas sun-centered system, then errors in observations lead you to toss out the epicycles and start thinking about elliptical orbits.
What you observe depends on what you expect. And what you expect depends on your theoretical framework.
Empiricism and logical positivism don’t work as a foundation for living your life because separated from the theoretical foundation, observations are just meaningless data. You don’t know which are bad data points, and what the trend curve plotted out by that data means.
Out of curiousity, how often does Google spider balloon-juice.com?
What does that ‘BS’ stand for? Oh, got it: Brooks Speak!