Romantic Science (or, something missing is missing)

This review by Daniel Dennett of Incomplete Nature by Terrence Deacon is interesting in itself, but one passage leaped out at me, where he attempted to delineate a split between reductionists and holists, or Enlightenment and Romantic science:

There are no entirely apt labels for the opposing sides of this gulf… Reductionism, fie! Holism, fie! …“Enlightenment” versus “Romanticism” is pretty close, as the reader can judge by considering what the following team players have in common; on the Enlightenment side: Darwin, Turing, Minsky, Dawkins, both Crick and Edelman (in spite of their antagonisms), Tibor Gánti, E. O. Wilson, Steven Weinberg, Paul and Patricia Churchland, and both Raymond Kurzweil and me (in spite of our antagonisms). On the Romantic side are arrayed Romanes and Baldwin, Kropotkin, Stephen Jay Gould, Humberto Maturana, Francisco Varela, Stuart Kauffman, Roger Penrose, Ilya Prigogine, Rupert Sheldrake, and the philosophers John Haugeland, Evan Thompson, Alicia Juarrero, John Searle, ... Jerry Fodor and Thomas Nagel.

I’m not sure Varela and Gould would appreciate being lumped in with Rupert Sheldrake, and they aren’t around to defend themselves, but never mind. I suppose another label for the gulf would be greedy mechanistic reductionists on one side, vs. those who have some qualms about it. The qualms might be similar, they probably all stem from a sense that viewing minds and organisms as machines leaves out something important – the “aching void” that Dennett refers to. But the things they choose to fill the void aren’t all the same, which is why the lumping is inappropriate.

Romantic science is a recognized thing, but that wiki page makes it seem like a purely historical phenomenon, whereas Dennett recognizes as a living force today. Most remarkably, he says that Deacon’s book has caused him to shift his views in the Romantic direction, to the degree where he is “re-examining fundamental working assumptions”. That I guess is a pretty big deal from someone as prominent as Dennett and someone so identified with straight-ahead materialism.

It reminded me of a similar-but-different dichotomy I came up with the other day, between what I think of as mainstream science, roughly the same as Dennett’s first group, that is, mechanistic and reductionist, and people like Haeckel, D’arcy Thompson, Rene Thom, Buckminster Fuller, Christopher Alexander, and Adrian Bejan whose book instigated the discussion. This group, which probably also includes Prigogine, tend to be more obsessed with form and geometry than mechanism, which gives them a somewhat marginal quality, even when they are obviously right. Their tendency to reinvent metaphysics from the ground up also tends to make them marginalized, even crankish, although their very real achievements undercuts this. I’d say this is a cousin or sub-family of Romanticsm, and driven by some of the same underlying forces. Alexander especially makes this very explicit, that his entire work in architecture, aesthetics, and metaphysics is driven by a need to make a place for what he calls “the quality of life”, which has been exiled from the mainstream mechanical universe.

I guess I need to read Deacon’s book, but my sense is that this dichotomy is never going to be solved or go away. Science by its nature takes a depersonalized (or more precisely, de-subjectivized) view of the universe. Attempts to re-graft subjectivity onto the results of science always seem forced and unsatisfactory. The work of the Romantics, valuable though it may be, has its value in design philosophies or moral philosophies or something else that is on the borders of science but is not itself science. That’s one reason it is attractive, most of us aren’t practicing scientists and are hence are more interested in the consequences of scientific results to our standing in the universe than in the science itself.

The current apex of human cleverness

may be these unremarkable-looking machines, which are Illumina gene sequencers. The other day I had the pleasure and privilege of getting a tour of the DOE Joint Genome Institute in Walnut Creek, which has 8 of these babies. With these and the other sequencing machines, they sequence well over 100 gigabases of DNA a day.

Something about these machines intrigues me, the way they combine the accumulated cleverness of chemistry, optics, lasers, robotics, computation, microfluidics, all in service of gathering even more knowledge. The way they have literally black-boxed these techniques so that the entire astonishingly complex process can be treated conceptually as just one thing (sort of). The way they unite academia, industry, and government efforts, and blur the lines between them. The way all these forces are concentrated serially on the base pairs of a strand of DNA, one little atomic-scale bit at a time, until it's all read. How many people contributed to this concentrated ingenuity in one form or another? It gave me a little burst of the technological sublime.

Apparently the hot thing in genomics now is the ability to read the genomes of individual cells (single-cell genomics). Another little room had the physical device that makes this possible, which takes a drop of cell-containing water and physically breaks it into into tiny droplets with ultrasound, some small fraction of which contain cells, and then plops each of the desirable droplets into wells on a plate through manipulation of electric fields. Then each cell can be separately amplified and sequenced. Before this technique, genomes could only be sequenced from cultured cells, and most organisms can't be cultured.

The good news for me is that while the technology for producing biological data is getting more powerful at faster-than-Moore's-law rates, the ability to interpret it is lagging, which means plenty of opportunity for the software to manage, analyze, interpret, and navigate the deluge.

Speaking for Nature

Awhile back I made the point that social construction is not arbitrary -- that is, when people talk about the social construction of knowledge, they don't generally mean that in such a way as to imply that knowledge can be just anything at all -- a point which seems to elude many participants in the science wars. Also, I commented on the Climategate affair, noting that it was a pretty good example of the Latourian view of science -- exposing the messy human processes by which scientific knowledge is made. Well, I have Bruno Latour right here and he agrees with me on both counts:

What I found so ironic in the hysterical reactions of scientists and the press [to Climategate] was the almost complete agreement of opponents and proponents of the anthropogenic origin of climate change. They all seem to share the same idealistic view of Science (capital S): "If it slowly composed, it cannot be true" said the skeptics; "If we reveal how it is composed, said the proponents, it will be discussed, thus disputable, thus it cannot be true either!". After about thirty years of work in science studies, it is more than embarrassing to see that scientists had no better epistemology to rebut their adversaries. They kept using the old opposition between what is constructed and what is not constructed, instead of the slight but crucial difference between what is well and what is badly constructed (or composed).

from An attempt at writing a "Compositionist Manifesto" (pdf)

And, in the first chapter of his book Pandora's Hope, he pokes fun at those who seem to doubt that he believes in reality:

Has reality truly become something that people have to believe in, I wondered, the answer to a serious question asked in a hushed and embarrased tone? Is reality something like God, the topic of a confession reached after a long and initimate discussion? Are there people on earth who don't believe in reality? ... I could not get over the strangeness of the question...If science studies has achieved anything, I thought, surely it has added reality to science, not withdrawn any from it. Instead of the stuffed scientists hanging on the walls of the armchair philosophers of science of the past, we have portrayed lively characters, immersed in their laboratories, full of passion, loaded with instruments, steeped in know-how, closely connected to a larger and more vibrant milieu.

I am kind of smitten with Latour lately, not sure why. He seems to have gone from anthropological studies of science that gleefully undermined naive notions of realism, to philosophizing on a much grander scale, where he attempts to undo the whole structure of modernity and undo mistakes that he traces all the way back to Socrates. And for a French philosopher, he's engaging, and usually quite comprehensible, although sometimes the abstractions veer off into the ether. Unlike many cultural theorists, it seems like he's trying to be clear rather than obfuscatory; he even includes a helpful glossary of his technical terms in the back pages of Pandora's Hope. But I can't quite figure out if he is useful, if these radical reconceptualizations have any implications or applications to how science and/or democracy is done.

One of his goals (as far as I understand it) is extending politics all the way down to the supposedly inanimate world of objects. Or more precisely (and this is exactly the kind of subtlety that confuses his critics), he wants to reveal the existing politics that inhere in our view of nature and society and the relations between them. Science, as it is while it is being constructed, is a set of networks that connect people, institutions, representations, and objects. These networks are often in political contention until one emerges as the victor and we now have Sicentific Knowledge with a capital letter. The connection to objects (aka "nature" or "reality") is crucial, of course, and it distinguishes science from mere politics. Social construction of science is not culture making up arbitrary stories; it's a way to let nature speak, to enter into discourse and society.

But Latour goes beyond this. He's not willing to let science take the role of spokesman for nature (which reduces nature itself to an inarticulate and lifeless thing), but wants nature and objects to not only speak for themselves, but act for themselves. Hence his idea of actants to describe the various human and non-human entities that form networks of knowledge and influence. The BRAF gene (eg) is just as much an actor, with goals and interests of its own, as the scientists to work to uncover its functions.

This sounds a bit crazy, but less so on reflection. From the strict materialist viewpoint, there are no actors and actions, just a seamless web of strictly deterministic causality. The stories we tell about each other involving selves, beliefs, intentions, choices, and actions, are just so many useful fictions imposed over this clockwork reality. Given that, why is it any less legitimate to extend these stories to what we normally think of as the inanimate?

For the record, these issues have been obsessing me for awhile. A section of my long-ago dissertation (ostensibly about educational programming environments) dealt with the conceptual underpinnings of animacy and its role in computation. I've moved on professionally to other things but the ideas won't leave me alone, hence this blog.

The Social Construction of Scientific Sausage

[updated below]

Bruno Latour is a sociologist of science best known for his insistence that scientific facts are constructed rather than discovered. He's widely reviled by scientists and critics of postmodernity; I rather like his work myself and feel that it's usually misinterpreted. Anyway, my guess is he's laughing his Gallic ass off right now at the so-called Climategate affair, which seems like a perfect test case for his point of view.

Thanks to the acts of some hackers, we got to see some politicized science construction in action. The scientific consensus around climate change is revealed to be the result of an ongoing human process, involving a variety of academic politics and a few somewhat shady-sounding practices, including:

• applying pressure to journals to keep conflicting articles out of press
• instructing people to delete emails, possibly in violation of the law
• applying fudge factors to data to make it conform to theory

It appears that at least in this case the making of science is like the making of laws or sausage: not for the squeamish. The import of any of this stuff is up for contention. Is it improper? Unusual, or just normal behind-the-scenes stuff? Does it imply anything at all about the physical facts of the matter? After a couple of hours of scanning the web and looking at some of the email and code that was stolen, I don't know. I have a couple of insights which I may post about later; for now I'm just interested in the sociology.

One thing is clear -- it's pretty clear that even a relatively science-literate layperson can't hope to have a reasonable opinion on this stuff which relies only on raw data and not its suspect interpretations. I suppose if I spent 2-3 months on this stuff I might understand it well enough to be useful, but without that much time to spend I am stuck relying on the judgement of others. And given that, my most important decision is deciding which others to trust. There have been some self-proclaimed experts diving into the material and proclaiming fraud; my impression is that while there is some suspicious stuff there it is impossible to tell exactly what they mean at this stage.

It's somewhat amazing to me how closely the opinions on AGW line up with people's political sensibilities, considering that this is presumably over a physical question which has a matter-of-fact answer (unlike, say, the definition of "conservatism"). But the climate of the past is something that cannot be measured directly, so there is plenty of room for variant interpretation (I suppose a proper sociologist of science would say that nothing is "measured directly"). Almost all the global warming sceptics I come across are also libertarians or right-wing cranks of one stripe or another, who have a vested ideological interest in the denial that there is a vast collective action problem to be solved. And of course, the scientists who support AGW have a vested interest in climate being a significant social problem because it affects their funding. There seems to be no light between the politics and the science, which of course is one of Latour's points.

There have been calls as a result of this for greater openness and transparency in the process of science. The Open Notebook Science movement is already on this. At first glance this seems like a great idea, but I'm somewhat sceptical about the workability, because science is no different than any other field in being a staged performance requiring a backstage zone where the performers can work on the non-public parts of the show (that's a link to Erwin Goffman, another favorite sociologist).

So, channeling my inelegant, Americanized version of Bruno Latour again: here's a situation where there is presumably a physical matter of fact (the actual state of the earth's climate in the past, and the various proxy measurements like tree-rings that we hope reflect it). There are scientists, journals, computer programs, datasets, and funding agencies which are all being deployed to construct competing theories. Only one gets to win, and the battle is enjoined on many levels with little regard for the supposedly neutral gentlemanly rules of combat that the Whig version of science history gives us. That's what the social construction of science is, only now it will occasionally get spread out all over the Internet for all to see. The rubes are shocked; scientists and sociologists are not.

Extras:

Full 61M of released materials.

A bunch of podcasts with a variety of philosophers of science.

Mapping Controversies on Science for Politics is a really interesting-sounding project by a bunch of European labs (including Latour's) to do just what it says.

[[update: I called Latour's attitude fairly accurately:

What I found so ironic in the hysterical reactions of scientists and the press was the almost complete agreement of opponents and proponents of the anthropogenic origin of climate change. They all seem to share the same idealistic view of Science (capital S): "If it slowly composed, it cannot be true" said the skeptics; "If we reveal how it is composed, said the proponents, it will be discussed, thus disputable, thus it cannot be true either!". After about thirty years of work in science studies, it is more than embarrassing to see that scientists had no better epistemology to rebut their adversaries. They kept using the old opposition between what is constructed and what is not constructed, instead of the slight but crucial difference between what is well and what is badly constructed (or composed).

from An attempt at writing a "œCompositionist Manifesto" (pdf) ]]

Reductionist epiphanies

I sometimes troll the libertarians at Overcoming Bias, but there's a lot of insightful stuff there. A recent post on Fake Reductionism was particularly good, I thought. It dealt with the varying emotional experiences related to reductionist explanations, and it inspired me to write down this long-simmering thought.

Maybe it's peculiar to my own thinking, but I have over the years had a number of experiences which I call "reductionist epiphanies". These are moments when all of a sudden I understand some phenomenon that previously either has mystified me, or that I've simply taken for granted. Generally the answer to the mystery becomes so glaringly obvious that I can't quite remember what it was like to not have that understanding, but enough of my previous state is left around that I can at least recall the fact that I had a thunderbolt of revelation.

The suddenness of these insights seems to be a product of having large chunks of knowledge that were previously separated suddenly come together in a more powerful, unified whole. There's an almost audible feeling of pieces clicking into place.

Anyway, here are three questions which have rather simple and elegant answers. Probably you already know what they are. In case you don't, I'm not going to give it away so you can have the pleasure of revelation when you figure it out.

• How does a computer translate instructions into action?
  
  
• Why are there three primary colors, when light is a continous spectrum of wavelengths? How does the three-dimensional representation of color used in computer displays map to color as a one-dimensional wavelength?
  


• What makes a gene dominant or recessive? 
  

The answer to the first question is one that anybody with a CS or EE degree learns in the classroom, but because I had a fairly broken and twisted route through higher education I managed to miss this, and instead figured it out on my own -- an entirely more satisfying experience, so I'm grateful for that.

Everything you know is wrong

If by "you", I mean "science", and by "everything", I mean "most things":
In PLoS Medicine, John Ioannidis says:

There is increasing concern that in modern research, false findings may be the majority or even the vast majority of published research claims. However, this should not be surprising. It can be proven that most claimed research findings are false.

This paper is written in an opaque style, but Alex Tabarrok clears up the murk. Bayes rules!

Turtles all the way down

Really good post at Cosmic Variance about the difference between physics and the sort of woo that goes into quasi-new-age philosophy as exemplified by the What the Bleep Do We Know? movie, quoting philosopher David Alpert:

It seems to me that what’s at issue (at the end of the day) between serious investigators of the foundations of quantum mechanics and the producers of the “what the bleep” movies is very much of a piece with what was at issue between Galileo and the Vatican, and very much of a piece with what was at issue between Darwin and the Victorians. There is a deep and perennial and profoundly human impulse to approach the world with a DEMAND, to approach the world with a PRECONDITION, that what has got to turn out to lie at THE CENTER OF THE UNIVERSE, that what has got to turn out to lie at THE FOUNDATION OF ALL BEING, is some powerful and reassuring and accessible image of OURSELVES. That’s the impulse that the What the Bleep films seem to me to flatter and to endorse and (finally) to exploit - and that, more than any of their particular factual inaccuracies - is what bothers me about them. It is precisely the business of resisting that demand, it is precisely the business of approaching the world with open and authentic wonder, and with a sharp, cold eye, and singularly intent upon the truth, that’s called science.

(Shouting in the original) I think this is very insightful and clarifying, although I draw somewhat different conclusions from it than the science guys.

It seems you can distinguish three very clear positions based on the above:

(1) the foundation of all being is the physical laws of nature as explicated by science. Intelligence is a contingent result of these mechanical processes, an epiphenomenon. This is the naturalist position.

(2) the foundation of all being is intelligence, which is somehow prior to the material universe. This is the religious position, seen in intelligent design theory and theology in general.

(3) there is no foundation, or none that we are have access to. This is the antifoundationalist position, seen in various philosophies (pragmatism, postmodernism, etc).

The appeal of quantum woo is that it seems to bridge the gap between (1) and (2). Certain other weird ideas in physics (anthropic cosmology) also seem to be somehow letting us have our cake (physics) and eat it too (give intelligence a central role). ID is what happens when believers in (2) try to ape the methodology of position (1), with generally laughable results.

Where am I? Well, I am in the position of a (1)ist with strong urges towards (2) deviationism. Humans are not the center of the universe, god knows, but they are at the center of something. As I said in the comments:

there is the physical world, which the hard sciences study, and the mental/social/cultural world, which is indeed part of and emergent from the physical world but is also in some sense a separate sphere, striving for independence from its physical substrate. This is the world people spend most of their time in, even physicists. In this world, people ARE the central and most important feature.

The relationship between these two worlds is hard to understand and What the Bleep sounds like an example of how NOT to think about it clearly.

Position (3) seems like giving up, or resigning yourself to linguistic gameplaying.

I just thought of position (4) -- there is a foundation, but it is not captured by either science or theology. I'm going to call this Taoism, but that is not its name.

One Thing About Science

So the Guardian asked a bunch of scientists "What is the one thing everyone should learn about science?" I guess I would pick Darwin's theory, as a number of respondents did, because it can be instantly understood yet the implications are vast and not always obvious. You couldn't teach someone quantum physics or even Newtonian physics in a hour, but you could get them bootstrapped into Darwin in that time.

A more original response: science is about making powerful abstractions about the world. Part of doing science is knowing what facts to pay attention to and which ones to discard in order to make a powerful theory. Physics is possible because scientists were able to learn how to ignore friction and air resistance and pay attention to more fundamental properties. Abstractions are powerful and necessary but they leave stuff out, so don't mistake your theories for reality.

Here's a response that amused/annoyed me:

John McCarthy Emeritus professor of computer science at Stanford University, and inventor of the term 'artificial intelligence'

Find the numbers, and compare them. As the physicist Lord Kelvin said in 1883, in a lecture to the Institution of Civil Engineers, "when you can measure what you are speaking about and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind".

It annoyed me because McCarthy is the inventor of Lisp and cofounder of symbolic AI, a school which mostly ignored numbers and quantitative techniques. Maybe it's unfair to McCarthy, but I blame him and his fellows for letting me feel I could neglect numerical techniques for the longest time, and now I feel like I'm scrambling to catch up.