Tangled Bank #35

Cognitive Daily is hosting the latest Tangled Bank. There were a lot of contributions dealing with pseudoscience, especially evolution vs. creationism, which don't really fit into the Tangled Bank theme, so they have been posted separately, at Word Munger. Happily, one of our posts, about how Darwin learned not to be too quick about anomalous data, has been included there. It is pleasant company.

The One Great Truth

I've just finished Alice Flaherty's The Midnight Disease, which is a fascinating review of current ideas about the neurobiological roots of creativity, particularly writing. Flaherty suffered from hypergraphia, or compulsive writing, and became interested not only in its neurological and psychological roots, but also in comparing it to the apparently opposite situation of writer's block, and to issues of creativity in general.

One little gem is embedded in the discussion of the impact of drugs on creativity. It is a long quote from Oliver Wendell Holmes, Sr., who is now known mostly for having been the father of the great Supreme Court justice, but who was a prominent public intellectual of the late 19th century. (He was especially known for The Autocrat of the Breakfast Table, a quasi-fictional setting of his conversational opinions on virtually everything.)

Holmes sought to understand the effects of ether by trying it himself. In this excerpt from Mechanism in Thought and Morals: An Address Delivered Before the Phi Beta Kappa Society of Harvard University, June 29, 1870, he describes the result. As I move from blog to blog, seeing what each purveys as the one great truth one which we must order our lives, I will try to keep this truth in mind:

The one great truth which underlies all human experience and is the key to all the mysteries that philosophy has sought in vain to solve, flashed upon me in a sudden revelation. Henceforth all was clear: a few words had lifted my intelligence in the level of the knowledge of the cherubim. As my natural condition returned, I remembered my resolution; and, staggering to my desk, I wrote, in ill-shaped, straggling characters, the all-embracing truth still glimmering in my consciousness. The words were these (children may smile; the wise will ponder): ‘A strong smell of turpentine prevails throughout.'

NRC Report on High School Science Labs

Last week, the National Research Council (an arm of the National Academy of Science) released a report on high school science labs. The story has been heavily covered by the mainstream media (e.g., the Associated Press story in USA Today), but has been mentioned in only a few blogs (e.g., in Education Wonks). I’m a little surprised, because many science blogs have been complaining about the state of science education in the US, particularly as it relates to understanding the theory of evolution. But it is in the hands-on experience, in the lab class (one hopes), that students get a real appreciation for how science is done.

The NRC report says that, on average, a high school student spends one class period per week having a “laboratory experience”, defined as direct interaction with the world or data drawn from it. The work is typically narrow in scope, often concentrating on the mechanics of lab work, and is seldom integrated into the classroom lesson plan. Ideally, the labs should follow the “four established principles for effective science instruction”:

• Design science lab experiences with clear learning outcomes in mind
  
• Thoughtfully sequence lab experiences into science instruction
  
• Integrate learning science content and learning about the processes of science
  
• Incorporate ongoing student reflection and discussion
  

That this doesn’t happen is blamed mostly on inadequate teacher training and (to a lesser extent) on the organization of schedules, space and resources within the school.

The report is available for on-line reading at the National Academies’ publication site.

I’ve had some experience with this problem, from having been a teacher (undergraduate physics, both for science and non-science majors) and from having been on a citizen advisory committee for the County School Board. I have a lot of respect for the preparation that K-12 teachers go through to give their kids the best instruction they can. Thus, I would look closely at the factors that dominated the concerns of the school teachers with whom we worked:

• There were usually only enough experimental setups to support groups of at least four students working together. That makes it hard to give each one a chance to get the full experience within one class period.
  
• Even if the classes are of modest size (on the order of 20 students), it’s hard for one teacher to circulate among them to provide each student with guidance and interpretation as they work.
  
• Most high school teachers have no support, so they are responsible for setting up the apparatus before class and taking it down afterwards. It is hard to find the time to do that, especially if the space has to be used by other classes during the same day.
  
• The crush for classroom space is so strong that there are usually no dedicated storage rooms for the experimental setups. Instead, they are either compressed into a closet, or into shelves on the side of the room. That makes it hard to maintain them, to check that they are working, and to protect them from the prevailing entropic forces.
  

It seems to me that this is one of those problems that can be solved by throwing money at it. Pay for an addition to the school to provide safe, organized storage space. Pay for more experimental setups. Pay for a teacher aide to be responsible for the pre-class setup and checkout and the post-class cleanup. After we do the obvious stuff, then we can see what else needs to be addressed with the teachers.

Predictions by Darwin – III – “Cross your fingers” and “I should have thought of that”

I have been looking at scientific predictions that are outside the popular image of critical experiments, for which the model was Einstein and the 1919 solar eclipse expedition. I argued that the future-directed form of this model distracted from the very useful concept of retrodiction, which is often the form of validation used when new theories are presented. I would claim that there are other valid means by which scientists become confident about theories.

Part of the problem of the critical experiment model is that it focuses on theories that are posed primarily to address anomalous or problem data. But more frequently, scientific hypotheses are extensions, augmentations, or more modest revisions of the large-scale theoretical structure within which the scientist is working, not addressing problem data. Some hypotheses only address a subset of the empirical problems with the theory. And some are only posed for fundamentally philosophical, non-empirical reasons, to mesh with the philosophical leanings of the scientist, or, most important, to make the structure more productive of new empirical questions

Are we only talking about background science, rather than revolutionary science as defined by Kuhn? I think not. Consider the case of Copernicus and the heliocentric theory, which was what Kuhn used as the archetype of revolutionary science.

While there were empirical problems with Ptolemaic theory (predicted planetary positions that weren’t right), the heliocentric theory didn’t resolve them. To a large extent, this was because Copernicus retained the model of circular orbits. It wasn’t until Kepler’s introduction of elliptical orbits that there was any significant improvement. Indeed, Copernicus wasn’t particularly concerned about collecting data to support his theory. Historical studies indicate that what observations he made were done to determine specific orbital parameters.

What the heliocentric theory did was provide a framework that Copernicus considered more sensible than the geocentric theory. He was much more comfortable putting the big Sun at the center because it was so much bigger than the Earth, and so much more glorious. There are plenty of arguments suggesting that he was a closet follower of Hermetic philosophy, which virtually deified the Sun. It also relieved him of having to think of the vast celestial spheres spinning at incredible speed, in order to account for their diurnal motion. Physical plausibility was as important as “saving the appearances”.

This importance of physical plausibility was later emphasized by Kepler in the lengthy letter that he wrote in rebuttal to Ursus’ De astronomicis hypothesibus. This was a serious treatise on the meaning of astronomical hypotheses and the methods one should adopt for deciding among them. It argues that hypotheses must both predict phenomena accurately and be physically plausible. It was only published in the 1858 volume of his works, but there is a translation and analysis of it in Nicholas Jardine’s The Birth of History and Philosophy of Science: Kepler’s A Defence of Tycho against Ursus with Essays on Its Provenance and Significance (Cambridge University Press, 1984). There is also a very nice discussion of these arguments in Owen Gingerich’s recent book, The Book Nobody Read (Walker & Co., 2004).

Of course, what was plausible for Copernicus was not plausible for others. For them, it was physically plausible that, if the Earth were rotating, they should be able to feel the motion. It was physically plausible that falling objects would, in their natural motion, fall toward the center of the universe, which meant that the Earth needed to be at that center. It also seemed implausible (although here I think we fairly part from the physical) that the sinful Earth would be a part of the celestial planetary scheme.

So if it wasn’t an improvement in predicting planetary data or a clear triumph of physical plausibility that made heliocentrism so attractive to Copernicus, then what was it? The current thinking – an argument actually made decades ago by Gingerich and others – is that Copernicus preferred the heliocentric model because it expanded his explanatory reach. It enabled Copernicus to calculate things that couldn’t be calculated before, such as the order of the planets. In the Ptolemaic model, the order was decidedly arbitrary; there was no particular reason for Jupiter to precede Saturn. Mercury and Venus were figured to be between the Earth and the Sun because they were never seen far from the Sun. But the rest was pretty much made up. In the heliocentric model, though, Copernicus could work out a geometric proof that required one particular order.

Here is an example, then, of revolutionary science being developed more for philosophical preference (increased explanatory power and a tip of the hat to Hermes Trismegistus) than to account for problem data. Predictive power in such cases takes different forms.

Copernicus did not make predictions. He did, however, note empirical problems that arose as a result of the new system, which would have to be resolved somehow. For example, even Aristotle had known that, if the Earth goes around the Sun, then we ought to see an annual oscillation in the apparent positions of stars (parallax). Copernicus figured that the fact that astronomers of his day didn’t see parallax had to have some answer, but he went ahead without having it. (He figured that it was some combination of imprecision in the data with a much greater distance to the stars than previously assumed. We now know that the answer is mostly the latter.) As a prediction, this is more like crossing your fingers. It has the form: “Such-an-such an empirical problem must eventually be explainable by some other observation or experiment, the nature of which I can guess at, but the answer I can’t.” (It is often the case that predictions like these aren’t even raised by the originator of the theory, precisely because they are problems that she hopes to deal with on her own terms later on.)

Another odd form of prediction is rooted in “virtual” empirical issues. Later investigators, not engaged in testing the theory per se, discover new phenomena that are surprisingly well explained by it. These predictions are of the form “If Copernicus is right, then we should have expected to see this, although we didn’t realize it until now.” An example is the fact that, as Galileo discovered, Venus shows phases like the moon, but Jupiter doesn’t. Of course, if Copernicus had thought it through, he probably could have predicted such an effect, and Galileo would then have been performing one of those cinematic critical experiments. But that’s not how it happened, and not how it usually happens.

In my next post in this series, I intend to identify some very significant Darwinian “predictions” that fall in these last two categories.

Predictions by Darwin – II – Retrodictions

I think that the image of predictions as critical experiments in the mold of Einstein and the 1919 solar eclipse expedition lead us to ignore other valid means by which scientists become confident about theories.

My guess is that the most common form of scientific prediction is retrodiction. These are predictions that take the form: “If the theory is correct then certain events will have had to have happened, so we’ll look for evidence that they did.” These should be familiar to fans of CSI: “If the gun was fired at close range, then gunpowder residue should have been left on the clothing, so let’s go check.”

Much of what happens when a theory is presented and justified is retrodiction. Indeed, it is often presented as a comparison of the retrodictive powers of the new theory and its competitor. Thus, Darwin presents data on the geographic distribution of species, and notes that, for similar groups, the degree of similarity is higher for groups that are closer in space or for which geographic barriers are lower. He then says that this is a logical result if these groups are slowly diverging from a common ancestor, as the more distant are more likely to have preserved variations from the ancestor rather than breeding with and blending back in to the central population. Recast into the retrodictive form: if the theory is correct, then such divergence will have had to have happened, and there it is. On the other hand, if the different groups were all specially created, it would be hard to see why such a correlation of divergence with distance should occur. Thus, in the comparison of retrodictions, descent with modification wins over special creation.

Of course, retrodictions included in the initial justification of the theory don’t have the punch of predictions resolved by future experiments. Again, we have that cinematic bias. But the fact is that Origin of Species is full of successful predictions, in the form of retrodictions, and given the fact that Darwin was pretty honest about including data that were clearly problematic for the theory (most famously the matter of the completeness of the fossil record), that we should not be discounting all these successes.

There are other forms of scientific prediction that relate to Darwin's work. I shall discuss them in the next post in this series.

The Sound of Medieval History Rocking

I'm not sure why I follow Blogenspiel. Partly, it's because I think history in general is interesting, and I'm not very discriminating about periods. Partly, it's because Another Damned Medievalist has such a sharp eye for the academic world. And, of course, the fact that ADM's attempt to secure a place in that world has been sheer melodrama tends to draw in the reader.

Whatever. None of that matters now.

In this post, ADM unleashes the ur-historian inside, and reveals the passion and perplexity of doing history. It's a glimpse of the real joy of scholarly work. It ought to be assigned reading for every student everywhere, to help them understand that the life of the mind can be awesome.

Go. Read it. And commit to memory that last line:

[I]t's good someties to remember that what we do, we often do because it's cool and geeky and because we, or at least I, love the fact that there is so much to argue about -- every question a hydra, every answer a target for someone else.

Predictions by Darwin – I – The Popular Image of Scientific Prediction

A persistent claim in the creationist literature is that evolution is not scientific because it doesn’t make testable predictions. I think that there are responses that can be made to this which have not been made a lot, but to justify them, I need to lay some foundations. In this post and the next, I’ll discuss what I think the problem is, and then I’ll try out some predictions that I think should be cited.

Of course, the argument that the neo-Darwinian theory does make testable predictions has been made at length, e.g. by Douglas Theobald. In addition to providing (considerable!) detail on evolutionary evidence, Theobald sketches out the basic form of the scientific method and the characteristics expected of testable hypotheses. This is all fine, but I fear that it hews to a narrow view of how scientists come to be confident about their theories, in that it emphasizes the concept of the critical prediction: The scientist, having formulated a hypothesis to account for the existing empirical data, including the problematic observations that necessitate a new theory, predicts that a certain critical experiment will have a particular result. Confirmation is thus dramatic – even cinematic. (Think of Paul Muni as Pasteur, returning to the farm to see whether his vaccinated sheep have survived.)

The modern image of the critical experiment is, I think, dominated by Einstein’s prediction of the bending of light by the gravitational field of the sun.

Einstein developed the General Theory of Relativity over the period 1907 to 1915. Although the argument was basically theoretical, justified as putting gravity in the form of a field theory, he did highlight its power for the explication of anomalous data. In particular, he was able to account for the perihelion advance of Mercury, itself a very public problem with the purely Newtonian model of the solar system. Mercury has a more elliptical orbit than most planets, and for some time it had been known that its closest approach to the Sun was shifting. (Imagine standing on the Sun and marking where in the zodiac Mercury appears to be at perihelion. That apparent position marches along the zodiac, at the rate of 0.04% of a constellation per century.) To account for this, astronomers had postulated the perturbing influence of a planet closer to the Sun, leading to well-publicized searches for (and some erroneous discoveries of) the elusive planet Vulcan.

In general relativity, the perturbation of Mercury’s orbit comes from an effective “tug” by the Sun, reflecting differences between Newtonian and Einsteinian gravity that appear when the gravity field is strong. Because Mercury is so close to the Sun and has such an elliptic orbit, it probes the gravitational field of the Sun more deeply than any other planet, and experiences the relativistic deviations from Newtonian gravitation more than the others.

Einstein’s 1915 paper included another observational test. In both Newtonian and Einsteinian theories, light rays passing close to the Sun will be deflected, so that the position of distant stars will appear to shift as the Sun passes by them. Of course, the only time you can see stars very close to the edge of the Sun is during a solar eclipse. Thus, there was a major expedition to plant instruments in the path of totality of the 1919 solar eclipse, on an island off the west coast of Africa, to measure the effect. The famous Arthur Eddington, of Cambridge University, was in the lead. Einstein predicted a deviation twice the Newtonian value (after correcting a mistake in the 1915 paper). Lo and behold, that was the result Eddington found and trumpeted about the world.

Most physicists, though, were already convinced by the 1915 paper. And Einstein himself was not so hung up on the experiment. When asked how he would feel if the confirming deviation was not measured, he said, “I should be sorry for the dear Lord. But the theory is correct.”

In my next post, I will talk about less dramatic, but more common, forms of scientific prediction.

Addendum

There's a good review of the Eddington expedition in Donald Fernie's Marginalia column in American Scientist.

DoD Lessons for NASA’s Vision for Space Exploration

Defense Industry Daily has highlighted the recent (July 12) testimony of Robert E. Levin, Director for Acquisition and Sourcing Management at GAO about the factors that lead to endemic cost overruns in the acquisition of space systems.

The root problem is simple:

"There is a widespread belief among DOD and other officials involved with space programs that DOD starts more programs than it can afford in the long run, forcing programs to underestimate costs and over-promise capability and creating a host of negative incentives and pressures.”

Specific issues cited by officials interviewed by GAO included:

Because programs are funded annually and priorities have not been established, competition for funding continues over time, forcing programs to view success as the ability to secure the next installment rather than the end goal of delivering capabilities when and as promised.

Having to continually "sell" a program creates incentives to suppress bad news about a program's status and avoid activities that uncover bad news.

When combined with the high cost of launching demonstrators into space, the competition for funding often encourages programs to avoid testing technologies in space before acquisition programs are started.

NASA’s Vision for Space Exploration has already raised worries that it plans far more than could be reasonably encompassed within the projected space budget. Given that the pressures of overspending tend to fall hardest on science programs (see, for example, complaints by the American Astronomical Society and the American Geophysical Union), it is hard to feel sanguine about the future of space-based science.
NASA does come out relatively better than DoD with respect to one of the GAO’s concerns:

… [w]hen faced with lower budgets, senior executives within the Office of the Secretary of Defense and the Air Force would rather make across-the-board cuts to all space programs than hard decisions as to which ones to keep and which ones to cancel or cut back.

NASA has generally taken the opposite approach, cutting whole programs rather than “sausage slicing” the budget. Administrator Mike Griffin has committed to upholding that policy.

And Now We Know

From the New York Times, August 3, 2003, under the byline of Elisabeth Bumiller:

At the White House, where intelligent design has been discussed in a weekly Bible study group, Mr. Bush's science adviser, John H. Marburger 3rd, sought to play down the president's remarks as common sense and old news.

Mr. Marburger said in a telephone interview that "evolution is the cornerstone of modern biology" and "intelligent design is not a scientific concept." Mr. Marburger also said that Mr. Bush's remarks should be interpreted to mean that the president believes that intelligent design should be discussed as part of the "social context" in science classes.

...

Mr. Marburger said it would be "over-interpreting" Mr. Bush's remarks to say that the president believed that intelligent design and evolution should be given equal treatment in schools.

But Mr. Bush's conservative supporters said the president had indicated exactly that in his remarks.

One would think that part of the job of the White House Science Advisor would be to advise the President about what is science and what ain't. It doesn't seem that Dr. Marburger is up to the task.

What Will Marburger Do?

In response to today's mind-boggling news that President Bush has advocated the teaching of intelligent design as part of the biology curriculum, one can only wonder what Presidential Science Advisor John Marburger will do.

Recall that, at this year's conference of the National Association of Science Writers, Dr. Marburger was explicit on the subject of ID. As Chris Mooney reported in The American Prospect:

Marburger fielded an audience question about "Intelligent Design" (ID), the latest supposedly scientific alternative to Charles Darwin's theory of descent with modification. The White House's chief scientist stated point blank, "Intelligent Design is not a scientific theory." And that's not all -- as if to ram the point home, Marburger soon continued, "I don't regard Intelligent Design as a scientific topic."

So what would a conscientious, principled advisor do when he is in clear conflict with his boss? Resign? We'll see.

A Lesson on Isolated Anomalies

I’ve finally been reading Volume 1 of Janet Browne’s biography of Darwin (Voyaging), having already read Volume 2 – partly because I was more interested in his later life, and partly because it was a gift. If anything, the first volume is even better at rendering the personages of Darwin’s life as true personalities, flawed and likeable.

I was particularly struck by one episode (on page 141). The year was 1831. Darwin has been at Cambridge, developing his interests in natural history. Darwin had a plan for going to Tenerife, to follow the footsteps of the German scientist/explorer Alexander von Humboldt. His mentor, John Henslow, encouraged him to develop some understanding of geology beforehand. Darwin was not particularly interested in the subject, having been unhappily bored with Robert Jameson’s lectures during his earlier tenure at Edinburgh. But Henslow, a mineralogist before he started teaching botany, made the case, and arranged for Darwin to accompany the renowned geologist Adam Sedgwick on a between-terms trip through Wales.

Darwin decided to impress Sedgwick, practicing his geology in the fields near his family home in Shrewsbury before Sedgwick came to meet him that August. Then before they set out, Darwin engaged him with a discussion of the nearby gravel pits. In particular, he told him that a day laborer had made the very surprising discovery of a tropical shell in the gravel. To his surprise, Sedgwick laughed and said that it had to have been thrown there, as there was no evidence for tropical species living at that location at the time the gravel had been laid down.

Darwin was disappointed that Sedgwick was not as delighted by the strange fact of the shell as he was. But recalling it later, Darwin said,

“Nothing before had ever made me thoroughly realize, though I had read various scientific books, that science consists in grouping facts so that general laws or conclusions may be drawn from them.”

Browne adds,

“[T]here must be a great deal of mutually supportive material for scientific theories of all denominations. Once such theories were established, it took more than an isolated shell to change them.”

This is a point that seems to get lost when discussing “teaching the controversy” about evolution. An important part (perhaps the important part) of science education is not teaching the facts but teaching the approach, teaching the way scientists approach the world (and how that differs from the way of the historian, of the artist, etc). That ought to include an appreciation for the fact that scientists use reasoned judgment to interpret the panoply of empirical data, and to determine which anomalies are problems and which are likely not. The development of that judgment – through the patient process of becoming familiar with the scale and interrelatedness of the evidence – is an important part of a scientist’s training.

I think part of the intensity with which scientists respond to challenges to evolution is rooted in the investment made to develop that scientific judgment. The public recognizes the existence of diagnostic authority, based on years of medical training. But they don’t seem to credit a similar authority in scientific training.