« There is a season. | Main | Pictures of plants. »

12 October 2006


David Moles

Crooked Timber is full of screaming arguments now about whether being able to do the math is a necessary or sufficient condition for "understanding the physics". Unfortunately Searle's goddamn Chinese Room has come up again....


I don't have the stamina to check out the Crooked Timber discussion, but has anyone raised the distinction between understanding the physics and being able to produce new work in the field? Because it makes perfect sense to me that there would exist a level of proficiency where one does actually understand the concepts as they exist, but cannot contribute to the advancement of the field. (Like when you're learning a foreign language, and you reach that point where you can understand what people are saying but can't actually produce new speech.)

Dan Percival

Yes, the "but can he contribute?" distinction seems to be the second-most-popular after "can he do the maths?"

I bailed out around comment 70-or-so, but someone pointed out that defining whether or not someone was contributing to the field can be a dodgy enterprise as well.

Oh, and here's this quote from comment #63: "[contribution to the field] is a definitively sociological test, and a rather more radical one than most sociologists of science would be comfortable with."

Jackie M.

I got the impression some of the questions on that test they gave Prof. Collins involved synthesizing new ideas. Anyway, while it's clear that he doesn't know the math -- or the top from the bottom of a power supply -- it wouldn't surprise me in the least if Prof. Collins is perfectly capable of having new and very testable ideas about gravitational wave theory and/or experimentation.

Furthermore... you know how physics is hard? Even for the lion's share of physicists? It also wouldn't surprise if, at this point, Prof. Collins would be able to complete a four-year physics degree, math and all, in two-and-a-half years. At MIT. While reading sociology papers in half of his classes.


This is a weird article. First of all, if you study something for thirty years, dude, you should learn something. That's not fake. The idea that "understanding" is the same as "doing the math" seems like--well, actually the Turing Test seems to be a pretty good analogy, what with the way it sounds good but turns out not to be a meaningful test at all.

But the other thing that I think is missing from the article, and apparently the discussion, is the way the form and genre of science writing is so defined and so specialized and so derivative (in a formal, not "unoriginal" sense). That has to be part of the answer to the question about being able to sound like a physicist vs. being a physicist. Because working with those forms is actually part of being a professional.

The thing about AI, for instance, is that whenever you make an objective test, you get an obviously non-conscious computer to pass it, and then you say, "Oh, well, then that must not be what consciousness is." I feel like you could do that with "what a physicist is" here.


Belated comment on the sociologist's knowledge of physics: according to this Slate article, he wrote the questions himself, so it's not exactly a surprise that he was able to answer them.

Jackie M.

That would certainly change my mind about Prof. Collins' ability to "synthesize new ideas." It would still be an interesting experiment... but not necessarily one demonstrating what the experiment sets out to demonstrate.

That said, I'm confused: at the top of page 21 of the paper Collins et al. submitted to History and Philosophy of Science, the claim is made that:

"in one or two cases, such as Question 5 in Table 2, Collins had no choice but to think the answer through since he had not encountered such questions before."

Which implies that he was not the author of the questions?


Thanks for the link; I hadn't seen the paper itself. On page 24 it says

Those composing the questions for the gravitational wave imitation game were asked to avoid mathematics.

which indicates that the Slate article was wrong. The paper then says

Were calculation or algebraic manipulation needed to answer the questions, Collins would have had no chance of passing the test.

This seems to me a peculiar condition to impose on an imitation game.

Jackie M.

But at least that takes us back to the original debate.

I suspect the Slate article mistake is a conflation of the original test with the secondary experiment mentioned on pg. 23:

"Collins also tried taking the role of both interrogator and judge and found he could use both questions he generated himself and those set by other gravitational wave scientists to distinguish gw physicists from other physical scientists who were not gw specialists and from Evans; he could do this without fail at confidence level 4 based on the technical mistakes or lacunae evident in the non-specialists answers."

Dan Percival

...which leads to, I think, a patch for the "it's just a trick" hole in the Turing test (and the application here to domain expertise): a successful candidate for "expert" status should be able not only to convince a known expert that they are so, but also be as good as other known experts in distinguishing experts from non-experts?

The comments to this entry are closed.

Susan Marie Groppi

  • Susan Marie Groppi is a historian and an editor, currently living in Berkeley, California.


  • www.flickr.com
    This is a Flickr badge showing public photos from Susan Groppi. Make your own badge here.
Blog powered by Typepad