Experiments Are Not Afterthoughts

There’s been a bunch of talk recently about a poll on quantum interpretations that showed physicists badly divided between the various interpretations– Copenhagen, Many-Worlds, etc.– a result which isn’t actually very surprising. Sean Carroll declares that the summary plot is “The Most Embarrassing Graph in Modern Physics, which I think is a bit of an overreaction, but not too much of one. I do strongly disagree with one thing he says in explaining this, though:

Not that we should be spending as much money trying to pinpoint a correct understanding of quantum mechanics as we do looking for supersymmetry, of course. The appropriate tools are very different. We won’t know whether supersymmetry is real without performing very costly experiments. For quantum mechanics, by contrast, all we really have to do (most people believe) is think about it in the right way. No elaborate experiments necessarily required (although they could help nudge us in the right direction, no doubt about that). But if anything, that makes the embarrassment more acute. All we have to do is wrap our brains around the issue, and yet we’ve failed to do so.

My problem is with the emphasized words, which are emphasized in the original. This plays into a pet peeve of mine, which I’ve ranted about before, namely the idea that experiments are somehow an afterthought, just cleaning up the loose ends once theorists have done the hard work of thinking about things.

This is emphatically wrong. Experiment is at least an equal partner in this, and every other scientific question. If we ever do determine that there is One True and Correct Interpretation of quantum mechanics, it will be because that intepretation produces makes concrete predictions that are testable by experiment. Full stop.

Everything we know in quantum foundations– and science generally, but wer’re talking particularly about quantum foundations here– is ultimately grounded in experiment. The reality of photons? Confirmed by experiment. The wave nature of matter? Confirmed by innumerable experiments– I particularly like this one. All that Alice and Bob stuff that has theorists tying themselves in knots about “firewalls”? Confirmed by experiment.

Experiment is not now and never will be some ancillary activity that maybe provides a “nudge” or two for the deep thinkers. Without experiment, nothing is ever settled. A subject where you try to settle questions only by thinking really hard is indistinguishable from theology, and it never ends.

In fact, there are times when over-reliance on deep thinking is part of the problem. Sean argues that the lack of consensus on interpretations is embarrassing because quantum mechanics has been established for eighty years now. He dates this from John von Neumann laying out the mathematical foundations in 1932, which is a little ironic, because about thirty of those eighty years of inaction can be laid at von Neumann’s feet. When he laid out his formulation of quantum mechanics, von Neumann asserted that hidden-variable theories were ruled out mathematically, and his reputation was such that most physicists regarded this as a settled question on that basis. The problem is, he was flat wrong on this point, relying on a mathematical theorem that didn’t actually say what he claimed it did.

The question was eventually re-opened in part by people thinking about it the right way– in particular David Bohm, and then John Bell. But what blew it open into an active field of research, and prompted all the exciting research in quantum information and so on was experimental work. First John Clauser and then Alain Aspect did experiments that showed unequivocally (well, mostly) that quantum reality is not compatible with local realism. All of the dramatic progress since the early 80’s is a direct result of people realizing that these questions are experimentally accessible.

So don’t be so quick to discount the role of experiment in this question. If it’s ever going to be settled, it will ultimately be settled by experiment, just like everything else in science.

(Now, there are a couple of dodges available to argue that I’m being too harsh on this. One would turn on the “elaborate” modifying “experiments.” While I agree that quantum optics experiments aren’t as expensive as particle physics (by about three orders of magnitude), if you think they aren’t “elaborate,” I will just point you to the ThorLabs catalog and invite you to duplicate Anton Zeilinger’s experiments. And if we’re going to argue about that, I’m also going to object to that first sentence, so let’s not go there, OK?

(The other out would be to say that “thinking in the right way” includes thinking about experimentally testable consequences of interpretations. But if you think you can hold that while also relegating experiment to the subsidiary role of providing the occasional “nudge” to theory, well, that’s an awfully narrow needle to thread.)

17 thoughts on “Experiments Are Not Afterthoughts

  1. I suppose there is a difference between Ptolemy, Copernicus and Kepler that is not much connected to experiment, which are largely different models for more-or-less the same experimental data (that is, all models depend on there being data *to* model, but sometimes there are different models for more-or-less the same data). Even Newton didn’t have that much more experimental data to work with; new data mostly came after him, when it was partly his theories that gave physicists the tools to imagine and construct new apparatus. Without Newtonian mechanics and Maxwell’s equations, it would be difficult to build Arecibo, say, or even to imagine *any* reason to build such a thing, amongst many other tools.

    “Think about it in the right way” could be construed as the construction of more parsimonious models, but perhaps it’s just “theorists should create something that captures the imagination”. It’s in part a theorist who fires up experimentalists to want to create what they can do. Experimentalists partly need theorists to give them ways to imagine and construct new tools. I hope this doesn’t slight experimentalists — if I didn’t think they were worth watching and listening to I wouldn’t read this blog, and I even think the creativity of engineers is crucial — but there’s give and take.

  2. I definitely agree that there’s interplay between theory and experiment– Clauser and Aspect wouldn’t’ve known what experiments to do without Bell’s theorem pointing in the right direction. Neither can function entirely without the other.

    But if you’re going to talk about finally settling questions, in the end it will always come down to experimental tests.

  3. Kepler had Brahe’s data, which was a qualtitative improvement on previous quantifications. Newton most certainly had access to new experimental data, notably that of Kepler and Galileo, but Halley! Newton also did his own experiments, eg in optics.

  4. I think that “thinking the right way” entails giving crisp operational meaning to ideas under consideration, something I think that field can use (especially when words like “real” and “exist” start making an appearance). Thinking about experimental consequences is an efficient way to do that, so maybe you and Sean are not in such a big disagreement here.

  5. @Peter: Your argument about Ptolemy vs. Copernicus might be granted, but by the time Kepler started looking at things, Galileo had disproven Ptolemy’s model–Galileo, using the recently invented telescope, observed phases of Venus that Ptolemy’s model predicted would never occur. And as Steinn points out, Kepler had new data that were not available to Ptolemy or Copernicus–data which in fact disproved the notion (adhered to based on philosophical arguments) that the orbits of celestial bodies were necessarily circular. Without Kepler’s data analysis, Newton probably would not have come up with the inverse square law. Without Newtonian mechanics, people would not have found the discrepancies in the orbit of Uranus which were the clue to finding Neptune, or the precession of Mercury’s perihelion which became an early test of general relativity.

    The main reason it’s worth the effort to think about these various interpretations of quantum mechanics is that it is the only way to come up with testable predictions that distinguish among them. Only then can you do the experiments to settle the issue. We probably don’t know yet what experiments might resolve the issue. We may not know in my lifetime what experiments we should do. But the proof will be experimental.

  6. Experimentation and theory are both part of any reasonable definition of science. They are both essential. I don’t read the Carroll quote as contradicting this, though; it seems to me that he’s talking about how people view his field. And his final word is that experiments will, likely soon, solve the problem he observes.

  7. Experiment tells you about the actualized possibility you happen to find yourself in. QM is that theory which describes the totality of all possibilities and how to expect them given your uncertainty about the present empirical records/causal story. The source of all possibilities of experimental outcomes cannot change according to experiment! QM is not just some other mechanics! It is the statistical correlation that follows from pure logic, and so, for example, the tautological modal realism that it re-motivates is the correct interpretation. Whoever does not understand this is simply wrong, period.

  8. Hi,
    Funny coincidence: I attended Carroll’s stimulating talk in Nottingham. I don’t think he disagrees with you about the importance of experiment. After all, he said we need experiments to answer questions about supersymmetry. But the problem of the interpretation of quantum mechanics is one of … interpretation. That means thinking. Very difficult – Carroll’s talk was not simple, and I confess I got a little lost. Maybe some additional experimental input would help, but I don’t know what.

  9. I wonder if this “primacy of theory” attitude can be traced back to how physics is taught? When I supervised undergraduate physics labs many years ago I was always struck by how the students would go to great lengths to explain how the experiments could be “improved” so that the results were more in line with theory. No-one ever suggested modifying the theory so that it better described the experiment! So lots of incredible schemes for eliminating friction, for example, but no attempt at adding a simple friction term to the equation of motion!

  10. Just a brief comment that of course I agree that a history of observations and improvements of method over several centuries is complex, nonetheless, insofar as Ptolemy’s models can be understood as a Fourier analysis of a 3n-coordinate time series, his system can accommodate any data (there’s a separate issue of converting bare experimental data of azimuth and altitude and time and place of observation into a Euclidean or other coordinate system). The Copernican system introduces no more than a different coordinate origin, so it is still capable of describing anything (however the mathematics of converting bare experimental data to theory coordinates is slightly more difficult). The move to ellipses introduces a different order of mathematical inverse problem, which is ratcheted up a notch by the sophistication of the Newtonian approach. Adopting the later theoretical models is a purely theoretical shift that is prompted by sophisticated ideas of simplicity, tractability, and beauty, etc., but is not *necessitated* by data. The theoretical shift, however, influences what new data it seems worthwhile to focus on gathering.

  11. @Hamish — indeed the spherical cow is queen of the school physics lab, and research labs perhaps are the same, but applied physicists and engineers tend to work with more realistic models, insofar as they have to use real cows, which resist sphericalization.

  12. Chad,
    On the main point, that science needs both theory and observation, I agree wholeheartedly. But, I have to point out a couple of problems in detail.
    First, contrary to what you say, some of what the black hole people say about Alice and Bob is definitely not supported by experiment. Quantum computing types wish we could create monogamous entangled pairs, but real particles share their entanglement with all comers. And quantum computing types dream of creating computers without information loss, but that’s not the reality we wake up to.
    Second, while the evidence for nonlocality is incontravertable, theorists have been extremely reluctant to embrace it, saying instead in effect “We don’t have to believe in nonlocality because we don’t believe in reality.”.
    You yourself say that local realism fails, which suggests that a local theory might still work if it’s sufficiently unrealistic.
    Before I accuse you of substituting gibberish for reasoned argument, I have to acknowledge that I have seen “reality/realism” used twice with sensible definitions, and so I’m asking you just what you mean, by “local realism”.

  13. A very nice post, Chad. I would like to add just one more thing. You mentioned how any theory, no matter how smart, needs to be confirmed by experiment before it is accepted. I think that experiments have an even more important role to play, namely to point theorists in the right direction. It is therefore doubly wrong to think that it is possible to understand quantum mechanics better by simply thinking deeply about it. To believe so would be to repeat the mistake of the ancient Greeks who, perhaps by analogy to Euclidean geometry, believed that natural phenomena could be understood through pure reason. When Galileo dropped two balls from the tower of Pisa (O.K., not really), he did not just falsify Aristotle’s theory but he also pointed the way out for future scientists, most notably Isaac Newton.

    — James Ph. Kotsybar

    “He went to catch a dicky bird,
    And thought he could not fail,
    Because he had a little salt,
    To put upon its tail.”
    — Simple Simon

    The Scientific Method doubts itself
    and runs experiments designed to fail —
    run once, or twice or all the way to twelfth,
    at least so long as funding can prevail.

    It’s just to test assumptions Science acts
    to prove the things we’ve guessed at all along,
    but, since we take too many things for facts,
    it shakes our balance, when it shows we’re wrong,
    but wrong and right are undermined by truth –
    the bottom line that Science always seeks,
    oblivious to the accepted couth
    or notions, when curiosity peaks.

    It’s narcissistic, obsessed with its care
    and our best guide to within or out there.

  15. I’ve been long peeved about how to interpret QM experiments. The idea of putting forth the experiments on equal footing as the theory is not only reasonable, it is necessary.

    I’m really not into any of the interpretations put forward so far. I think that noise plays an important part ,and noise is equivalent to making a measurement. Furthermore, noise is an error of some sort, and if it is intrinsic to the thing being measured, it means that some of the assumptions are faulty in a mathematical sense (not necessarily the assumption of intrinsic noise). In fact, it would be better to dispense with the idea of a probability distribution altogether as something apriori in the basic formulation of quantum mechanics. I would replace it with an acknowledgement that measuring produces an error that projects a complex numbered entity onto the reals. The intrinsic error not only affects the accuracy, it changes the thing in itself (the complex wave function) to the kind of thing that is perceived or measured (a real number expressing a property such as spin). That way the idea of making a quantum measurement is the same as making a classical one, but the interpretation is that the nature of the measured is intrinsically different from the nature of the corresponding unmeasured entity.

    I made the following mathematical error the other day. In trying to prove a proposition false, I labeled one term of the equation even, and expressed the other terms (all of them even) as functions of it. It turned out after factoring that I could take out two once more from only the term I labeled even. I was struck by the fortuitous coincidence and so I went back and relabeled the terms of the equation, this time calling another term even and expressing the other terms as functions of it. Again I found that the labeled term could be factored evenly twice. That meant I had proven that the proposition was false.

    Unfortunately, my proof was in error, but the idea stuck with me. What if there were good proofs by which someone proves solely by arbitrarily labeling a term among many even that it must also divide by 4? If it were not obvious that the equation is false for the integers (but true for the complex numbers), then the assignment of the properties of evenness and oddness become arbitrary and mysterious until you realize that the operation only produces false properties. By analogy, making measurements induces properties that are real valued, not complex, and these properties are not those of the wave function of the thing in itself.

    This sounds instrumentalist, but my feeling is that it has consequences. It avoids rolling dice in favor of simply measuring or invoking the axiom of choice, and there are not an infinite number of universes chosen among by some decision logic, there’s just one universe and the only beings playing dice are us.

Comments are closed.