Statistics

Robert Heinlein And The Distinction Between Scientist vs Academician

Heinlein was no academic.

Christos Argyropoulos ‏(@ChristosArgyrop) asked me to comment on his post of the same title at the blog Statistical Reflections of a Medical Doctor.

Heinlein:

One can judge from experiment, or one can blindly accept authority. To the scientific mind, experimental proof is all important and theory is merely a convenience in description, to be junked when it no longer fits. To the academic mind, authority is everything and facts are junked when they do not fit theory laid down by authority.

From this grist, our friend Argyropoulos bakes the following pita: “[Heinlein] summarises the essence of the differences between Bayesian (scientific mind) and frequentist (academic mind) inference or at least in their application in scientific discourse.”

To this we can only respond, frequentists are people, too (right, JH?).

Heinlein’s shorthand is a false dichotomy: there are more ways to judge than from evidence or blindly accepting authority. One can infer, extrapolate, guess, induce, deduce and so on. And theories can be junked, modified, or they can even be proved true (rare, rare).

But I take his point. Tradition, collegiality, the big C (Consensus), and the big G (grants), ego, prestige, hope for promotion, boredom, politics, politics, politics drive science just as much or more than any passion for uncovering truth.

Argyropoulos:

For objective Bayesians, models are only convenience instruments to summarise and describe possibly multi-dimensional data without having to carry the weight of paper, disks, USB sticks etc containing the raw points. Parameters do the heavy lifting of models and the parametric form of a given model may be derived in a rigorous manner using a number of mathematical procedures (e.g. maximum entropy)…

Now consider the situation of the frequentist mind: even though one can (and most certainly will!) use a hypothesis test (aka reject the null) to falsify the model, the authoritarian can (and most certainly will!) hide behind the frequentist modus operandi and claim that only an unlikely body of data was obtained, not that an inadequate model was utilized.

Yes to this last point; double-yes with flags waving. The vast, immeasurable multitude of models are never tested. Talk about taking things on faith! Entire fields look to software as the ancients used to consult oracles. If the chicken guts are spotted, i.e. the p-value is wee, the theory is true. Almost no one ever checks the model on data never yet seen. Models are taken “as in” and trusted implicitly.

The leading sin of statistics is failing to teach this distinction.

I’ll let Argyropoulos have the last word to get the discussion started. (I’m badly distracted by the news of the day.)

[O]ur systematic failure to respond to the financial crisis or even to advance science in the last 3-4 decades can be traced to the dominating influence of academicians over scientists. Rather than systematically evaluating evidence for or against particular models in specific domains, we seem to only judge models/explanations by the authority/status of their proponents, a situation not unlike the one in the 30s when Heinlein wrote the aforementioned piece.

Categories: Statistics

31 replies »

  1. “Heinlein’s shorthand is a false dichotomy: there are more ways to judge than from evidence or blindly accepting authority. One can infer, extrapolate, guess, induce, deduce and so on.”

    I think Heinlein would have responded: That’s what you do with evidence.

  2. Following a previous discussion I am tempted to point out that entropy never reaches a maximum (thank goodness) and thus the second law of thermodynamics is not falsifiable. It is, however, very useful for determining the maximum efficiency of heat engines. Thus it is falsifiable for all practical purposes? 🙂

  3. The second law of thermodynamics is falsifiable. You only need a counter-example to do so. The idea that the Universe is ultimately random OTOH is impossible to disprove.

  4. Hi DAV, Your statement “The second law of thermodynamics is falsifiable. You only need a counter-example to do so” is a little odd as it is logically equivalent to “X is falsifiable. You only need a counter-example to do so” and thus would seem to be a tautology that says nothing at all about X. You must also realize that the second law is a statistical law and thus fits in very well with Briggs’ articles.

  5. Scotian,

    Not sure how you find that odd. The statement “all swans are white” only needs a black one to falsify it. Thermo2 implies perpetual motion machines of the second kind are impossible (and not highly unlikely which would be a statistical statement). If I could find or build one then 2nd Thermo would be falsified and in need of modification.

    Granted, the impossibility is an “epistemic impossibility” but that’s just a fancy way of saying the current formulation doesn’t allow it.

  6. OTOH:
    Einstein once remarked to Heisenberg, “Theory determines what can be observed,” which Heisenberg later commented that it changed his life.

    Pierre Duhem put it this way:
    “Take two physicists who do not define pressure in the same manner because they do not admit the same theories of mechanics. One for example accepts the ideas of Lagrange; the other adopts the ideas of Laplace and Poisson. Submit to these two physicists a law whose statement brings into play the notion of pressure. They will hear the statement in two different ways. To compare it with reality, they will make different calculations so that one will find this law verified by facts which, for the other, will contradict it.” [Emph. added]
    – “Some Reflections on the Subject of Experimental Physics” (1894)

    Duhem, too, took an instrumentalist view of physics.

    While Heinlein was contrasting science and engineering with all those worthless humanities subjects, he also said that “a fact is self-demonstrating,” which is demonstrably false.

  7. Well DAV the thing is that the second law does only say that perpetual motion machines are very unlikely (very, very unlikely) and not impossible. This was the great discovery of Boltzmann. It is even engraved on his memorial stone as shown here http://artigosdomestre.blogspot.ca/2007/09/boltzmanns-epitaph.html and you see this equation is a statistical equation. The same equation expressed as probabilities (information) is shown here http://en.wikipedia.org/wiki/Entropy_(information_theory)#Characterization. Strangely enough I derived this equation this morning in my statistical mechanics class. Now what are the odds of that?

    Now the failure to build a functional perpetual motion machine means that we have failed to falsify the second law for all practical purposes but not in the Briggsian sense.

  8. Scotian,

    “Now the failure to build a functional perpetual motion machine means that we have failed to falsify the second law for all practical purposes but not in the Briggsian sense.”

    You said finding a perpetual motion machine wouldn’t falsify the 2nd law (perpetual motion machines are very unlikely (very, very unlikely) and not impossible) so exactly what do you mean by failure to build a functional perpetual motion machine means that we have failed to falsify the second law?

  9. DAV,
    I guess that I put too many negatives in the sentence, but it seems clear to me. Remember that this is all in reference to the debate between Briggs and Willis.

  10. Scotian,

    Well, OK, but if a theory amounts to a probability statement it is only useful in the sense of rule of thumb and the only way to “practically falsify” it would be to show that its estimates of the probability are way off base. In fact, I would tend toward saying such a demonstration genuinely falsifies it in every sense and not just “practically” as the only thing it can predict are the probabilities.

  11. DAV,
    I think that there is some confusion here. Since the second law is supreme it will limit the accuracy of all physical theories and we haven’t even talked about Heisenberg. Thus statistical limitations can not be avoided. Laws that appear to be exact are so only in mathematical form. Briggs has written many articles making a similar point although from a more abstract point of view and I am clearly being influenced in a slow insidious manner.

    To see what kind of probabilities that we are sometimes talking about consider an ideal gas in the left side of an equal volumed two chamber container. The partition is removed so that the gas molecules are free to roam throughout the container. What is the probability that at some instant in the future they are all found to have returned to the left side? For 332 molecules this is about one chance in a googol (ten to the power of one hundred). Is this useful enough for you?

    The randomness inherent in nature can be seen clearly in the phenomenon of Brownian motion. The real thing is here:

    http://m.youtube.com/watch?v=2Vdjin734gE&desktop_uri=%2Fwatch%3Fv%3D2Vdjin734gE

    A visual model is here:

    http://m.youtube.com/watch?v=6VdMp46ZIL8&desktop_uri=%2Fwatch%3Fv%3D6VdMp46ZIL8

    The discoverer was the botanist Robert Brown.

  12. “Since the second law is supreme it will … ”

    But since it can’t, so you say, be falsified, then you might as well be talking about the capability of gremlins. And since its not falsifiable what it can or can’t do is indistinguishable from a WAG.

    “For 332 molecules this is about one chance in a googol”

    The actual numbers don’t mean anything in this discussion. The point is that if all you get are probabilities then you all you have is a rule of thumb like saying grass is green when you really mean green should be your first guess. You don’t really have am explanatory theory — just a practical one that seems to work. (And one apparently you have practically failed to falsify).

    “The randomness inherent in nature can be seen clearly in the phenomenon of Brownian motion. ”

    How do you know its inherent in nature and not just an outcome of your limited knowledge?

  13. I should point out that if your theory cannot be falsified then you can never know if and when it’s working . It’s meaningless.

  14. DAV,
    One of us is losing the plot. Tell me, what do you expect from a theory about the physical world?

  15. Scotian,

    I expect it to tell me what will happen given circumstances and not what outcome might (along with the implied or not) happen which is effectively saying nothing. The former I can test for reliability the latter not at all.

  16. DAV,
    You are asking for something that never was and never will be.

    “I expect it to tell me what will happen” then you say “The former I can test for reliability”. If you know what will happen there is no need to test. You only test when there is a probability of failure.

    “which is effectively saying nothing” All or nothing? I don’t think so. When Einstein published an explanation of Brownian motion, a statistical phenomenon, did he explain nothing?

    When you say “what will happen” above, to what precision do you expect an answer? In your answer remember that precision (and accuracy) is limited by stochastic processes. Am I to assume that you expect an infinite degree of precision or else nothing is explained?

  17. Scotian,

    “You are asking for something that never was and never will be. ”

    If you are saying I will never find a theory that predicts what will vs. what might occur then what you are saying is utter nonsense.

    You are too wrapped up in numbers.

    Theory: If I mix table salt with water I will get salty tasting water. No ands, ifs or maybes. I can test this. I can’t test every situation but I can test a lot. However, if at some time I mix salt with water and don’t get salty tasting water then I’ve shown the theory to be wrong (falsified it) and must at least modify it to account for the exception if not completely discarding it. Note that a modified theory (call it T’) is not the original, falsified theory even if given the same name.

    If however the theory is mixing table salt with water might produce or usually produces salty tasting water then I really can never know if it false and it basically says: expect salty water but don’t count on it. Far less valuable and possibly not even remotely useful.

    If the theory amounts to A causes B but not always it is useless in my book.

    “Am I to assume that you expect an infinite degree of precision or else nothing is explained?”

    No, of course not. There’s nothing wrong with a theory that states a numerical result will fall within a given range as long as the range isn’t something impractical like zero to infinity or otherwise implies the answer could be anything.

    “When Einstein published an explanation of Brownian motion, a statistical phenomenon, did he explain nothing? ”

    No. Note he didn’t say (as you did) it was the result of “randomness inherent in nature” which (if you meant it as a fundamental property) is untestable. IIRC he gave expressions (which ARE testable) for the mean squared displacement of particles from apparent randomness when a precision is proffered.

  18. DAV,
    Your salt example supports my position not yours. For example “If I mix table salt with water I will get salty tasting water. No ands, ifs or maybes. I can test this”. Let’s examine this more closely. How much salt is mixed in how much water? How sensitive are your taste buds? You see, your test is not as reliable as you think. In practice you will just mix in enough salt until you taste it just as you salt food to taste, which means that you expect that the test might fail and you would need to add more salt, otherwise why test? The numbers are still there whether you measure them or not. If you failed to add enough salt to activate your taste buds, which might happen to be less sensitive than others, would you be forced to conclude that adding salt to water does not make it salty?

    “If the theory amounts to A causes B but not always it is useless in my book”. I find it very unlikely that you believe this in your day to day life. For example, if you contract an infection and your physician writes out a prescription for a course of antibiotics that you are told has a good chance, let’s say 70%, of curing you, will you refuse to take it because it is useless in your book?

    “There’s nothing wrong with a theory that states a numerical result will fall within a given range”. I am glad to hear this as long as you realize that the range will not have sharp edges, but this admission undermines your previous position.

    “Note he didn’t say (as you did) it was the result of “randomness inherent in nature””. He didn’t use those exact words as far as I know, but he and others (Boltzmann, Bohr etc) said much the same thing. The whole God doesn’t play dice with the universe was another topic for another day. I am curious as to why you think that randomness is untestable? We were talking about falsifiability. Here is Einstein’s paper translated into English.

    http://www.pitt.edu/~jdnorton/lectures/Rotman_Summer_School_2013/Einstein_1905_docs/Einstein_Brownian_English.pdf

    The variance of Brownian motion is testable within the precision limits of random motion, but to repeat myself we are talking about falsifiability and the difference between my interpretation of Briggs’ versus Willis’ positions.

  19. Scotian,

    ” Let’s examine this more closely. How much salt is mixed in how much water? How sensitive are your taste buds? You see, your test is not as reliable”

    You are merely listing reasons why my original salt water theory might be falsified. You are forgetting the “no ands, ifs and maybes”. I very well could end in a position where I couldn’t create a testable theory but at that point I would say I don’t have a good theory and additional research is needed; please send more money.

    “There’s nothing wrong with a theory that states a numerical result will fall within a given range”. I am glad to hear this as long as you realize that the range will not have sharp edges, but this admission undermines your previous position.

    Nonsense. In practice, the range should be specified such that a reliable (whatever I decide that means) sensor would register the expected result and if it doesn’t then the theory is faulty. The salty water example may not have been the best example but, frankly, I’m trying to convey a concept instead of making a concrete, bulletproof example.

    your physician writes out a prescription f[with] say 70%, of curing you, , will you refuse to take it

    Maybe, maybe not but that’s wandering away from the central topic as I rarely consider a doctor’s guess a scientific theory.

    I am curious as to why you think that randomness is untestable?

    Because you can never distinguish between unknown cause vs. a fundamental property of randomness.

    we are talking about falsifiability and the difference between my interpretation of Briggs’ versus Willis’ positions.

    So am I whether accept it or not. Any theory that is forever untestable (therefore unfalsifiable) is pure (and permanent) speculation and its only value is whatever value you place on speculation. You may choose to think it true but you are on shaky ground. If the theory is U(=untestable) + V(=testable and effectively not dependent upon U) and you test V you are only testing the V part and the U part is irrelevant.

    As far as “practically falsifiable” goes: you are choosing to disbelieve without hard evidence. But if the theory is, in fact, unfalsifiable I doubt it matters either way.

  20. DAV,
    The more you explain your position the less that I understand you. I guess the more caveats (i.e. whatever I decide that means) that you add the more unfalsifiable your position is. Saying “nonsense” isn’t an argument. “doctor’s guess”: You may need a better doctor. I guess that, channeling Briggs, I should ask you what you mean by the word random.

    “untestable (therefore unfalsifiable)” Untestable may mean unfalsifiable but the reverse isn’t true. The words can’t be used interchangeably.

    “choosing to disbelieve without hard evidence” What does this mean and what does it have to do with the limits to falsifiability?

  21. Saying “nonsense” isn’t an argument.

    Wasn’t an argument.

    “doctor’s guess”: You may need a better doctor.

    Maybe so but a doctor’s opinion is a guess. That’s why a second opinion can be a good idea.

    I should ask you what you mean by the word random.

    Same as Briggs (I think): unknown; unknowable

    “untestable (therefore unfalsifiable)” Untestable may mean unfalsifiable but the reverse isn’t true. The words can’t be used interchangeably.

    Possibly but as far as I’m concerned unfalsifiable means untestable now and in the future. So, even if something is untestable now because of current limitations but the limitations could conceivably be removed at some future time then I consider it testable.

    “choosing to disbelieve without hard evidence” What does this mean and what does it have to do with the limits to falsifiability?

    It means there is no such thing as “practically falsifiable”. If you consider a theory falsified even though you are getting answers allowed for by the theory then you have no hard (conclusive) evidence. Trying to claim falsification with such is expressing an opinion. You may as well put it to a vote.

  22. Thanks Matt for commenting!
    I Heinlein’s paragraph refers to the things we condition on to reason: either data or other propositions whose truth is assumed. Such propositions almost always come from an external authority (Jayne’s weather man example comes to mind at this point!).

    In any case, there is more to Heinlein’s statement and the Bayesian vs frequentist approaches than meets the eye. In particular, there are cases of possible scientific fraud where both the accusations and the responses are driven by the convolution between the two interpretations of probability and the dichotomy of academicians vs scientists (as defined by Heinlein)

  23. DAV,
    Quote “Maybe so but a doctor’s opinion is a guess. That’s why a second opinion can be a good idea”. A doctor’s opinion is a guess? I sure hope not because if it is I have been extraordinarily lucky in my past dealings and how long can my luck hold out? What has made you so bitter? Of course the medical profession can succumb to political pressure and fads come and go, so it pays to be an informed consumer. I am curious, how many second opinions have you obtained?

    “Same as Briggs (I think): unknown; unknowable”. It may very well have these characteristics but this must be an insufficient definition. For example, what I ate for supper on this date one year ago is unknown and unknowable but it clearly wasn’t random.

    “If you consider a theory falsified even though you are getting answers allowed for by the theory then you have no hard (conclusive) evidence”. Where does this come from? I don’t remember anyone saying this. Could you supply a direct quote that says this.

  24. what I ate for supper on this date one year ago is unknown and unknowable but it clearly wasn’t random.

    Makes my point. “Random” obviously refers to reasons unknown to the observer. It’s not a property of an event itself . OTOH, it’s impossible to show that it isn’t when the causes are hidden. So, I should add “hidden” to the list.

    “If you consider a theory falsified even though you are getting answers allowed for by the theory then you have no hard (conclusive) evidence”. Where does this come from? I don’t remember anyone saying this. Could you supply a direct quote that says this.

    Earlier, you said:… the second law of thermodynamics is not falsifiable. It is, however, very useful for determining the maximum efficiency of heat engines. Thus it is falsifiable for all practical purposes?

    You did frame it as a question so I may be misunderstanding. However, I fail to see how you could be claiming it falsifiable (practical or otherwise) when the theory, if non-falsifiable, must be allowing for (effectively) any answer — as it must for if it didn’t then it would be falsifiable in fact.

    So, I ask: is the limit of heat engine efficiency true if and only if Thermo2 is true? If so, I submit the second law of thermodynamics is not falsifiable is a false statement. If not then the claim of “practical falsifiable” must mean you would have no hard evidence to refute it.

    A doctor’s opinion is a guess?

    Yes. An educated one perhaps but a guess nonetheless. I suppose you don’t think misdiagnosis could occur.

    I sure hope not because if it is I have been extraordinarily lucky

    Indeed.

  25. DAV,
    If you readily confuse the words “falsified” and “falsifiable” I fear that any further discussion is pointless.

  26. Heinlein was just voted No. 2 overwhelmingly above the trailing pack in a recent Internet poll held by the Missouri State House for inclusion into the Hall of Famous Missourans.

    [More than one choice may be selected per year.]

    What else would one expect from a Favorite Son of the “Show-Me” State?

    JJB

Leave a Reply

Your email address will not be published. Required fields are marked *