Workplace Freedom: A Primer for Alan Dershowitz

I just know someone is going to bring up the damn Firefox thing….

As does the author of that piece, quoting Dershowitz’s opinion,when he should be used as an avatar of political bias.

“On several occasions where the research has suggested opposite conclusions, his students say, he has asked them to go back and look for other cases, or simply to omit the discrepant information.”

Assuming that this is true, that is clear-cut misconduct in any academic discipline.

“I would bet anything that 99 percent of the people who are demanding that [Salaita] be restored tenure would be on the exact opposite side of this if he’d been making pro-Israel but equally uncivil statements,” he said.

He’s not wrong here.

There’s an obvious difference between not “dedicating equal time” and being “on the exact opposite side.”

Like AF, I’m not convinced. What we think about the guy’s thesis-driven research approach is liable to get entangled with what we think about his let-the-grad-students-do-the-reading methodology. There’s something boneheadedly arrogant about responding to a rebuff from the data by asking someone else “to go back and look for other cases, or simply to omit the discrepant information”, but if you were writing a paper on an under-documented phenomenon which you were pretty damn sure was there – say, worker discontent in un-unionised workplaces in the Southern states in the 1960s – would you feel honour bound to reference all the accounts you found that suggested that, nope, there was nothing to see there, everything was fine? Or would you go back and look for other cases, or omit the discrepant information?

“‘On several occasions where the research has suggested opposite conclusions, his students say, he has asked them to go back and look for other cases, or simply to omit the discrepant information.’

Assuming that this is true, that is clear-cut misconduct in any academic discipline.”

No, it’s not. It depends on what the proposition was that the research assistant believed to be contradicted by the evidence, what evidence led the research assistant to believe the proposition to be untrue, and what statement ended up in the final work. There are a number of ways the RA’s statement could be true without it amounting to academic misconduct.

“But I can imagine you imagining it. So your point is refuted.”

The original Dershowitz argument was in the form of “a bet,” so we’re talking about speculative counterfactuals. It’s not sufficient that I can mentally conjure the image Corey Robin being such a stalwart defender for academic freedom that he would flood his blogs defending the unhiring of a pro-Zionist academic. Colloquially “I can’t imagine” implies my certainty based on my experience that he wouldn’t — ie: how I might bet. It is self-evident imho (though obv not yours) that Salaita’s political defenders are using the argument of “academic freedom” to buttress their defense of the particulars of Salaita’s POV. This wasn’t like Burton Joseph of the ACLU defending the National Socialist Party of America in the Skokie Affair.

Instead, as a lawyer might, he writes his conclusions, leaving spaces where he’d like sources or case law to back up a thesis.

This is something we see quite a bit of in contra-scientific efforts to undermine the climatological community’s research program on global climate change. Is there a formal name for logical constructs of this sort? Affirming the consequent doesn’t seem to quite fit the bill.

“If a scholar deliberately discards relevant data that is inconvenient to his or her preferred conclusions, then he or she isn’t a scholar, but a hack masquerading as one.”

Not a scholar in a normative sense, I agree. But not guilty of academic misconduct either, I would still argue. If being a hack masquerading as a scholar is academic misconduct, there are thousands of professors who could be plausibly characterized as guilty. I don’t think this is or should be the standard. Can anyone cite an example of someone in the humanities or legal academia who was disciplined for academic misconduct simply for not dealing without relevant counterevidence?

“It’s in the same class as this case in which a foreman at a West Virginia mine was pressured to make contributions to GOP candidates (through a centralized process, in which her boss would be able to see who contributed and who did not), and alleges that she was fired when she failed to comply.”

On the other hand, union members are compelled to contribute to Democratic party candidates as a matter of course, whether they agree with them or not.

(I did communications for SEIU Local 6434 for four years. )

Dershowitz uses student researchers both for academic articles and briefs used in litigation. In either case, failing to cite contrary legal holdings is misconduct, academic and professional respectively. Litigation is different from a scholarly work in being goal directed, advancing the client’s case, so there the “backwards” research model is reasonable. But by failing to cite contrary precedent the lawyer runs the risk of being thought at best sloppy even if not dishonest. Not a good strategy for your professional reputation.

I have wondered about whether Dershowitz would have such a high-profile reputation if he had to do his own work instead of relying on a small army of very bright, highly motivated student workers. Sort of like some Supreme Court justices and their clerks.

Interesting, if brief, relevant piece from the Chancellor of my university, an economist, about who has what in common with whom:

http://www.chancellor.wisc.edu/blog/is-an-english-professor-a-scientist/

You pay your dues or you lose your job, all the while knowing that the bulk of your dues are going to end up in Democratic party coffers.

Wow. Which union pays the bulk of its dues into political races?

“AF – I think that there’s a significant difference between “not dealing with” and “deliberately suppressing.” It’s obviously often hard to prove the latter – but I think that e.g. if a historian were found to have cherry picked a couple of convenient quotes from an archival source that he or she had clearly consulted, while failing to report the substantial body of evidence that weighed against his or her thesis, he or she would be in serious disciplinary trouble.”

Sure, if a historian did that, it would be a problem. But Dershowitz isn’t a historian. And he doesn’t deal with archival sources — which are the real source of the misconduct in your hypothetical. If a professor of literature or political theory or law — or for that matter a historian — cites portions of a generally available text in support of an argument and ignores (even “deliberately”) portions of the text that go against the argument, that is bad scholarship, not academic misconduct.

AFAIK, there are requirements in the law for non-Union members to pay at least that part of of the union dues for the collective bargaining services and other benefits union members receive, but that they can’t charge for political activity in what the non-union worker pays in said dues. Unfortunately, SEIU has had some problems that I’m aware of, but none of them linked to the contributions to candidates.

And as someone who was the only person in my workplace as the unofficial shop steward, things look a lot different on the ground then when you’re even slightly above the fray.

I’ve been talking with physical science colleagues here at UIUC The numbers aren’t clear yet but a significant minority of us are very concerned about this as an academic freedom issue. A few of those who signed the pro-administration letter are even a bit concerned about the implications for academic freedom, although they’re more concerned about institutional stability and funding. One economist (now elsewhere) – very conservative, and somewhat Islamo-phobic- is strongly on the academic freedom side. I just spoke last night with a philosopher who is boycotting. He hates identity politics, has a low opinion of the humanities departments here, but is sure that this blow to academic freedom will just make them worse.
So I’d hesitate to throw around crude labels of the sides.

Brett Bellmore 09.22.14 at 4:54 pm

“Sure he is: If Salaita had been making similar public statements in the opposing direction, they’d never have the opportunity to take the opposite side on his unhiring. Getting unhired required getting hired in the first place, after all.”

Pretty rich, when the article starts with Alan Derschowitz.

So, let’s see. By Julians logic, having my employer fire me for not making a personal contribution to a political cause that I oppose is exactly the same thing as being a member of a union – where I can vote on the policies – and having some of the money go to causes that I disapprove of. But – let me guess! – it’s completely OK to have corporations spend their money on causes that I disapprove of, even though I’m a shareholder.

AF, if hypothetically a legal scholar asked a student to research a legal issue, explicitly predefined the expected outcome, the student came up with relevant case law that did not support the expected outcome, and the scholar then told them to discard that relevant precedent, then I think most academics including most legal scholars would agree that this constituted misconduct.

Of course we don’t know in detail what happened and can only guess based on those student statements. If it were true and could be proven, it is still unlikely that formal action would be taken. Academic misconduct is often difficult to prove and gets punished rarely, and only in the most egregious cases. But that doesn’t mean that the standards aren’t there and aren’t well understood.

“Which union pays the bulk of its dues into political races?”

Taking into account in-kind contributions of union labor? I don’t know that any do, but their political spending is typically a LOT higher than 1%. More like 10-20%.

Political Spending by Unions Far Exceeds Direct Donations

Since unions will typically only exempt non-members from the reported political expenditures, and their total political expenditures including in-kind contributions amount to maybe 5 times as much, writing that letter only slightly reduces the extent to which you’re forced to subsidize the union’s political agenda.

Peter, that’s what I figured.

Julian F 09.22.14 at 8:20 pm

I surrender :^)
————
You trolled CT and sucked in Brett. You win an Internet Service Star. Wear it proudly.
~

Henry — you misunderstand the “candor towards the tribunal” rule of professional conduct. You have indicated that this rule prohibits “suppress[ion] of citations to relevant cases” or “knowingly omit[ting] relevant information or citations.” These statements are not remotely true.

The model rule, like the rule where I practice, provides that a lawyer may not: “fail to disclose to the tribunal legal authority *in the controlling jurisdiction* known to the lawyer to be *directly adverse* to the position of the client and not disclosed by opposing counsel.”

If I know of directly adverse authority from a different state I can “suppress” that authority all I want. If I am litigating a case under Oregon law, and there is a US Supreme Court decision that reaches a directly contrary conclusion applying federal law, I need not cite it (though I would certainly expect the other side to).

Moreover, if I know of authority that I (or a judge) might consider relevant, but that is not directly adverse to the position of my client, I need not disclose it. For example, a case may exist in a related area of law that would provide a great and persuasive basis for my opponent to ask the court to adopt its reasoning in the present case. I will not cite that case.

Indeed, among the less than punctilious members of the bar (a fair proportion, as all the jokes suggest), this rule is viewed as virtually toothless, because nearly any case can be distinguished in such a way that it is not “directly” adverse. In other words, if you have some argument to work with, then the legal authority is not “directly” against you.

And this is only regarding legal authority. Regarding facts, I cannot make an affirmatively false statement (or an omission that is the equivalent of a misrepresentation), but in no way is a litigant required to volunteer damaging information just because it might be relevant to the judge’s decision. Our system is an adversarial one. It is the other side’s job to highlight the weaknesses of your case.

I would hope that I don’t really have to remind Henry or commenters of Feyerabend’s “Against Method” and related studies on the history of science.

If discarding/hiding/forgetting data that didn’t fit the hypothesis wasn’t a critical part of science, we’d have to do without a number of subsequently very well substantiated theories. Millikan’s estimation of e (electron charge; the “oil drop” experiment) is perhaps one of the most well known, Even though there is some disagreement over the extent and nature of Millikan’s exclusion of results, there is no disagreement that he did collect a substantive number of results that were not included in published work and that at best cast the nature of his results in less focus than the published set did.

And then there is longer term effect on other researchers, neatly summarized by this quote from Feynman on the wikipedia page:

We have learned a lot from experience about how to handle some of the ways we fool ourselves. One example: Millikan measured the charge on an electron by an experiment with falling oil drops, and got an answer which we now know not to be quite right. It’s a little bit off because he had the incorrect value for the viscosity of air. It’s interesting to look at the history of measurements of the charge of an electron, after Millikan. If you plot them as a function of time, you find that one is a little bit bigger than Millikan’s, and the next one’s a little bit bigger than that, and the next one’s a little bit bigger than that, until finally they settle down to a number which is higher.

Why didn’t they discover the new number was higher right away? It’s a thing that scientists are ashamed of—this history—because it’s apparent that people did things like this: When they got a number that was too high above Millikan’s, they thought something must be wrong—and they would look for and find a reason why something might be wrong. When they got a number close to Millikan’s value they didn’t look so hard. And so they eliminated the numbers that were too far off, and did other things like that…

So, as awful as Dershowitz may or may not be, lets not be quite so sanctimonious about “misconduct”, eh? Just as forgetting is critical part of self-formation, convenient exclusion of contradictory data is a key part of the evolution of scientific hypotheses. Of course, there comes a point at which that has to stop for the hypothesis to evolve towards a theory, but that isn’t always (and perhaps isn’t ever) the responsibility of the initial author.

AF has a point actually — I don’t think it’s terribly uncommon in the humanities (to speak of what I sort of know) to start with an idea that functions as a sort of thesis, and then to do the best job one can to marshall the strongest argument(s) for it. Rather than going wherever the text/evidence/whatever leads you. It’s also quite common — indeed, sanity demands it — that one doesn’t look at _all_ the countervailing evidence/arguments.

What AF seems to be missing though is that simply “to omit the discrepant information” sounds really bad, at least borderline misconduct (though admittedly it might refer to something less bad than it sounds).

AcademicLurker @14: I have a different theory about the tepid participation of natural scientists on either side of the affair. Scientists have their own distinct academic ecosystem and their own distinct news sources/social media/blogosphere.

It may also be relevant that a cloropleth map of the physical and biomedical science academic ecosystem would show Israel large enough to cover both Urbana IL and Blacksburg VA at once. Some sleuthing (and a bit of coding) would be needed to filter on scholars who have no first-degree collegial ties (coauthor, advisor, coeditor) with any Israeli citizen, but that might be all it takes to make the alleged science vs. humanities bias, if such there be, vanish into the noise.

“if in that instance, I not only refused to support that person’s rights by, say, not signing a petition (and if anyone can find an instance of me doing so, you should let me know), but actually spent a good part of my time over a period of nearly two months, working to deny them those rights.”

Isn’t that essentially what Cary Nelson did? Defender of academic freedom, and then working to deny someone those rights?

I don’t speak for the people who are trying to smear you in particular. But there does seem to be something of a belief here that the left cares about basic freedoms non-hypocritically and the right doesn’t. In the Age of Obama, I haven’t observed this to be true. The archetypical figure for these years is Marty Lederman, strong opponent of secret detention and expansion of executive power when Bush was President, and author of secret law once Obama became President allowing the President to have people assassinated at will. The difference between Lederman and Yoo vanished when Lederman went into Yoo’s job.

So I’m willing to congratulate anyone who’s on the right side now, but to presume that people will on average stay on the right side when their preexisting commitments say otherwise would be foolish.

#60 JWB

If science were what Descartes and Popper told you in high school, you’d expect to see normally distributed scatter, with decreasing variance as the measurements improve.

Yes.

If science were what Kuhn and Feyerabend told you in sophomore year of college, you’d expect to see paradigms shifting stochastically without convergence.

We do see that, just not very often. Very few scientists have both the imagination and the clout to present a new paradigm and get it accepted.

Marvelously, you instead see the human endeavor of science, in which measurements are “sticky” around the “known” value — not out of malice or even sloth, but simply for the reason your keys are in the last place you look, namely that when you think you’ve found them you stop looking — but still narrow over time toward objective truth.

I’m with you here up to the point you decide that what they narrow in on is objective truth. Why would you think that?

The more prior research has already been done, the stickier the values get. Right up until somebody makes a successful paradigm shift after all.

Fifty years ago, the best value for the neutron lifetime was just over 1000 seconds; a value as low as 880 seconds was a 3.5 σ surprise, a chance in five thousand. Today, it’s 880 seconds with less than a second of systematic error; to roll it back above 1000 seconds would be a 130 σ jump, one chance in 10 to the 3700th power. Buddy, can you paradigm?

Let’s say for the sake of argument that the current value is correct. The original value was way off, but they didn’t think it was off. They got consistent results to the point that if they had gotten a correct result it would have been an unusual outlier.

And I will assume that this fit the usual pattern where the observed values gradually drifted down toward the correct one. And now more reliable equipment gets the same result more consistently, so that what used to be 3.5 sigma has become 130 sigma.

Now, imagine that the results were drifting in the correct direction, but they had not drifted to the right answer yet. But the equipment got so reliable that it gave much more reproducible results. Wouldn’t that tremendously slow the rate of drift?

Imagine that the correct value is in fact 130 sigma on the other side, and it has drifted about halfway. When researchers compute the value fresh, how likely is it they will publish a value that’s more than, say, one sigma away from the consensus? It will take a very long time to get there at this rate!

Well, but we could argue that the current value can’t be 130 sigma off but must be correct. And yet, how was it that the early values were consistently around 3.5 sigmas off? 130 sigma is just more of the same. It takes a lot more calibration to adjust your numbers that far if you in fact start to measure the correct value at the beginning, but 130 sigma is also a lot more powerful as an incentive. How could you possibly be right when everybody else agrees on something so reliable and so far from your answer?

Of course I don’t want to say it has to be that way. But consider the possibility that we might not converge to objective truth. Possibly it’s a race between drift in the correct direction versus increased precision. When the precision gets high enough, maybe the drift mostly stops whether the result is true or not.

Imagine that the correct value is in fact 130 sigma on the other side

No, I won’t.

“(And in fact even the ACLU is not the ACLU; it also has to choose its battles.) ”

Not really, so much. The crucial difference between individuals and organizations is that individuals have largely fixed limits, while organizations are capable of growing to meet the challenges they take on. An organization like the ACLU takes on a cause, it accretes new members who think that cause important, and grows to have the capacity to take it on.

That’s why, unlike an individual, it really is fair to look at an organization, and draw conclusions about what it’s leadership believes from the fights they take, and the fights they pass on. The ACLU, for instance, didn’t decline to defend 2nd amendment rights, and invent specious excuses for why they weren’t really civil liberties, due to a lack of resources. (The giant the NRA grew to, taking on that cause, ought to tell you that.)

They passed on that fight, and even enlisted in the other side, because they hated that right. And really didn’t want to associate with the folks who didn’t hate it.

But you’re certainly right about the limits of the individual, and I draw no conclusions from whose causes YOU fail to exert yourself on behalf of. Individuals can’t do everything, so they really do have to pick their fights, and only the fights they DO pick say much about them.

>>>“That’s the way it’s done; a piecemeal, ass-backwards way,” says one student who has firsthand experience with the writing habits of Dershowitz and other tenured colleagues. “They write first, make assertions, and farm out [the work] to research assistants to vet it. They do very little of the research themselves.”<<<

No surprise. This is how judges make law.

“while organizations are capable of growing to meet the challenges they take on.”

This is pretty weak. If charitable organizations could grow infinitely we wouldn’t have poverty now would we? There is a limit to the number of dollars out there for orgs like the ACLU so they have to decide how they want to spend those dollars.

It could have been that science (in this case turn-of-the-millennium particle physics) had that behavior, but in fact it doesn’t. Instead, for fundamental constant after fundamental constant, people do the same experiment a bit better, subject to subtle confirmation biases, and the data points shuffle along guiltily in what will turn out to be the right direction, with suspiciously less variation than the error bars demand.

This is the biased behavior we described, the behavior you said it could have had but doesn’t.

Then — because this is physics and we have many orders of magnitude in sight and we conquer them boldly — somebody does a radically better experiment.

Now you have something! They do a different experiment, and they don’t feel any obligation to respect the old results!

Oops, my keyboard did something weird and posted early.

They do a different experiment and feel no obligation to respect the old results because their new results are better! So the answer takes a random jump, and now this is the value that others compare against. Does it include calibration errors and judgement errors? I dunno. It’s something new, so it might. On the other hand the new precise experiment may be simpler and easier to perform and so there simply may not be as many errors to make. It stands to reason that when your experiment gets an answer that’s precise to 6 decimals instead of 3, there will be fewer fiddly little things to adjust for that could be adjusted wrong. And when it’s precise to 12 digits then you just *know* that all the systematic errors are gone from this brand-new experimental technique.

I tend to think that careful experimental scientists, thinking carefully, will tend to actually get correct results, and those results will consistently get better with time instead of following a series of random jumps with small incremental improvements between them.

But the data you point to suggests that I’m likely to be wrong about that. For 3/4 of the constants on that page, the currently accepted value is outside the error bars from 1990. For how many of them, will some new technique justify a new shift outside the current error bars within the next 25 years?

I was a bit heartened by the K-short lifetime. The graph dipped down, and then it came back up. As if they overshot significantly and then the data started coming in on the other side so they gradually increased again. That makes it plausible that likely the true value is somewhere between the highest and lowest reported.

Joshua Burton, let me just tell you how much I’m enjoying your posts in this thread. Thanks for holding the line; that poem from Hein is just gold.

We’ll be outside the current error bars, because we’re not Cartesian angels of the scientific method; we’re human and can fool ourselves. But, we’ll be outside the current error bars by an amount that is tiny compared to the 1970 error bars, because the scientific religion is better than its priesthood.

Again, if scientists accidentally (but mostly consistentl) fudge their results to get them close enough to the previously-established results, then as the precision of their work increases, that makes the allowable variation decrease — independent of any objective accuracy of the precise results!

I figure though that when everybody already knows the answer is zero, that’s a special case. It’s a great big deal to get a nonzero answer, it’s even a big deal to get errorbars that don’t include zero. So all there is to do is to get errorbars that get closer and closer to zero — unless you actually do measure something astounding.

“This is pretty weak. If charitable organizations could grow infinitely we wouldn’t have poverty now would we? There is a limit to the number of dollars out there for orgs like the ACLU so they have to decide how they want to spend those dollars.”

Who said anything about “infinitely”? There aren’t infinite dollars out there, but there aren’t infinite battles to spend them on, either.

In practice, the ACLU has a half million or so members, the NRA 4.5 million. Somewhere in between is how big the ACLU would have been if it hadn’t decided that particular civil liberty was too offensive to defend.

Again, if scientists accidentally (but mostly consistentl) fudge their results to get them close enough to the previously-established results, then as the precision of their work increases, that makes the allowable variation decrease — independent of any objective accuracy of the precise results!

No. It’s really easy to fool yourself into a couple of standard deviations of error. It is gross, career-limiting incompetence to go to press with hundreds of standard deviations of error. Major advances in precision therefore imply (not logically, but empirically with overwhelming force in publishable work) comparable advances in accuracy. When an error-bar I becomes a dot, you are looking at the true answer in the back of the book . . . and a new book opens with new puzzles when you expand the vertical scale so that the new data points look like I’s again. Particle physicists have a saying: yesterday’s discovery is today’s background is tomorrow’s detector calibration.

Null experiments are no different in this regard, by the way. In fact, the main trick to get order-of-magnitude gains in precision is to find a difference between signal and expected signal that you can turn into a null experiment. At least seven, and possibly all, of the PDG graphs I cited involve measurement in a setup that nulls out a low-order (that is, yesterday’s) result, and observes the residuum.

Thanks to Abbe Faria (comment 22), I now know about the Suzanne Sisley case. Not sure why it didn’t pop up through my media filters before, though.

It brings up an interesting question, though about the different reactions of “hard” and “soft” fields to political pressure: It seems like there has always been more regulation of science research (at least the last sixty years or so) in the form of “you can’t research this or that.” But because the government structures it as ‘funder’s preference’ rather than an outright ban, it seems to get around some of the academic freedom arguments.

This worked for me: Σ, σ, ∫, η

J Thomas: I don’t know why you’re persisting in this, but you clearly don’t have a very good idea of how science actually works, either at the implementational or at the institutional level. The claim that physicists have gotten it pretty much right over the last hundred years or so, in incremental fashion, isn’t really in dispute for any reasonable definition of “dispute.” JWB has nothing to “prove” to you that the historical record hasn’t proven many times over.

I am always grateful when opposing counsel fails to cite relevant authority or mischaracterizes the cases they do cite, because it gives me an opportunity to tear them apart. This is why, ethics aside, you should never do this. If there is authority that goes against your position, you should deal with it up front. You are much better off distinguishing it, down-playing it, or explaining why it’s simply wrong, in your opening brief than having to explain why you omitted it in your reply, or at oral argument.

I know you believe this, but you have not presented the ghost of a reason why anybody should believe it.

I think this side-conversation has essentially come down to where we ended up on one of the Zionist threads: cheap, obdurate and hence ultimately irrefutable (but also boring) skepticism on your part about the lived experience and measured judgment of people who know things, oddly misapplied as a strategy to chat them up.

How would you go about testing that?

As an exercise in context-free internet wanking, this is half-clever. As applied to the actual datasets we are discussing, it’s obtuse: when we say that a formerly cutting-edge measurement (say, the W mass, known at the 20% level in the 1980s and the 0.02% level today) is settled at the 1% or 50 σ level, we mean that the value is so certain it is now baked into the deep engineering design of the LHC detectors as well as into the software cuts. You might as well ask how it would affect the Higgs search if Michael Faraday turned out to be totally wrong about electromagnetic induction. And, of course, for a Kuhnian something like that might happen, at any time. The wonder is that you dare to sit so close to a computer while talking about it.

I think that the point of AM was that actual scientists do not follow particular practices in order to arrive at results, nor to judge the quality of the results. Research is ad hoc, judgement is ad hoc within the field; the only judgement that isn’t is the broader one rendered by human societies: does this help us to improve some aspect of our interaction with the world (by efficiency, knowledge, or both).

I raise Feyerabend again because I think the idea that you could watch what scientists do for a century and conclude that there is no shadow of a doubt that they are on the right track would have struck him as mostly laughable. Whether science is on an ascending staircase towards some better understanding of the world, or whether it took a wrong turn somewhere before it reached a nearly stairway – surely this is something that only hindsight reveals – this has been true for 5000 years of human history, and I see no particular reason to believe that it is any different today.

There’s no questioning the apparent continued string of consistent and improving results (in the sense of the depth and breadth of the world that can be explained) within the natural sciences during the 20th century. But is this sense of things unique to our time? It seems unlikely to me, and either way, I would not want to sit in the same chair of judgement used by Michelson to declare the end of physics in the late 19th/early 20th century.

#87 Jerry Vinokurov

J Thomas: I don’t know why you’re persisting in this, but you clearly don’t have a very good idea of how science actually works, either at the implementational or at the institutional level. The claim that physicists have gotten it pretty much right over the last hundred years or so, in incremental fashion, isn’t really in dispute for any reasonable definition of “dispute.”

It’s clear that physics has gotten lots of results over the last hundred years or so.

The immediate question though is the value of these particular constants.

Joshua and I agree that physicists very often find their results biased to agree with wrong prior research, for whatever reason. Looking at his graphs, it appears to happen far more often than not. Do you disagree?

Next, the error bars are a description of the reproducibility of their work, how well they can do the same thing again and get the same result. Uncontrolled variables cause the results to vary. Controlled variables do not.

So for example, imagine an experiment in which a 60-cycle hum makes a difference, where samples are taken in much less than a sixty’th of a second. You get different results depending on what part of the cycle you happened to be on when the sample is taken. Now imagine that you find a way to isolate the 60-cycle background, but there is a constant 5-volt/meter field. Will that make a difference compared to a minus 5-volt field, or zero? If different instants in the cycle made a difference, it might. But if it stays constant, it will probably make the same difference all the time. It won’t make the results less reliable, just different from what they would be at zero volts.

There are lots and lots of potential errors like that, which can change the result but which only reduce precision when they vary. I don’t know how to measure a neutron’s half-life, but I’ll use that as an example anyway — if it doesn’t work then something similar would work.

I expect you have a way to tell when a neutron reaches its end of life. But it’s an indirect event. You measure something that ought to be produced. And you don’t detect every event. So you need a way to calibrate that measurement, and lots of things can affect it. There can also be some false-positives, and you can probably do a good job of reducing those. You can measure how many of those you get when there are no neutrons present, and figure out how to keep them from happening. Except for the false-positives that only happen when neutrons are present….

Then there’s a question how many neutrons you have to start with. You can produce them at some average frequency, and you can measure how many there are by some indirect method. You have to calibrate that method also. After you get your raw data you must massage it to deal with known errors and biases. So you need a good way to estimate those errors and biases, because the more accurately you can do that the better your final number will be.

In all of this, doing the same experiment repeatedly and getting similar answers tells you only that the various variables are controlled. It does not say that your estimates of them are correct.

Rebuild the same apparatus at another lab and lots of things will be different. You will have to calibrate it all over again. You figure you’ve done it right when you test the calibration and get the same result the other guys did….

Scientists have more faith in their results when they try entirely different methods and come up with similar results. There’s lots of room for error, but if three entirely different approaches give the result, it’s probably more-or-less real.

Joshua pointed out the method of nulling out results. Sometimes you can detect very small results if you are careful. Say you want the weight of a battleship captain and you can’t get it directly. Maybe you can weigh the battleship with the captian on it and off it, and compare the difference. Sometimes this approach can work wonders. Sometimes it can have unexpected errors. If you can’t control the number of seagulls that land on the ship, then you get observable errors. If the captain always takes his duffel bag off the ship when he leaves, you may get a consistent wrong number.

J Thomas: I don’t know why you’re persisting in this, but you clearly don’t have a very good idea of how science actually works, either at the implementational or at the institutional level.

I have a clear idea how it actually works. Far too many people misunderstand how it ought to work. This is basic stuff people really ought to pick up in grade school and I run into PhDs who have completed their dissertations who don’t get it. Not to mention MDs.

#95 JWB

“How would you go about testing that?”

As an exercise in context-free internet wanking, this is half-clever. As applied to the actual datasets we are discussing, it’s obtuse: when we say that a formerly cutting-edge measurement (say, the W mass, known at the 20% level in the 1980s and the 0.02% level today) is settled at the 1% or 50 σ level, we mean that the value is so certain it is now baked into the deep engineering design of the LHC detectors as well as into the software cuts.

That is not responsive. I ask how you’d test it, and you tell me that people deeply believe it, to the point that they would be willing to accept extremely expensive consequences if it turned out to be wrong.

But they would not test it to find out even if they knew a way, so that’s no problem.

It’s times like this that I begin to wonder if possibly Brett Bellmore might have a point about climate science after all….

It’s clear that physics has gotten lots of results over the last hundred years or so.

The immediate question though is the value of these particular constants.

There is no actual question about the value. There is only a misunderstanding on your part of the experiments that have obtained these values.

Joshua and I agree that physicists very often find their results biased to agree with wrong prior research, for whatever reason. Looking at his graphs, it appears to happen far more often than not. Do you disagree?

Those graphs don’t mean what you think they mean. Notice how most of them have the same characteristic shape: They start out with large error bars that tamp exponentially as a function of time, reducing the region of space that represents an “admissible” result. That makes sense when you think about such things as the improvement of understanding in the construction of experiments, more factors being taken into account, and the improvement of the technology itself. Yes, a scientist might have institutional reasons for preferring a result that is a half-sigma closer to the previous accepted value (assuming the scientist has several results to choose from). A value that is off by 10-sigma either means that we’ve had it all wrong before (possible, but unlikely) or your experiment has done something wrong. If you’re a good Bayesian, you should, in the parlance of our times, check yourself before you break yourself.

So for example, imagine an experiment in which a 60-cycle hum makes a difference, where samples are taken in much less than a sixty’th of a second. You get different results depending on what part of the cycle you happened to be on when the sample is taken. Now imagine that you find a way to isolate the 60-cycle background, but there is a constant 5-volt/meter field. Will that make a difference compared to a minus 5-volt field, or zero? If different instants in the cycle made a difference, it might. But if it stays constant, it will probably make the same difference all the time. It won’t make the results less reliable, just different from what they would be at zero volts.

What is this thought experiment intended to accomplish? First of all, we have ways of isolating experiments from things like electric fields; that’s called a Faraday cage and has existed for 150 years. Second, we (presumably) have some sort of fundamental theory that guides our understanding of whether or not these factors are relevant. The magnetic field of the earth is relevant for stratigraphic analysis because you can see the direction of the field reflected in magnetic domains frozen into the rock. We know this because we have a fundamental theory of magnetism. Whether or not some other factor is relevant would, again, depend on the theory. Maybe there are factors that are relevant and we don’t know them, which would be a problem, but under certain reasonable assumptions of the distribution of errors, those factors can be subsumed into “noise.” There are methods for dealing with such things; this isn’t exactly a surprise to anyone.

I expect you have a way to tell when a neutron reaches its end of life. But it’s an indirect event.

Everything is an indirect event; direct events just don’t exist.

You measure something that ought to be produced. And you don’t detect every event. So you need a way to calibrate that measurement, and lots of things can affect it. There can also be some false-positives, and you can probably do a good job of reducing those. You can measure how many of those you get when there are no neutrons present, and figure out how to keep them from happening. Except for the false-positives that only happen when neutrons are present….

Then there’s a question how many neutrons you have to start with. You can produce them at some average frequency, and you can measure how many there are by some indirect method. You have to calibrate that method also. After you get your raw data you must massage it to deal with known errors and biases. So you need a good way to estimate those errors and biases, because the more accurately you can do that the better your final number will be.

In all of this, doing the same experiment repeatedly and getting similar answers tells you only that the various variables are controlled. It does not say that your estimates of them are correct.

There’s a pretty simple experiment that you can do in any advanced undergrad lab (and which I’ve done twice) and that’s to measure the lifetime of the muon (and simultaneously test the special theory of relativity). It turns out that with some PMTs and a vat of scintillator liquid (plus some fancy electronics), you can get pretty damn close to the true value (under the assumption that muon arrival times are Poisson-distributed). Of course this requires you to assume there aren’t any other factors that affect the muon lifetime and you can continue adding complications as you like, but it’s not like there’s no answer for that. Actual laboratories that measure these things think deep and hard about various confounding signals and expend a great deal of time and money in modeling and understanding their sources of error (both systematic and stochastic). At some point, you will exhaust those sources of error down to the randomness inherent in the idea of a measurement. That’s just how it goes; this is actually one of those common methods that Feyerabend seems to think don’t exist. If this is an unsatisfactory situation for you, then I suppose you will remain unsatisfied, but it’s emphatically not a problem for the discipline as a whole.

Rebuild the same apparatus at another lab and lots of things will be different. You will have to calibrate it all over again. You figure you’ve done it right when you test the calibration and get the same result the other guys did….

That’s not how you “know you’ve got it right.” If I do my muon lifetime experiment and get a value 10-sigma outside the mainstream, I will assume that I have made a mistake because lots and lots of other people who are much smarter and better experimentalists than me have done it many, many times and it makes sense to seek the fault in myself rather than in them. And of course, because these measurements are part of a whole tapestry of results, an anomalous muon lifetime would mean that our physics is actually radically different from what we thought it was.

The crucial difference people are missing between the late 19th century and the early 21st is the sheer volume of what we know. We just didn’t know all that much circa 1890 or whenever it was that Kelvin thought the earth was 100,000 years old. And now we do know a lot more things, and if one of those things turns out to be wrong, then a whole bunch of other results would also be wrong, and then everything would be wrong and no longer work at all. But it does work! When you flip a light switch and your bulb fails to turn on, you don’t immediately assume that all of physics is wrong, you assume your light bulb must be broken. The constraints on what counts as a plausible hypothesis today are much tighter than they used to be.

Joshua pointed out the method of nulling out results. Sometimes you can detect very small results if you are careful. Say you want the weight of a battleship captain and you can’t get it directly. Maybe you can weigh the battleship with the captian on it and off it, and compare the difference. Sometimes this approach can work wonders. Sometimes it can have unexpected errors. If you can’t control the number of seagulls that land on the ship, then you get observable errors. If the captain always takes his duffel bag off the ship when he leaves, you may get a consistent wrong number.

Again, I have no idea what this is supposed to mean or what relevance it has to scientific practice other than pointing out that errors exist and need to be modeled.

@98: I wish I had the time to repeat Feyerabend’s historical examination with one rooted in current lab observation. Maybe someone is doing it. I was once a molecular biology PhD candidate, gave that up for software development, but retain many friends and neighbours involved in many different areas of research. My gut feeling is that if you really went into labs (and offices) and observed what goes on, you’d walk away with the same conclusion that Feyerabend did: scientists use whatever they can find, and their practices follow norms only to the extent that they have to tell a story of their work to their peers. I think you’d still conclude that despite observing many common practices in laboratories across disciplines – the noise of “whatever works” would drown out “well, first you do X and then you do Y”. I think that one of the take home lessons from AM and from other sociology of science studies that have occured since it was published is that scientists tell a just-so story about what they are doing even to themselves and even when that story is at odds with their day to day reality. I wish I could give you a citation on that.

On the uniquess of “now”: I suspect that except for some relatively brief periods in the history of various human civilizations, it has always been the case that we have never “known as much as we know now about how the natural world works”. Even when wrong turns were being taken (phlogiston, geocentrism, the humors etc. etc), this was still true.

I have a very strong faith that when viewed on a long enough arc, human knowledge of the natural world tends towards increasing depth and correctness. But I have no idea whether the arc needs to be 100 years or 10,000 years long. If it is closer to the latter, there’s no reason that there may not be periods in which we make serious mistakes, both in measurement and in conception.

However, I don’t really subscribe to any notion of science drawing us closer to some knowledge of a “reality”, and view it as something to be judged based almost entirely on efficacy. By that measure, the 20th century advances have been astounding and I suspect that further Kuhnian shifts in understanding will do nothing to alter that, much as both the relativistic and quantum revolutions didn’t really change the impact of newtonian mechanics.

Joshua’s belief is that this calibration (and other things) gets adjusted to make the results closer to the norm.

Norm?

#103 The Temporary Name

“Joshua’s belief is that this calibration (and other things) gets adjusted to make the results closer to the norm.”

Norm?

For attempts to measure some fundamental constant, I mean the latest previous attempt, or anyway the one that is considered the best.

The argument is based on an indirect measurement. Usually each successive measurement is farther in the same direction. If the changes came because each new researcher found new errors and fixed them, at random, then I’d expect the changes to be in random directions.

So one hypothesis is that they keep actually getting results that are close to correct, but each time they assume they are wrong since all the other previous experimentalists were every one of them off to the same side. So they look for errors they can correct that will push their result closer to the others.

Another possible hypothesis is that if they get results that are just somewhere in the middle of the pack it won’t be very interesting, but if they get something that’s to one side but close enough to be believable, then it’s potentially important. That assumes a level of cynicism on their part that I don’t believe.

There might be some other explanation we haven’t thought of yet.

@106: nice post.

quibble mode though: it is precisely the tangled web of interdependencies of various constants that might lead one to question whether the model is actually correct. obviously this is a deeply philosophical question (somewhat akin to the pros and cons of Bohm’s hidden variables interpretations of quantum theory). i can see at least three possibilities: (1) a world as apparently rich in phenomena as ours is inevitably well described by a model with this kind of interdependency (2) the interdependency is not part of the world at all, but is a consequence of the mathematics used for the model (3) it is ridiculous to believe that a model with this level of interdependency describes the universe.

we’ve had consistent models of the universe before which have been broken apart by new observations. it doesn’t seem absurd to think that as the number, accuracy and precision of our observations explodes over time that this becomes less and less likely (because as you note, the model involves so many more interdependent components). but it doesn’t seem absurd either to believe that further kuhnian/copernican/quantum shifts are still part of the future of our understanding (i.e. that despite the lovely interdependencies and consistencies between fundamental constants, they are just artifacts at some level).

#106 JV

“Yes! That’s the part that can make it work! To the extent that the people practicing that art do it really well, they can get specially good results.”

Which is what JWB and I have both been trying to say. What are you arguing about, then?

It sounded like you both were saying that you were confident in the results because they are measured to such great precision. (Which you both called “accuracy”, which is something entirely different.)

Getting precision is something different from the art of discovering and correcting systematic errors.

Here’s a climate-change example since I looked at it recently for Brett Bellmore. Say you have a bunch of weather stations around the country that have been measuring daily temperature for decades or sometimes even centuries. If you start replacing them with electronic thermometers that record the temperature to .01 degrees, your new readings are maybe more precise. That isn’t a bad thing, but it doesn’t much matter in a climate-change context.

Say that over the past 40 years a lot of your non-automated stations have switched from measuring the temperature in the afternoon to measuring it in the morning. You can either remove their data, or you can come up with the best correction you can manage to account for the change. Some people say you should just use the raw data, but if you do you will get a general cooling effect that isn’t real, because most places it’s cooler in the morning than in the afternoon.It makes no sense to use the data as if it isn’t biased, but if you make your best effort to correct it, then political partisans will say that you fudged it to get the answer you wanted.

It’s a subtle art to correct for systematic errors, and you can never be completely sure you’re doing it right.

“First you say that isn’t how you know you have it right. Second you say that if you get a different result, that’s how you know you have it wrong.”

No, you didn’t read what I wrote. I didn’t say “that’s how I know I got it wrong.” I said I would assume, in the face of overwhelming contrary evidence, that I had made a mistake somewhere.

I don’t see the difference between what I said and what you said.

This is radically different from the Millikan situation, where people took his one single value as authoritative and adjusted their reports accordingly. Simply put, when the countervailing data is one point, then the authority of the person producing that one point can drive others’ reactions to it. When the countervailing data is almost literally the whole of known physics, it is maximally parsimonious to assume a mistake on my own part.

Yes, of course. And this explains the graphs Joshua pointed to. If the original estimate was wrong by 20% or more, then maybe each later researcher gets a result which is closer to the truth than that, but each of them assumes that they have made a mistake and looks for ways to “correct” it toward the previous answer. Each time, the result is closer to correct than the last published result and gets adjusted to be close to that one. So the answers keep creeping closer and closer. My thought is that as the precision goes up, the creep toward better answers slows down. Because too many physicists stupidly assume that precision implies accuracy, so they stop considering systematic errors that have big effects and consider only systematic errors that have effects small enough to keep them close to the currently-accepted value.

Of course, it’s possible that the reason the progress slows down is that they have found all the large systematic errors and correctly fixed them. But how likely is that? ;-)

It might be particularly likely in physics where you can study something extremely simple in isolation, and the mechanisms you use to keep your simple items isolated are also beautifully elegant with hardly any moving parts, changing fields, or measurements required.

It’s called the quantum metrology triangle, and it describes the three-sided relationship between Planck’s constant and the electron charge (three sided because you need three independent experiments to determine both). Now, the electron charge is known down to something like 9 significant figures, and from the QMT, once you know e, you know h, so if e were different by 2% then h would be different by 2% and if h were different by 2% then all of black-body physics would look very different and if black body physics blah blah blah etc. and so on.

So yes, if that happened you would get a different result. A lot of other numbers that were calculated from these constants would have to be revised. After those revisions, would you get results that were as compatible with reality as the existing results? I guess that could be determined, but it looks like a lot of work. It might be more interesting to look specifically for nonlinear results, things where the shape of the curve would be distinctly different. Then check what difference that would make. You might possibly get changes that could not be fixed by recalibrating measurements.

The whole thing has to hang together coherently or it just doesn’t work.

This is an excellent reason for physicsists to doubt their lying eyes. Get a result that appears to be a little off, so you look for errors in your work that can be corrected to get it closer. Because the whole thing does work, therefore it’s right. Your observation doesn’t quite fit, therefore your observation is off.

We have evidence that it goes that way for measurement of constants, why would it not work like that in a wider context?

I don’t see the difference between what I said and what you said.

The difference is between claiming what you know, and evaluating how likely you are to be right given other data points; that is, estimating (if only notionally) a Bayesian posterior. And that’s a sensible way to read those PDG plots too, because if you’re a good Bayesian, you’ll take the previous data point into account.

If the original estimate was wrong by 20% or more, then maybe each later researcher gets a result which is closer to the truth than that, but each of them assumes that they have made a mistake and looks for ways to “correct” it toward the previous answer. Each time, the result is closer to correct than the last published result and gets adjusted to be close to that one. So the answers keep creeping closer and closer.

But that’s not how it works, and this is where it all bottoms out. It’s not that every successive researcher adjusts their values towards the previously best value. Of course, it’s not impossible that one or two people do this, especially with tabletop experiments, but the institutional nature of a large-scale team-based practice like particle physics mitigates against that. When you’re just one person in your own lab, you could conceivably “get away” with doing this, but when we’re talking about large collaborations where data is semi-public, it’s really difficult (I’d say pretty much impossible) to get away with any actual malfeasanse.

Of course “adjusting” values is not the only way to be influenced by past results. If a past result was obtained by some experimental technique that has some systematic error built in, and you replicate the experiment with the same technique, you might incur the same systematic error even if your technology is better. Or if you pick an old method of analysis where a new one would give you more correct results because an older experiment used that older method. All these things are possible and do happen; I’d suspect, though I don’t know for sure, that when you see the jumps in those graphs it’s because some systematic error was eliminated either through a more careful experiment or through better analytic tools. But most scientists, especially those working in large collaborative projects, just aren’t sitting around picking values that accord better with past measurements.

My thought is that as the precision goes up, the creep toward better answers slows down. Because too many physicists stupidly assume that precision implies accuracy, so they stop considering systematic errors that have big effects and consider only systematic errors that have effects small enough to keep them close to the currently-accepted value.

I don’t know why you would think this. What do you know about the actual practice of what physicists do that would lead you to believe such incredible things? I mean, if taken seriously, this is a claim of serious scientific malpractice!

Here’s a story that’s playing out right now. You may have heard back in March that an experiment called BICEP2 had claimed a detection of primordial gravitational waves. Along with that detection, BICEP2 reported a number, called the tensor-to-scalar ratio, r, and said that r = 0.2 to some pretty high certainty. Nevermind what this number represents, the important thing is that we have it. Now, as it turns out, upper bounds on r have previously been produced by the Planck satellite mission, and Planck has high certainty that r < 0.11. Obviously this is a huge discrepancy. Initially the BICEP2 team claimed that their data was very clean but upon closer examination, it now appears possible that they used an incorrect model for the dust foreground and that the signal they're actually observing is either the inflationary signal plus the foreground or even swamped by the foreground entirely. This analysis was carried out by the Planck team, with assistance from cosmologists unaffiliated with either project. You can read a pretty good summary of it here, but the short of it is that BICEP2 has acknowledged the flawed analysis and gone back on their claims (which, if true, would be an explosive discovery in cosmology; the “incentive” here is not to hedge towards the previous value but to move away from it) and people are now digging through the data to try and figure out what happened here and what can be done better. Future experiments will obviously seek to learn from this and avoid BICEP2’s problems (partly caused by measuring at only a single frequency).

That’s how these things actually work. Huge amounts of human intelligence are being directed precisely towards figuring out just what the systematics are and how they can be accounted for. To be sure, there are imperfections in any institution, but the suggestion that scientists aren’t taking systematic errors seriously is unwarranted and contrary to every standard practice I’ve ever encountered.

So yes, if that happened you would get a different result. A lot of other numbers that were calculated from these constants would have to be revised. After those revisions, would you get results that were as compatible with reality as the existing results? I guess that could be determined, but it looks like a lot of work.

I am reminded of Babbage’s famous quote: ‘On two occasions I have been asked, — “Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?”… I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.’

How would it be possible to get results that were “as compatible with reality” if the fundamental constants were different? That’s the whole point, they wouldn’t be compatible with reality; the reality of a different set of fundamental constants would look radically different; it’s not just a question of changing some values around and getting out the same physics as you had before but just with a different set of numbers.

It might be more interesting to look specifically for nonlinear results, things where the shape of the curve would be distinctly different. Then check what difference that would make. You might possibly get changes that could not be fixed by recalibrating measurements.

Imagine a simple threshold, like the work function of some metal or the ionization energy of an ion. If that number is different, then the physics of everything based on that is different. All sorts of implications cascade right out of that. I guess you could call it a nonlinearity if you like.

This is an excellent reason for physicsists to doubt their lying eyes. Get a result that appears to be a little off, so you look for errors in your work that can be corrected to get it closer. Because the whole thing does work, therefore it’s right. Your observation doesn’t quite fit, therefore your observation is off.

We have evidence that it goes that way for measurement of constants, why would it not work like that in a wider context?

Yes, because the whole thing does work. The entire structure has to cohere; if according to your experiment some piece of it is sticking out in contravention of other known results, then you should be suspicious. It doesn’t necessarily mean you’re wrong, but it very well might. In a case like BICEP2, it’s plausible that the BICEP2 number for r is correct. It doesn’t seem consistent with Planck’s number, but there’s no a priori reason to say that BICEP2 is wrong. If you measure a radically different lifetime for the neutron or a radically different electron charge (or you claim to detect superluminal neutrinos, say), then odds are overwhelming that you are wrong and you should start examining your experiment and analysis pipeline for errors. It’s a process which has actually proven wildly successful, as those graphs indicate, so I’m not sure what exactly the problem is supposed to be here.

As to the wider context: of course it happens. It’s called the conventional wisdom for a reason, and even in some scientific fields it’s stronger than others. But that’s why things like “academic integrity” actually matter; if physicists behaved like Dershowitz, the Planck and BICEP2 teams would be throwing feces at each other, not sitting down like civilized human beings and trying to figure out which one of them is actually right. All that shows you is that reasoning norms are important and need to be enforced.

Is there any way we could estimate which of these fits the data better? Are there any ways to do experiments that might reveal the effects of both?

There’s no general answer to this question. You have to consider the specific theory to figure out what can be done in this regard.

I dunno, you’d have to try it and see. There might be more than one way, unless you already have a system that explains everthing. If you have a system that explains everything perfectly with no error and nothing unexplained, then any alternative that worked as well would have to be completely equivalent — just as Copernican orbit theory was equivalent to a sophisticated ptolemaic system.

I’m not aware of any theory that explains everything. But the dependencies are important; you can’t “try it and see” because there’s only one universe (that we live in, anyway) and we can’t run it multiple times to see what happens, so you have to rely on what other factors of your system are going to be affected. And the thing is that we have lots of results that we know, to a ridiculous level of precision, would look very different if those constants were different. It’s just not a good counterfactual at all, because you can just look at the equations and the differences will jump out at you. There’s not a good distinction between “reality” and the “model of reality” at the fundamental level. Like, what you’re trying to say is that the equations could be different and reality would still be the same, but that’s not possible if you accept that the equations are an accurate model of reality.

Nowadays we know that it doesn’t really matter where you decide the center of the system is, the math will work out with any center, just some approaches are more tedious to compute than others. But if your system doesn’t explain everything, then it’s possible an alternative might work as well if it had as much effort put into it as this one has.

Yes, we have the benefit over the Ptolemaics and Copernicans of having a more sophisticated mathematics. What you’re really doing is making the old Quinean argument about theory being underdetermined by the data. And while it’s logically true, in practice, it doesn’t actually end up being a problem. It turns out that if you have a special theory of planetary motion over here and another special theory of projectile motion over there, you might come up with a way of changing one without affecting the other. But as soon as you realize that these special theories are really cases of a general principle of gravitation, then the interdependency of the theories becomes clear. And by and large the story of physics is the (wildly successful!) story of unifying these special theories under general principles.

Astrologers have gone to a lot of trouble to make a self-consistent system.

That’s because the governing meta-principle is not intra-theoretic consistency but rather inter-theoretic consistency.

I think the farther you get from things that can be measured directly, the more room there is for fudge factors that will warp incorrect constants into acceptable results.

I guess, but it’s not like biology or sociology or even history consider quantification to be unimportant. I don’t know of any subdiscipline of science that doesn’t fundamentally boil down to making measurements. Of course, there’s a lot less room in physics for divergence than there is in other fields; even the “central dogma” of biology is not anywhere near as restrictive as the laws of thermodynamics, and the landscape of “constants” (whatever those might be in biology) is much less firmly established. But from everything I know of biologists (many of my friends in grad school were biologists) they’re much more likely to push a result that’s out of the mainstream than one that adheres to “known values.” There are a lot of reasons for this: partly, there’s the fact that “replication studies” are not considered terribly respectable. Partly, the fact that the landscape is less well-established means that there’s more of an opportunity to stake out your own corner of that landscape. And of course a more sensational result, if true, leads to higher-profile publications. So all the incentives that I am aware of drive scientists in the direction not of conformity but of pushing the boundaries.

I’ll try one last time and then let this die.

Have you noticed how much of your reality consists of fudge factors? It might possibly turn out that with a different set of fundamental constants you could get a different set of fudge factors which mostly work out adequately. It might possibly work out better on average. Maybe not with exactly the same fundamental laws, but you just might find a way to make it work better.

What does “work out” mean? I’m trying to take your point seriously, but you have to understand that any other valid description of reality would have to reproduce what we already know to be true, and as such it (or a subset of it) would be isomorphic to the current formulation. There are fudge factors and then there are fundamental constants and they really do have a different status within the theory. What you’re describing is not possible so long as you’re committed to a mathematical formulation of your theory, because the outputs of your theory just are determined by the mathematical structure.

You are sure that is impossible because you are impressed at how well the current system works. But a large number of very smart people have worked hard to fit the current system together. I say you can’t know how well they’d do with an alternative system unless they actually try it.

You know what the capitalists say: there is no alternative. Seriously, any alternative would, in a fundamental sense, look just like what we have now, except rotated 90 degrees (or whatever). This has nothing to do with what I want and everything to do with the structure of the description. Any other description has the constraint that it must explain all the things that the current description explains, so in some sense it will be the current theory. It’s like if you took the QMT I talked about above and said that the fundamental constants were actually the Hall resistance and the flux quantum, and then we went about rewriting everything with the appropriate substitutions.

For this reason, and for the reason that the current system works splendidly, there’s no incentive to try anything else. All those smart people know this, which is why there isn’t a cottage industry of alternatives in physics like there is in, say, philosophy, where you aren’t bound by the constraint of having your theory stand the test of experiment.

Jerry @115: philosophy, where you aren’t bound by the constraint of having your theory stand the test of experiment

Yes, but does physics have trolley problems?

#18 JV

“What you guys are doing is not likely to lead to much progress. Somehow physics has been dominated by people who want to believe that they are right. It isn’t just you, I run into this pretty consistently.”

And here we bottom out at the fact that you really don’t know what you’re talking about.

You might be delving too deep into the roots of the trees to see the forest.

You gave me the impression that you aren’t ready to have an informed opinion on the topic when you argued that tight error bounds imply the systematic errors have been fixed. I had mostly failed to get Joshua to understand about this — he didn’t respond to hints, and finally you sort of started using the argument after the second time I presented it, but you kept using the wrong argument too.

It really helps to play Eleusis.
https://en.wikipedia.org/wiki/Eleusis_%28card_game%29

Also it might possibly be useful to play Black Box.
http://www.bibeault.org/blackbox/

I suspect we have nothing left to say to each other.

Probably. But I hope you’ll try out those games. The second is online and is kind of fun to try for awhile. Eleusis is a great game that can teach a whole lot about scientific method. Scientists at work don’t get to actually do scientific method all that much because there’s so much else that has to be done, but in the game there just isn’t much else to do.