What Scientific Ideas Are Ready for Retirement?

Every year we look forward to the Edge Annual Question, and as usual it’s a provocative one: “What scientific idea is ready for retirement?” Part of me agrees with Ian McEwan’s answer, which is to unask the question, and argue that nothing should be retired. Unasking is almost always the right response to questions that beg other questions, but there’s also an argument to be made in favor of playing along, so that’s what I did.

My answer was “Falsifiability.” More of a philosophical idea than a scientific one, but an idea that is bandied about by lazy scientists far more than it is invoked by careful philosophers. Thinking sensibly about the demarcation problem between science and non-science, especially these days, requires a bit more nuance than that.

Modern physics stretches into realms far removed from everyday experience, and sometimes the connection to experiment becomes tenuous at best. String theory and other approaches to quantum gravity involve phenomena that are likely to manifest themselves only at energies enormously higher than anything we have access to here on Earth. The cosmological multiverse and the many-worlds interpretation of quantum mechanics posit other realms that are impossible for us to access directly. Some scientists, leaning on Popper, have suggested that these theories are non-scientific because they are not falsifiable.

The truth is the opposite. Whether or not we can observe them directly, the entities involved in these theories are either real or they are not. Refusing to contemplate their possible existence on the grounds of some a priori principle, even though they might play a crucial role in how the world works, is as non-scientific as it gets.

I’m also partial to Alan Guth’s answer: “The universe began in a low-entropy state.” Of course we all know that our observable universe had a relatively low entropy at the Big Bang; Alan is making the point that the observable universe might not be the whole thing, and the Big Bang might not have been the beginning, so it’s completely possible that the universe as a whole was never in what one might call a “low-entropy” state. Instead, starting from a generic state, entropy could increase in both directions, leading to a two-sided arrow of time. This has been one of my favorite ideas for a while now, and Alan and I are writing a paper with Chien-Yao Tseng that examines toy models with such behavior.

Here are some other interesting/provocative answers, picked unsystematically out of over 100,000 words overall. Remember that the titles are what the person wants to retire, not something they’re in favor of.

  1. Either science is a tool to answer practical questions, or it is a tool to satisfy our anxiety over the uncertainty we experience in our existence. It can’t be both. Falsifiability makes science practical rather than psychotherapeutic.

    Put another way, In deciding to spend our money, how much should we devote to speculative imagination, and how much to practical solutions? Does it really matter how well we understand string theory if its not going to be testable or yield practical applications in any foreseeable time? Are there better problems to focus on? Who gets to choose, the academics, or the indigent?

  2. I’m glad you point out this is meant to be provocative because I have a hard time believing you are genuinely against the idea of Falsifiability (assuming *your* understanding of it is sufficiently nuanced).

    Falisifiablity is as much about improving theories as it is about demarcation. I’ll take for granted you are an empiricist and believe hypothesis are only that until they can be confirmed by evidence (especially theoretical physics!). They must also have some predictive power.

    The problem with relying solely on confirmation is that many hypothesis can stand up to endless confirmation if you are only looking in one place. To really make your theory strong you need to search for evidence which could refute it. For example, edge cases. The classic example of “water boils at 100c” is a hypothesis and you can test it a million times at sea level. Every time it will be confirmed. It’s not until you take the experiment to the top of Everest and “break” (falsify) it do you truly begin to learn and find the limits of the theory.

  3. Doc C,

    Why can’t it be both, to the same person, or either to different people? I’d have thought it was both. Some people may start out in science wanting to satisfy uncertainty (curiosity rather than anxiety – why anxiety?), but in a science career can change the focus to practical questions. I guess the nature of theoretical physics/cosmology retains the quest for understanding more so than much applied science which is more about practical answers.

  4. Ron, it’s an epistemological question. Do you really “know” anything if it cannot ever be confirmed or make testable (empirical) predictions? In my opinion you cannot. Empiricism is king. Therefore if you are doing purely theoretical science that can NEVER be tested in practice I would say you are no longer searching for knowledge. At which point its no longer science.

  5. Danny Hillis : I’m immediately turned off by the phrase “we humans”…

    the rest of the links are taking an extremely long time to open.

  6. It is unclear what qualifies Nina Jablonski to urge that we ignore almost the entirety of genomics and biomedical research as people “seeing what they want to see.” A more unscientific attitude would be hard to imagine.

  7. Edge is down, but I already agree with your article. It’s worth knowing that falsification was basically proposed as a solution to the problem of induction, and is meant to replace inductive reasoning. IMO, falsification makes no progress in solving the problem of induction, and is nothing but a clunkier version of inductive/bayesian reasoning.

    The other thing is that though scientific theories must ultimately be subject to empirical investigation, I don’t think it’s necessary for this to be true of every single nut and bolt of science. Scientific practice contains both induction and deduction, and there can be entire careers which focus only on deduction.

  8. not to get off topic here, but Doc C, you can do whatever the fuck you want in life if you can find someone who will pay for you to do it.

    Bill Gates made billions in the tech industry so that he could try to eradicate diseases in arguably the most disease ridden regions of the planet. Bill Gates can do whatever the fuck he wants to do with his $50 Billion because “WE” have no say in what he does with that money. We can donate money to the Bill and Melinda Gates foundation if we choose to do so, as could the government of any country.

    Since understanding String Theory doesn’t cost a damn penny, I’d say the freedom of choosing what you do with your career and/or free time is a decision that every individual is granted the right to. If you’re focus is string theory, then chances are you get paid in various other practices.

  9. Not physics, but psychology: I still see many people quoting Sigmund Freud, even though most of his findings have been shown to be wrong, or even harmful. Of all the scientists with overturned theories, Freud seems to have the longest-lasting and unwarranted prestige.

  10. If you have a location in the past, and you are trying to describe the energy that makes up that past , I think you have to keep looking even when you come up against the horizon? So the attempts to describe the interior and ways to approach it are still necessary. Confirmations have been made up to this point, and a method for examination is still recognized.

    Whether or not we can observe them directly, the entities involved in these theories are either real or they are not. Refusing to contemplate their possible existence on the grounds of some a priori principle, even though they might play a crucial role in how the world works, is as non-scientific as it gets.

    So would we say in this respect this is non scientific? Just wondering?


  11. Karl Popper replaces the focus on the justification of beliefs with the process of forming critical preferences for theories or policies formulated to solve problems. Better theories explain more than lesser theories, predict more precisely, articulate with other theories and they stand up to tests.

    With respect to the latter, we should attempt to state our claims in a form that will invite refutation by existential statements (if true). Thus falsifiability is a logical property of a statement. Falsification is the process of demonstrating that a proposition has been falsified. Unlike the decisive logic of the process, the real-world process of falsification can never be decisive due to the Duhem problem, the uncertainty of observations and the many more or less disreputable ways that people can protect their views.

    Thus, science is that branch of the arts that presents its explanatory claims in such a form as to be testable against the world. The demarcation between science and non-science has nothing to do with methods of discovery or technologies of discovery but rather with the perpetual and normative requirement to make scientific statements logically amenable to being proven wrong, even if in practice it is a matter of decision. Furthermore, if one is asked what is the aim of science, it is to explain whatever one deems requires an explanation. It begins with problems, not with data. How can minds or their technological surrogates even perceive data if they do not possess modifiable expectations or propensities to solve problems?

    Furthermore, building tools is not the same as doing science. Scientific theories may be instruments but they are not only instruments. They explain. A shovel or an electron microscope is not true or false, a scientific theory can be; mind you any claim to be true is only tentative – otherwise it is not science! Our conjectures come first as in all arts. Let us praise imagination, but remain humbled by our propensity for error. It is the humility of science that gives us hope that we will not be consumed by technological hubris.

    Discard the norm of trying to make scientific statements able to be proven wrong if you like? What will you replace it with?

  12. Brett,
    1. You are way off topic.
    2. We all paid handsomely for the development of string theory in the taxes that went to supporting research grants and to institutions of higher education who used that money to subsidize faculty and grad students who work on it.
    3. The question I addressed regarding falsifiable science is not whether anyone can do anything they want with their money, but how groups of people who organize into a cooperative society should decide how their common pool of resources are used.
    4. The use of vulgarity does not improve an argument. It generally reveals a lack of one.

  13. Pingback: Time to ditch falsifiability? | SelfAwarePatterns

  14. You can’t prove that you’re not a BB, just like you can’t prove that you’re not dreaming, or being fed false information by a whimsical demon. Nor can you make any progress on understanding the world in reliable ways if any of those scenarios is true. Therefore, we work to develop models in which they are not true.

    This is how you previously rejected the theory that I’m a Boltzmann Brain dreaming of a non-thermal-equilibrium universe. A theory that actually satisfies both of your “definite” and “empirical” criteria. It’s “definite” because it “say[s] something clear and unambiguous about how reality functions”, and it’s “empirical” because it of “its ability to account for the data”.

    Whether or not we can observe them directly, the entities involved in these theories are either real or they are not. Refusing to contemplate their possible existence on the grounds of some a priori principle, even though they might play a crucial role in how the world works, is as non-scientific as it gets.

    Question: If the dog didn’t bark in the night-time, should we take that as a clue, or should we get rid of it?

  15. Pure mathematics has propositions that have not been proved
    or disproved and as such are an enjoyable challenge for the professional or amateur. Number theory is replete with such unproven (e.g. Goldberg’s Conjecture) nuggets that await being proven but are yet open to being shown a counter example. But the sciences per se seem to be more susceptible to pressure from non-science agendas
    then the science’s supporting mathematics. This was true in Galileo’s time as well as now. So I see Popper’s rule as a bulwark against such pressures.

    Whenever I hear the phrase ‘the data show’ the expectation is that one cannot argue against ‘the data’, as if that itself were irrefutable and therefore the argument is moot. The old saying among statisticians was ‘torture the data until it confesses’. I have been involved in research as a support for scientists and they can get quite opinionated as any. Being rational and unbiased is an ideal goal, as difficult for the scientist as the lay person.

    Current topics that are highly politicized such as; differences between the sexes, IQ and race, the concept of race, global warming (now called climate change), et al. Can a researcher even pursue, unbiasedly, such fields without being attacked. Hardly.

    So for me, as a former practicing statistician, my concern is that the scientist has many pressures to assume a theory to be true ipso facto. To discard the Popper Rule is to weaken the scientific community’s defense from non-scientist pressure to conform to the zeitgeist.

  16. I started reading but was stopped in my tracks when I got to this in the second paragraph:

    “We knew, for example, that the electron mass is the same everywhere in the observable part of the universe, so the obvious assumption was that it must take the same value everywhere, it is just a constant of nature”.

    That’s wrong! There’s the mass deficit or mass defect. The electron’s rest mass varies with say gravitational potential. If you drop a 1kg brick into a black hole, the black hole mass increases by 1kg, no more. And on the way in the brick was moving very fast. Its kinetic energy comes from its rest-mass energy. I also noticed Freeman Dyson saying wavefunction isn’t real. He should talk to Jeff Lundeen who said this:

    “…So what does this mean? We hope that the scientific community can now improve upon the Copenhagen Interpretation, and redefine the wavefunction so that it is no longer just a mathematical tool, but rather something that can be directly measured in the laboratory.”

    I have to say Edge sounds like a promotional clique for people who want to tell everybody else that they’re the elite.

  17. Ian Mcewan is right… retire nothing …but we can reasses and think things over from new perspective.
    Much trouble in cosmology comes from years of missunderstanding …miss interpreting.
    Newton went as far as he could …or dared . Einstein did the same
    an apparent attraction … curved space time …
    all just words trying to explain observations

    We need to re think Gravity from the universal law on wards .
    What masses are doing .. in producing tjis force ..
    they do not create gravity on their own .
    Force comes from Energy .. the energy in space …from all directions .. radiations and partickes .

  18. A theist can can make the same argument about yhwh or the spaghetti monster.

    In the realm of physics, the math can support e.g. tachyons. Does it mean we need to accept them as real? If not, why not?

  19. I find the falsifiability argument very useful, for myself and in my work. At work I often get fundies arguing for Intelligent Design or Young Earth Creationism. My approach is to ask them if there is anything that would convince them they are wrong. They usually look suspicious (of the trick question), and agree there is not. (I think they feel this strengthens their argument.) I can then reply that there is no point in talking further.

    There are certainly things that would change my mind. If the heavens opened up and a booming voice said, “I didn’t use evolution!” – that certainly would, even though I seriously doubt it will happen.

  20. It is tempting to think that science needs the precision of philosophical analytics, as science (especially physics) make improvements that far exceed a clear path to falsifiability. This is however what makes Popper even more relevant now than when he proposed demarcation. The further we step into mathematical concepts as a pure explanation of physics without the potential for observation, the less likely we are to further fundamental truths. It is good that the best minds are working on the most complicated problems from a standpoint which is not limited to falsifibilty, however letting our notion of a physical reality float endlessly into intelligent mathematical concepts, leaves us unable to make the steps necessary to provide a practical understanding of stuff, and therefore loses experimental rigor at the expense of mathematical rigor. Both require and should have rigor. Perhaps Popper is more relevant now than ever.

  21. “I have to say Edge sounds like a promotional clique for people who want to tell everybody else that they’re the elite.”

    It certainly does have a waft of that about it.

  22. Dear Sean, I’m flabbergasted! Here’s your 2nd paragraph with only a few point modifications:

    “The truth is the opposite. Whether or not we can observe Him directly, the Creator involved in this theory is either real or He is not. Refusing to contemplate His possible existence on the grounds of some a priori principle, even though He might play a crucial role in how the world works, is as non-scientific as it gets.”

    No further elaboration is called for. But let me only add that, unlike in mathematics, nothing in science is really definitely provable. (That rigorous proofs do exist in mathematics only highlights the fact that math is tautological in nature.) So if one exempts oneself of falsifiability, one is left with nothing but speculation. It may be a very “scientific” speculation, construed as science due to the question asked being conventially categorized as scientific, or because it uses exclusive math tools and layperson-dazzling lingo typical of science, yet it is speculation all the same.

    And one more thing. Science fights for its life—against antiscience, pseudoscience and post-modernist relativism—in the public opinion arena. This kind of transgression provides the enemy with ammo: New Agers may rightly ask “Why, then, are you any better?”

  23. Well said DEL. And Dan and Andy and Matthew and Doc and Jack.

    Sean, if you retire falsification you retire science. I’m not happy with that. And a lot of other people won’t be either. This is not going to do your career any good at all.

  24. I doubt that you want to retire falsifiability. You just don’t think it is a requirement for an idea to be science.

  25. I am not bothered by falsifiability. It is one solid criterion but not the only one. I am more concerned of this blatant gimmick to redefine the scientific method to suit theories which fail to meet any of its criteria. It’s not going to fool anybody. If we reach a point where the scientific method fails, then we say the scientific method can not tell us anything beyond this point here. At that point, we can use other means to explain what’s going on but we don’t pretend that we are conducting scientific inquiry because we’ve cleverly just redefined what we mean by science.

  26. “Amanda Gefter, Andrei Linde, and Seth Lloyd all suggest that we get rid of the idea of a unique universe, each from a slightly different perspective. Sorry, universe: the tide is turning against you.”

    I suspect the universe is still sleeping quite well at night with such weak arguments arrayed against it.

  27. Whether or not we can observe them directly, the entities involved in these theories are either real or they are not.

    That’s setting the bar very low. Santa Claus is real or he is not. The easter bunny is real or it is not. Invisible pink unicorns are real or they are not.

    Does science tell us anything about whether these things are real?

  28. So if we can’t touch it it can’t be real? If ‘Santa Claus” inspires a child to buy a toy for an indigent child (because Santa can bring that child toys, but his family can’t, so the first wants to help Santa), is Santa real? He has a real effect.

  29. If you drop falsifiability, an infinite number of theories will all be right. On what new basis will you know the right one?

    “Darkness gathers. The demons begin to stir.”

  30. Sean Carroll is an excellent theorist, and I wish I had his deep understanding of string theory and most of the rest of physics for that matter, but he’s really wrong here. And it’s quite clear why he’s really wrong– what does he think is the actual process whereby “nature is the ultimate guide”? Falsifiability, obviously! Seriously, we have a huge body of evidence from which to answer that question, it is called the history of physics. In that history, it is completely demonstrable that the process for discarding theories is not coming up with ones that a consensus of people like Sean Carroll prefer for personal reasons, they get discarded for only one reason: they got falsified, because they were falsifiable. What happens to what is unfalsifiable is also a matter of record: it becomes folklore.

    A classic example of this is the geocentric model of the solar system. For millennia, that model was adopted because the consensus of the Sean Carrolls of the day was that the stars could not be so far away that they would not appear to move, it just didn’t make sense to them for the universe to be built that way. So if personal preferences, rather than falsifiability, ruled the day, we’d still be teaching the geocentric model. But when Galileo saw that Venus went through phases, it meant Venus had to orbit the Sun, period, end of debate. Thank goodness for falsifiability!

  31. Dear Mr. Carroll
    I was wondering if you can answer the question or consider writing an article on the subject:

    Has dark energy or dark matter changed in the life of the universe?
    Is it possible for one to slowly change into the other over the life of the universe?
    For instance, if dark energy can change into dark matter, then perhaps the universe will slow it’s expansion, perhaps even stop it’s expansion, reverse and a big crunch will occur.

    Is there any evidence or reason (beyond my pure conjecture and perhaps wishful thinking) for this to be?

    Yours truly

  32. Mr. Carroll,
    You might seriously consider writing a retraction, or something that perhaps clarifies your position to some degree better than you have. Your position on jettisoning falsifiability from science is not logically sound, namely because you are abandoning the underpinnings of logical argument itself, without which there is no science.

  33. Dear Sean Carroll

    In your ‘Edge’ piece you advocate for the retirement of ‘falsifiability’ and yet you end your piece with declaring Nature our ultimate guide. To me, this is completely self-contradictory. If Nature has no way of saying “you are wrong!” – which is falsifiability – then to me it appears you’re taking it along, not as a guide, but more like a Teddybear to hold your hand.

    As a physicist I’m deeply troubled by seeing someone in a significant position advocating the abandonment of falsifiability. Yes, it might be that string theory is correct without being falsifiable. Yes, it might be true about the multiverse too. And it might be true about God and about the giant teacup theory. But for Science to have any value, we have to stick to the perhaps limited position of only accepting statements and theories, where Nature is left a right of veto. Otherwise we’re not doing Science anymore.


  34. Falsifiability is just one in a whole set of requirements. Others are, e.g., internal consistency, mathematical distinction, Occam’s Razor compliance and, lest some priviledged Edger might suggest to retire it, agreement with the complete existing body of experimental and observational results to within their respective error bars.

  35. Pingback: Shtetl-Optimized » Blog Archive » Retiring falsifiability? A storm in Russell’s teacup

  36. Hi Sean,

    It’s been long time since I publicly, sort of, insulted you—I mean, online. (I vaguely remember the development of a facility to shuttle in between the universe(s (!)) of multi-verses. … Come to think of it, I got it almost right.)

    Never mind.

    With that (Marathi) “namanaalaa ghaDibhar tel” [rough English sense of the term even if not the translation: When you pour a pound of oil to the lamp of the God right before the beginning of a prayer… [meaning: you are so stupid, such a newcomer, you wouldn’t even know how to make a candle burn *all* through the night at a church]], even without the full of your post or your readers’ replies, let me answer in brief:

    Instantaneous Action at a Distance.

    We could do without it.


  37. Maybe it is worth recognising that Popper’s idea of falsifiability was a vast improvement on the positivist attempt to separate science from non-science in terms of sense or meaning. Popper is not saying that non-falsifiable statements are meaningless nor even less true than scientific statements. For instance realism and anti-realism are metaphysical and cannot be refuted. This does not mean they are beyond argument. Popper liberated metaphysics from a narrow positivist treatment.

    Popper would say: We explore the universe by conjecture and we explore our conjectures through logic (rationality filters our guesses). Falsifiability is a logical property of scientific statements.

  38. Pingback: Methods: testing an idea

  39. For many years I have been an e-mail subscriber to Edge and I usually read most of the monthly conversations there. I made it a rule, however, to skip the annual question and all the answers given to it, for three main reasons: 1) The sheer volume of them. 2) The mostly mediocre quality of the answers, which reafirms what Buddhini Samarasinghe wants to get rid of:

    ◾Buddhini Samarasinghe, Scientists Should Stick to Science.

    3) If I’m technically prevented from even proposing an interesting answer, since Edge excludes mortals such as I, (i.e., those that do not cater to Mr. Brockman’s publishing business,) then I’m not interested in their tedious answers—let them read one another ad nauseam.

    Several years ago I attended a Nobel laureates seminar in the Technion, Haifa. One of the speakers was a co-winner of the 1991 prize for the invention of the patch clamp neuroscience lab technique. He told the audience that, being a rather shy person, he had hesitated whether to accept the invitation. Consulting his mentor, the latter had told him briskly: “A shoemaker should be making shoes.” (But he had accepted anyway.)

    Take, for example, the first answer, by

    ◾Danny Hillis, Cause and Effect.

    As an example for the unsoundness of the cause-and-effect concept, even in classical mechanics, he brings up the case of Newton’s 2nd law, F=ma, where one cannot say which of the three quantities is cause and which are the effects. But this only shows a misunderstanding of that law, which is nothing but a definition of inertial force (mass being defined, as by Newton himself, to be “the quantity of matter,” [Principia, Book 1, Definition 1]), and a failure to realize that causality entails that the events involved be time-sorted. Incidentally, in the LHC, which is the cause and which is the effect, the hadrons colliding or the Higgs boson signature detected? If you know the answer to that you know also what to think of Hillis’ suggestion.

    Yes, shoemakers should be making shoes and scientists should stick to science.

  40. I am troubled by the treatment of falsifiability as a single theory. There are strong and weak versions with many gradations. Is the theory not falsifiable right now with present methods? In the near future with plausible methods? Eventually with purely theoretical methods? Probably unattainable methods? Not falsifiable with any conceivable test? I would submit that at least the last is certainly a valid notion.

  41. From reading the enlightening comments, the post are against your idea of rejecting the idea of falsifiability as a scientific concept to reject. The posts give excellent arguments against your proposed idea. Floating out such ideas is a wonderful way to learn.

    Science is based on experimental data. Theoretical and experimental physics ultimately make their advancements based upon data. Theories need to be tested with data. If a theory cannot be tested against data it is not science, it is philosophy. A scientific theory must have the ability to be shown wrong with experimental data.

  42. The practice of Science has always been ad hoc, a gradual groping
    for a way of proceeding that seems reasonable– “falsifiability”
    is one an attempt at firming this up, and it seems funny to me
    that so many people here are trying to treat it as though it’s
    some sort of primary revealed doctrine.

    Falsifiability itself isn’t a scientific idea, right? Most of us
    would call it a philosophic principle (after all: “Is
    falsifiability falsifiable?”). So if it strikes you as a
    critically important idea, you’ve already conceeded that there
    are things outside of science that matter.

    An attempt at advancing science like “string theory” might not be
    precisely “scientific” (as of yet) if you’re a true believer in
    falsifiability, but in that case you could just put string theory
    in another category (“proto-scientific”?). String theory need
    not be abandoned as Unscientific and Unclean.

  43. @ Joseph Brenner.

    “… So if it strikes you as a critically important idea, you’ve already conceeded that there are things outside of science that matter.”

    – yes, of course! The concept of falsifiability – which simply means that you’re giving Nature a right of veto – does not mean that only falsifiable statements are worth while considering. It simply means that only (in principle) falsifiable statements can be called scientific theories.

    Most things that matter is outside of science – obviously.

  44. If you have a (mathematically) complex theory which currently does not make predictions, but you think it will after 1 year of research, is it scientific? What about 5, 10, 50, 100 years? At what point does it become nonscientific?

  45. @ Richard. In my opinion, to call a mathematical framework (or something else) a theory requires predictions (they may not be practically verifiable). If your framework doesn’t make predictions, then it may of course still be very interesting and worthwhile pursuing, and it is still scientific – its just not a theory.