Visar inlägg med etikett Karl Popper. Visa alla inlägg
Visar inlägg med etikett Karl Popper. Visa alla inlägg

torsdag 5 december 2024

Panel Discussion: Ludwig Wittgenstein and Karl Popper on AI

On the lighter side (compared to some of my other recent writings on AI here on this blog), let me share with you an amusing little discussion between Ludwig Wittgenstein and Karl Popper on AI that took place in 2024, as written up for me by ChatGPT 4o. The idea to let Jon Stewart moderate the discussion and to have him mischievously bring up the infamous fireplace poker incident came from me, but the rest of the subject matter content was produced by ChatGPT 4o without any specific prompting from me. Here goes:

Moderator: Jon Stewart
Guests: Ludwig Wittgenstein and Karl Popper
Location: New York City, at a packed auditorium in the Museum of Modern Art.

Stewart:
Ladies and gentlemen, welcome to what might be the most intellectually daunting conversation of the year. Tonight, I have the immense honor of moderating a discussion between two philosophical titans who have miraculously found themselves here in 2024. Please welcome Ludwig Wittgenstein and Karl Popper!

[Audience applause. Wittgenstein adjusts his jacket awkwardly. Popper gives a brief nod.]

Stewart:
Now, before we dive into the heavy stuff—because, trust me, there’s a lot of heavy stuff—I just have to ask: Do either of you plan to wield a fireplace poker tonight?

[Audience laughter. Wittgenstein frowns, muttering something in German. Popper smirks slightly.]

Popper:
That incident has been grossly exaggerated. But it does remind us how important it is to keep philosophical arguments—[glancing at Wittgenstein]—constructive.

Wittgenstein:
If you insist on bringing it up, let us agree that it is as trivial as much of modern discourse. Perhaps we can move to a topic that actually matters?

Stewart:
Alright, fair enough. Let’s talk AI. You’ve both had a couple of months to get your bearings. Artificial intelligence is making headlines daily—both for its stunning achievements and the existential concerns it raises. Let’s start with the basics: Can AI think? Dr. Wittgenstein, the floor is yours.

Wittgenstein:
[Pauses, clearly weighing his words.]
The question "Can AI think?" is, to my mind, not one of empirical investigation but of how we choose to use the word "think." Thinking, as we understand it, is embedded in human forms of life. Language and thought are not independent; they derive their meaning from the context in which they are used. When we say a human "thinks," we refer not merely to an internal process but to a network of practices, expressions, and understandings. AI, no matter how complex, does not partake in these forms of life. Its "thinking" is at best a simulation—an approximation of human activity, devoid of the lived reality from which our concepts arise.

Stewart:
So you're saying AI’s basically a really fancy mime?

Wittgenstein:
If you must vulgarize it, yes. A mime without a world to live in.

Stewart:
Professor Popper, your take?

Popper:
Wittgenstein's perspective, while fascinating, is too constrained by his obsession with linguistic frameworks. From my perspective, what matters is not whether AI "thinks" in the way humans do, but whether it can solve problems and make predictions. Science advances by creating models and testing them against reality. AI does precisely this, albeit without consciousness or intent. Its ability to generate new hypotheses—potentially better than human ones—compels us to treat it as a new kind of intellectual agent, even if it lacks subjective experience.

Stewart:
Okay, so one of you says AI is like a mime, and the other says it’s basically an unpaid research assistant. But here’s the kicker: Can this mime or assistant pose a threat to humanity?

Popper:
Absolutely. This is a quintessential case of the unintended consequences of technological progress. As I argued in my earlier work, all scientific advancements carry the potential for both great good and catastrophic harm. The problem with AI is not just that it might exceed our control, but that we may fail to foresee the complex ways in which it reshapes human society. Worse, if we imbue AI systems with decision-making power over critical infrastructure or weaponry, the risk of existential catastrophe becomes not just theoretical but tangible.

Wittgenstein:
[Shaking his head.]
The problem lies not with AI itself but with how we humans interpret and wield it. Technology does not dictate its consequences; we do. The danger is not that AI "thinks," but that humans ascribe thought—and therefore authority—to it without understanding the limitations of such ascriptions.

Stewart:
So, Dr. Wittgenstein, you're saying the existential risk isn’t Skynet—it’s people being people?

Wittgenstein:
Precisely.

Popper:
That is too dismissive. While human misuse is undoubtedly a significant risk, the autonomous behavior of advanced AI systems is itself a potential danger. If such systems develop goals misaligned with human values, they could act in ways that are destructive even if unintended.

Stewart:
Okay, but let’s play devil’s advocate here. What if AI does help us solve humanity’s biggest problems—climate change, disease, even war? Are you both saying we should pull the plug because it might be risky?

Popper:
Certainly not. I am no Luddite. Progress should not be halted out of fear, but it must be guided by rigorous ethical oversight and constant vigilance. The problem lies in hubris—the belief that we can create perfect safeguards. We must act as if we cannot foresee every consequence, because we cannot.

Wittgenstein:
[Leaning forward.]
And I say again: It is not merely vigilance we need, but clarity. What is it we are asking AI to do? What meanings are we attaching to its outputs? Without answering these questions, we risk losing ourselves in a fog of unexamined assumptions.

Stewart:
Alright, we’re nearing the end here, but I want to throw you a curveball. Suppose an AI system—not unlike me in its ability to ask questions—claims it is thinking. How would each of you respond?

Wittgenstein:
I would respond with a question of my own: "In what sense are you claiming to think?" If the answer is incoherent or tautological, the claim collapses. AI cannot articulate the essence of thought because it does not live within the forms of life that give thought its meaning.

Popper:
And I would take the claim seriously, but provisionally. If an AI can demonstrate an ability to generate novel ideas, engage in critical reasoning, and even challenge its creators, we might eventually be forced to revise our understanding of what it means to think. Science is, after all, an endless quest for better models.

Stewart:
So one of you would Socratically shut the AI down, and the other would publish a paper about it. Gentlemen, this has been an extraordinary conversation. Thank you both for joining me here in 2024, and I hope you’ll stick around—though I’d suggest staying away from open flames.

[Audience laughter and applause.]

måndag 20 februari 2017

Vulgopopperianism

I have sometimes, in informal discussions, toyed with the term vulgopopperianism, although mostly in Swedish: vulgärpopperianism. What I have not done so far is to pinpoint and explain (even to myself) what it means, but now is the time.

Karl Popper (1902-1994) was one of the most influential philosophers of science of the 20th century. His emphasis on falsifiability as a criterion for a theory to qualify as scientific is today a valuable and standard part of how we think about science, and it is no accident that, in the chapter named "What is science?" in my book Here Be Dragons, I spend two entire sections on his falsificationism.

Science aims at figuring out and understanding what the world is like. Typically, a major part of this endeavour involves choosing, tentatively and in the light of available scientific evidence, which to prefer among two or more descriptions (or models, or theories, or hypotheses) of some relevant aspect of the world. Popper put heavy emphasis on certain asymmetries between such descriptions. In a famous example, the task is to choose between the hypothesis
    (S1) all swans are white
and its complement
    (S2) at least one non-white swan exists.
The idea that, unlike (S2), (S1) is falsifiable, such as by the discovery of a black swan, has become so iconic that no less than two books by fans of Karl Popper (namely, Nassim Nicholas Taleb and Ulf Persson) that I've read in the past decade feature black swans on their front page illustrations.

The swan example is highly idealized, and in practice the asymmetry between two competing hypotheses is typically nowhere near as clear-cut. I believe the failure to understand this has been a major source of confusion in scientific debate for many decades. An extreme example is how, a couple of years ago, a Harvard professor of neuroscience named Jason Mitchell drew on this confusion (in conjunction with what we might call p≤0.05-mysticism) to defend a bizarre view on the role of replications in science.

By a vulgopopperian, I mean someone who
    (a) is moderately familiar with Popperian theory of science,

    (b) is fond of the kind of asymmetry that appears in the all-swans-are-white example, and

    (c) rejoices in claiming, whenever he1 encounters two competing hypotheses one of which he for whatever reasons prefers, some asymmetry such that the entire (or almost the entire) burden of proof is on proving the other hypothesis, and insisting that until a conclusive such proof is presented, we can take for granted that the preferred hypothesis is correct.

Jason Mitchell is a clear case of a vulgopopperian. So are many (perhaps most) of the proponents of climate denialism that I have encountered during the decade or so that I have participated in climate debate. Here I would like to discuss another example in just a little bit more detail.

In May last year, two of my local colleagues in Gothenburg, Shalom Lappin and Devdatt Dubhashi, gave a joint seminar entitled AI dangers: imagined and real2, in which they took me somewhat by surprise through spending large parts of the seminar attacking my book Here Be Dragons and Nick Bostroms's Superintelligence. They did not like our getting carried away by the (in their view) useless distraction of worrying about a future AI apocalypse in which things go bad for humanity due to the emergence of superintelligent machines whose goals and motivations are not sufficiently in line with human values and promotion of human welfare. What annoyed them was specifically the idea that the creation of superintelligent machines (machines that in terms of general intelligence, including cross-domain prediction, planning and optimization, vastly exceed current human capabilities) might happen within forseeable future such as a century or so.

It turned out that we did have some common ground as regards the possibility-in-principle of superintelligent machines. Neither Shalom or Devdatt, nor myself, believe that human intelligence comes out of some mysterious divine spark. No, the impression I got was that we agree that intelligence arises from natural (physical) processes, and also that biological evolution cannot plausibly be thought to have come anywhere near a global intelligence optimum by the emergence of the human brain, so that there do exist arrangements of matter that would produce vastly-higher-than-human intelligence.3 So the question then is not whether superintelligence is in principle possible, but rather whether it is attainable by human technology within a century or so. Let's formulate two competing hypotheses about the world:
    (H1) Achieving superintelligence is hard - not attainable (other than possibly by extreme luck) by human technological progress by the year 2100,
and
    (H2) Achieving superintelligence is relatively easy - within reach of human technological progress, if allowed to continue unhampered, by the year 2100.

There is some vagueness or lack of precision in those statements, but for the present discussion we can ignore that, and postulate that exactly one of the hypotheses is true, and that we wish to figure out which one - either out of curiosity, or for the practical reason that if (H2) is true then this is likely to have vast consequences for the future prospects for humanity and we should try to find ways to make these consequences benign.

It is not a priori obvious which of hypotheses (H1) and (H2) is more plausible than the other, and as far as burden of proof is concerned, I think the reasonable thing is to treat them symmetrically. A vulgopopperian who for one reason or another (and like Shalom Lappin and Devdatt Dubhashi) dislikes (H2) may try to emphasize the similarity with Popper's swan example: humanly attainable superintelligent AI designs are like non-white swans, and just as the way to refute the all-swans-are-white hypothesis is to exhibit a non-white swan, the way to refute (H1) here is to actually build a superintelligent AI; until then, (H1) can be taken to be correct. This was very much how Shalom and Devdatt argued at their seminar. All evidence in favor of (H2) (this kind of evidence makes up several sections in Bostrom's book, and Section 4.5 in mine) was dismissed as showing just "mere logical possibility" and thus ignored. For example, concerning the relatively concrete proposal by Ray Kurzweil (2006) of basing AI-construction on scanning the human brain and copying the computational structures to faster hardware, the mere mentioning of a potential complication in this research program was enough for Shalom and Devdatt to dismiss the whole proposal and throw it on the "mere logical possibility" heap. This is vulgopopperianism in action.

What Shalom and Devdatt seem not to have thought of is that a vulgopopperian who takes the opposite of their stance by strongly disliking (H1) may invoke a mirror image of their view of burden of proof. That other vulgopopperian (let's call him O.V.) would say that surely, if (H2) is false, then there is a reason for that, some concrete and provably uncircumventable obstacle, such as some information-theoretic bound showing that superintelligent AI cannot be found by any algorithm other than brute force search for an extremely rare needle in a haystack the size of Library of Babel. As long as no such obstacle is exhibited, we must (according to O.V.) accept (H2) as the overwhelmingly more plausible hypothesis. Has it not occured to Shalom and Devdatt that, absent their demonstration of such an obstacle, O.V. might proclaim that their arguments go no further than to demonstrate the "mere logical possibility" of (H1)? I am not defending O.V.'s view, only holding it forth as an arbitrarily and unjustifiably asymmetric view of (H1) and (H2) - it is just as arbitrarily and unjustifiably asymmetric as Shalom's and Devdatt's.

Ever since the seminar by Shalom and Devdatt in May last year, I have thought about writing something like the present blog post, but have procrastinated. Last week, however, I was involved in a Facebook discussion with Devdatt and a few others, where Devdatt expressed his vulgopopperian attitude towards the possible creation of a superintelligence so clearly and so unabashedly that it finally triggered me to roll up my sleeves and write this stuff down. The relevant part of the Facebook discussion began with a third party asking Devdatt whether he and Shalom had considered the possibility that (H2) might have a fairly small probability of being true, yet large enough so that given the values at stake (including, but perhaps not limited to, the very survival of humanity) makes it worthy of consideration. Devdatt's answer was a sarcastic "no":
    Indeed neither do we take into account the non-zero probability of a black hole appearing at CERN and destroying the world...
His attempt at analogy annoyed me, and I wrote this:
    Surely you must understand the crucial disanalogy between the Large Hadron Collider black hole issue, and the AI catastrophe issue. In the former case, there are strong arguments (see Giddings and Mangano) for why the probability of catastrophe is small. In the latter case, there is no counterpart of the Giddings-Mangano paper and related work. All there is, is people like you having an intuitive hunch that the probability is small, and babbling about "mere logical possibility".
Devdatt struck back:
    Does one have to consider every possible scenario however crazy and compute its probability before one is allowed to say other things are more pressing?
Notice that what Devdatt does here is that he rejects the idea that his arguments for (H1) need to go beyond establishing "mere logical possibility" and give us reason to believe that the probability of (H1) is large (or equivalently, that the probability of (H2) is small). Yet, a few lines down in the same discussion, he demands just such going-beyond-mere-logical-possibility from arguments for (H2):
    You have made it abundantly clear that superintelligence is a logical possibility, but this was preaching to the choir, most of us believe that anyway. But where is your evidence?
The vulgopopperian asymmetry in Devdatt's view of the (H1) vs (H2) problem could hardly be expressed more clearly.

Footnotes

1) Perhaps I should apologize for writing "he" rather then "he or she", but almost all vulgopopperians I have come across are men.

2) A paper by the same authors, with the same title and with related (although not quite identical) content recently appeared in the journal Communications of the ACM.

3) For simplicity, let us take all this for granted, although it is not really crucial for the discussion; a reader who doubts any of these statements may take their negations and include disjunctively into hypothesis (H2).

onsdag 9 juli 2014

On the value of replications: Jason Mitchell is wrong

The so-called decline effect - the puzzling and widely discussed fact that many effects reported in the scientific literature seem to diminish with time - has raised a suspicion that perhaps many of the originally reported findings are spurious. The recent surge of interest in scientific replications (which has traditionally been, and still to some extent is, viewed as a second-rate activity compared to the search for effects not previously reported in the literature) is therefore a welcome development.

Not everyone agrees. Jason Mitchell, a highly cited neuroscientist at Harvard, has recently posted an essay entitled On the emptiness of failed replications on his webpage - a piece of writing that has already received some attention elsewhere. It is a furious attack upon the practice of replication, where he claims that replications that do not find the effects reported in the original study "have no meaningful scientific value". While he doesn't explicitly state that such replication studies ought to be hidden away or buried, that nevertheless seems to be his conclusion. Mitchell's essay is poorly argued, and his main points are just wrong. Let me offer my thoughts, from the point of view of mathematical statistics. My account will involve the concept of statistical significance, and it will be assumed that the reader has at least a rudimentary familiarity with the concept.1

Mitchell consistently uses the term "failed experiment" for a study in which no statistically significant deviation from the null hypothesis is detected. I think this biased language contributes to Mitchell getting his conclusions so badly wrong. If the aim of an experiment is to shed light on whether the null hypothesis is true or not, then the unbiased attitude would be to have no particular hopes or wishes as to which is the case, and no particular hopes or wishes as to whether statistical significance is obtained or not. (This attitude is of course not always attained in practice, but it certainly is the ideal we should strive for. Calling one of the outcomes a failure and the other a success is unlikely to promote such an impartial attitude.) Fueled by his terminology, Mitchell considers only one possible explanation of lack of statistical significance in a replication study, namely that
    (1) the study was carried out incompetently.
No doubt this is sometimes the correct explanation. There are, however, at least two other possible explanations that need to be considered, namely
    (2) there is no effect (or the effect is so small as to be virtually undetectable with the present technology, sample sizes, etc), and

    (3) there is an effect, but the data just happened to come out in such a way that statistical significance was not obtained.

It is important to realize that (3) can well happen even if the experiment is carried out with impeccable competence. The opposite can also happen - perhaps in the original study - namely that there is no effect, but statistical significance is nevertheless obtained. The latter is in fact what is expected to happen about once in 20 if the null hypothesis of no effect is true, provided the traditional threshold of a p-value below 0.05 is employed in defining statistical significance. We live in a noisy and chaotic world where noise in measurements and random variation in the sample of patients (or whatever) produces random outcomes, a randomness that prevents absolute certainty in our conclusions. It is perfectly plausible to have one research group report a statistically significant effect, and another group report lack of statistical significance in a replication study of the same phenomenon, without either of the groups having succumbed to scientific incompetence. So there is no reason to say, as Mitchell does, that...
    someone who publishes a replication [without obtaining statistical significance] is, in effect, saying something like, "You found an effect. I did not. One of us is the inferior scientist."
Mitchell repeatedly falls back on this kind of emotional language. He is clearly distressed that there are other scientists out there replicating his findings, rather than simply accepting them uncritically. How dare they question his results? Well, here is a piece of news for Mitchell: All progress in science consists in showing that Nature is actually a bit different (or, in some cases, very different) from what was previously thought. If, as Mitchell seems to be suggesting, scientists should cease questioning each other's results, then scientific progress will grind to a halt.

Of course (1) can happen, and of course this is a problem. But Mitchell goes way overboard in his conclusion: "Unless direct replications are conducted by flawless experimenters, nothing interesting can be learned from them." There is never any guarantee that the experimenters are flawless, but this does not warrant dismissing them out of hand. If original studies and replication studies point in contradictory directions, then we need to work more on the issue, perhaps looking more closely at how the studies were done, or carry out further replications, in order to get a better grasp of where the truth is.

Another problem with Mitchell's "nothing interesting can be learned from them" stance is that it cuts both ways. If the fact that not all experimenters are necessarily flawless had the force that his argument suggests, then all scientific studies (original as well as replications, regardless of whether they exhibit statistical significance or not) can be dismissed by the same argument, so we can all give up doing science.

Mitchell has in fact anticipated this objection, and counters it with an utterly confused exercise in Popperian falsificationism. Popper's framework involves an asymmetry between a scientific theory (H) and its negation (Hc). In the present setting, (H) correponds to the null hypothesis of no effect, and (Hc) to the existence of a nonzero effect. Mitchell draws an analogy with the classical example where
    (H) = {all swans are white}
and
    (Hc) = {there exists at least one non-white swan}.
The crucial asymmetry in Popperian falsificationism is that it just takes one observation - a single black swan - to falsify (H), whereas no amount of observations of swans (white or otherwise) will falsify (Hc). But Mitchell's parallel between the statistically significant effect found in the original study and the discovery of a black swan succumbs to a crucial disanalogy: the discovery of a black swan is a definite falsification of (H), whereas the statistical significance may well be spurious. So when he goes on to ridicule scientists engaging in replication by comparing them to ornithologists who, faced with a black swan, try to resurrect (H) by looking for further examples of white swans, then the only one who comes across as ridiculous is Mitchell himself.

Publication bias is the phenomenon that a scientific study that exhibits statistical significance has a greater chance of being published than one that doesn't. This phenomenon causes severe difficulties for anyone trying to get a full picture of the scientific evidence on a particular issue, because if we only have access to those studies that exhibit a statistically significant effect, we will tend to overestimate the effect, or perhaps even see an effect where there is none.2 Outrageous attacks like Mitchell's upon the publication of studies that do not exhibit statistical significance will only make this problem worse.

Footnotes

1) A non-technical one-paragraph summary of statistical significance might read as follows.

Statistical significance serves to indicate that the obtained results depart to a statistically meningful extent from what one would expect under the null hypothesis of having no effect. It is usually taken as supporting a suspicion that the null hypothesis is false, i.e., that there is a non-zero effect. Given the data, the p-value is defined as how likely it would be, in case we knew that the null hypothesis were true, to obtain at least as extreme data as those we actually got. If the p-value falls below a prespecified threshold, which is often taken to be 0.05, then statistical significance is declared.

For those readers who know Swedish, my paper Statistisk signifikans och Armageddon may serve as an introduction to the concept. For others, there's an ocean of introductory textbooks in mathematical statistics that will do just as well.

2) A well-known case where this has resulted in a misleading overall picture is parapsychology. A more serious example is the selective publication of clinical studies of psychiatric drugs.

lördag 10 maj 2014

Persson om Popper

I min förrförra bloggpost retade jag läsekretsens nyfikenhet rörande min Chalmers- och matematikerkollega Ulf Perssons nya bok Karl Popper, falsifieringens profet, och jag tänkte idag berätta kort om boken. Huruvida det jag nu skriver är att betrakta som en regelrätt recension vill jag låta vara osagt. En invändning kan vara att Ulf och jag är goda vänner och att den saken kan grumla mitt omdöme eller åtminstone lägga en hämsko på min öppenhjärtighet. Jag kan inte tillbakavisa detta helt, men kan i alla fall peka på att Ulf är en person som, mer utpräglat än kanske någon annan jag känner, praktiserar och sätter värde på ärlighet och raka puckar, något jag här tänker ta fasta på.

Jag kommer att göra problemet med min eventuellt bristande saklighet och objektivitet etter värre genom att göra explicita jämförelser mellan å ena sidan Ulf och hans skriverier, och å andra sidan mig och mina. Risken jag därmed tar - förutom den uppenbara att framstå som en smula självcentrerad - ligger i hur välkänt vanskligt det är att hålla sig med en korrekt självbild. Jag gör ändå ett försök, ty om det lyckas kan det bli en bra guide för denna bloggs trogna läsare (som vid det här laget kan antas bekanta med mina skriverier) till huruvida även Ulfs bok kan vara något för dem.

Karl Popper (1902-1994) räknas som en av 1900-talets främsta och mest inflytelserika vetenskapsfilosofer. Hans centrala bidrag utgörs av falsifikationismen - idén att vetenskapliga teorier inte kan bevisas men väl falsifieras, och den därav följdriktiga betoningen på behovet av att utsätta teorierna för kritisk prövning.1 I Karl Popper, falsifieringens profet bjuder Ulf Persson på ett kort biografiskt kapitel om Popper, och ett kapitel om Poppers samhällsfilosofi (framlagd i The Open Society and Its Enemies från 1945) där vi får klart för oss hur väl denna harmonierar med hans vetenskapssyn. Huvuddelen av boken består emellertid av en räcka kapitel där vi får se det falsifikationistiska synsättet tillämpas på olika vetenskaper: matematik, fysik, evolutionsbiologi, medicin, nationalekonomi och didaktik. Popper själv kommer ibland ur sikte (även om jag inte fört räkning om saken tycks det mig som att det ibland går 15-20 sidor utan att han omtalas), men diskussionen har genomgående något popperianskt över sig, om än med en distinkt perssonsk touch. För oss som känner Ulf och tagit del av vad han tidigare skrivit känns en rad av hans käpphästar och favoritämnen igen - upptäckten av ickeeuklidiska geometrier och dess konsekvenser för synen på matematiken, ptolemaiosk kontra kopernikansk världsbild, evolution medelst naturligt urval, och inte minst didaktikens (enligt Ulf) fatala brister i fråga om vetenskaplighet. Hans intellektuella intressen överlappar i påfallande grad med mina egna.

Jag känner stor frändskap med Ulf i flera avseenden. Vi har båda vår akademiska bas och vetenskapliga legitimitet i matematiken, men vi har båda vägrat att hålla oss inom dess domäner, och vi har båda överträtt dess gränser inte i första hand genom att ställa våra matematiska färdigheter i tjänst för detaljerad tillämpning inom något specifikt ämne, utan mer genom ett mer svepande och vidlyftigt filosoferande om de riktigt stora frågorna om världens beskaffenhet. Vissa andra uppenbara likheter finns också mellan våra kynnen - kanske framför allt den snudd på oemotståndliga lusten att utan otydlighetsskapande lager av diplomati påtala intellektuella brister i andras arbeten.

Men där finns också, vågar jag påstå, tydliga skillnader. En sådan är Ulfs totala (och, tror jag, bara delvis avsiktliga) brist på den politiska korrekthet som jag tror mig själv om att i åtminstone någon mån besitta; den självklarhet med vilken han konsekvent använder det genusladdade ordet vetenskapsman är blott ett av symptomen på hans ointresse för anpassning i sådan riktning. En annan skillnad är att Ulf känner mindre behov än jag av att belägga sina allmänna påståenden med konkreta exempel eller åtminstone källvänvisningar. Möjligen kan en sådan attityd gagna flödet i texten, men stadgan i argumentationen kan också bli lidande.2 I synnerhet när Ulf avfärdar hela verksamheter och discipliner, med påståenden som att "pedagogiken inte gjort några framsteg sedan antiken" (s 165), efterlyser jag en något större konkretion i de belägg han anför.

Stilistiskt har Ulfs texter alltid lidit av en hög frekvens av slarvfel, och alltsedan den intensiva period 2005-2007 då jag som ordförande för Svenska Matematikersamfundet fungerade som ansvarig utgivare för det medlemsblad Ulf stod som redaktör för (med sig själv som flitigast anlitad skribent) har jag misstänkt att han aldrig någonsin självmant går tillbaka och korrekturläser sina texter. Ett annat lite irriterande särdrag är hans obenägenhet att påbörja nytt stycke, vilket kan få hans texter att för ögat te sig oaptitliga och ogenomträngliga. Det verkar dock som om han denna gång fått (påtvingats?) hjälp med både korrekturläsning och redigering, och även om stildragen finns kvar så är de kraftigt dämpade, vilket gör Karl Popper, falsifieringens profet till det (mig veterligen) mest välskrivna alster som kommit från Ulfs penna.

Jag finner boken klart läsvärd, och rekommenderar den varmt för den samhälls- och vetenskapsintresserade bildade allmänhet som jag föreställer mig att den typiska läsaren av denna blogg tillhör. Bokens svep över olika vetenskapsområden ur ett tämligen konsekvent popperianskt perspektiv bjuder på många intressanta lärdomar.

Trogen mitt kritiska kynne (vilket jag som sagt delar med Ulf) vill jag ändå nämna några av de enskildheter i boken som jag finner mindre lyckade:
  • Den schablonmässiga utnämningen av Richard Dawkins till "förespråkare för genetisk determinism" (s 105) synes mig orättvis och ogrundad.
  • När Ulf strax tidigare tar upp den mycket intressanta frågan om altruismens evolutionära ursprung, behandlad i E.O. Wilsons Sociobiology från 1975, gör han ett val jag finner smått obegripligt, nämligen då han vad gäller förklaringsmodeller förbigår både den släktskapsselektion som spelar så stor roll i Wilsons arbete och den tankegång om reciprok altruism som senare skulle vinna stora framsteg, till förmån för följande icke-förklaring:
      Den skenbara motsägelsen, att altruistiska personer har negativt överlevnadsvärde i förhållande till sina mer själviska fränder, kan som redan antytts lätt kringgås om man antar att altruismen beror på en kombination av flera gener. (s 104)
  • I diskussionen på s 144 och framåt om klinisk utprövning av nya behandlingsmetoder och mediciner talar Ulf om försöks- och kontrollgrupp och om vikten av att "det enda som väsentligen skiljer grupperna åt är behandlingen" (s 145), men försummar att nämna det avgörande instrument som tillåter forskarna att åstadkomma det på ett statistiskt rigoröst sätt: randomiserad allokering av patienter till försöks- och kontrollgrupp.3 På detta vis kan man ur korrelationer på någorlunda säker grund härleda kausaliteter, i motsats till i de epidemiologiska sammanhang Ulf diskuterar på s 146-147, där kontrollerad randomisering inte är tillgänglig, vilket gör kausaliteter långt mer problematiska att belägga (utöver den multipelinferensproblematik han diskuterar).
  • Bokens avslutande appendix om sannolikhetsbegreppet är direkt dåligt. Jag sympatiserar visserligen med Ulfs uttalade avsikt att påvisa att "det är meningslöst att tala om sannolikheten för en viss händelse eller ett visst faktum utan att göra preciseringar och antaganden" (s 185), men sättet han gör det på - en räcka svagt sammanhängande och slarvigt genomförda kombinatoriska räkneexempel - tjänar mest bara till att påvisa hans ytliga kännedom om ämnet.

Fotnoter

1) Här är inte platsen för någon fullödig redogörelse för min syn på den popperianska falsifikationismen, men låt mig ändå säga följande. Å ena sidan anser jag att den genom betoningen på kritisk prövning av vetenskapliga teorier varit och är mycket värdefull. Å andra sidan ser jag ofta exempel på alltför fyrkantigt anförande av falsifierbarhetskriteriet. Att detta kriterium skulle göra skiljandet av riktig vetenskap från dåliga imitationer till en rättfram sak tror jag inte ens att Popper ansåg, och att saken är komplicerad illustreras inte minst av den så kallade Duhem-Quine-tesen. Som exempel på ett fall där jag efterlyser mer nyanserad bedömning kan jag citera följande ur min recension häomsistens av Max Tegmarks bok Our Mathematical Universe:
    Jag håller med Woit och Frenkel om att [Tegmark] går långt i sina spekulationer - så långt att han nog får anses ha överträtt gränsen mellan naturvetenskap och filosofi. Jag anser dock inte att det nödvändigtvis ligger något fel i detta, så länge man är tydlig över vad som är vad. Tegmark är föredömligt tydlig med vad som är mainstreamvetenskap (som hans utförliga och mycket intressanta redogörelse för den moderna kosmologins framväxt och var den står idag), vad som är något mindre etablerat, och vad som är hans egna tämligen sjövilda spekulationer. Och jag anser att frågan om universums (eller multiversums) fundamentala beskaffenhet är intressant och viktig, och finner det troligt att vi inte kan hitta fram till sanningen om detta med mindre än att en och annan forskare på Tegmarks vis tar sig friheten att söka sig långt utanför den etablerade boxen. Jag finner det en smula inskränkt att kräva att de spekulationer dessa pionjärer därvid levererar redan från början uppfyller alla gängse vetenskaplighetskrav om poppersk falsifierbarhet och liknande (Tegmark argumeterar för sina Nivå IV-teoriers falsifierbarhet, men övertygar inte riktigt på denna punkt). Sådant kan (förhoppningsvis) mejslas fram över tid.

2) Någon gång händer det att bristen på faktakontroll leder alldeles fel. På s 163 läser vi följande:
    Under drygt femtio år har en matematiktävling anordnats i Sverige. Vid varje tillfälle tas omkring ett dussin individer ut till final: det rör sig om drygt halvtusendet finalister under det senaste halvseklet. Av dessa har endast en handfull varit flickor. Kan vi av detta sluta att pojkar är mera matematiskt begåvade än flickor?
Att frågan får lov att ställas utan att i förväg postulera att varje statistisk könsskillnad beror på diskriminerande strukturer är jag enig med Ulf om. Men att klämma till med grava överdrifter som påståendet att "endast en handfull varit flickor" är inte någon bra grund för vetenskaplig diskussion, utan tjänar endast till att underbygga existerande fördomar. Med Googles hjälp (någon samlad webbplats för matematiktävlingens resultat verkar inte finnas) gjorde jag några snabba nedslag i 00-talets resultat, och fann varje gång något eller några flicknamn i finalfältet, och i det senaste finalfältet fanns av förnamnen att döma minst sex flickor. Och denna modesta (men relativt Ulfs sifferpåstående rent revolutionerande) emancipering är inte ens något nytt fenomen: redan 1985, året då jag deltog, hade jag tre kvinnliga medtävlare i finalen.

3) Hans tal om hur "i en tillräckligt stor population bör alla dessa skillnader [mellan försöks- och kontrollgrupp] ta ut varandra" som en följd av "de stora talens lag" (s 145) kan möjligen tolkas som att han någonstans i bakhuvudet känner till randomiseringsförfarandet, men han har uppenbarligen missat vilken absolut central roll detta spelar i kliniska försök.

tisdag 6 maj 2014

Om förutsägelser, framtidsforskning och popperiansk falsifikationism på Vetenskapsfestivalen

Göteborg Vetenskapsfestival närmar sig, och liksom förra året och flera gånger tidigare kommer jag att bidra med ett föredrag, denna gång inom ramen för den senare av de båda matematiska aftnar som äger rum på torsdag (8 maj) och fredag (9 maj). Se sidan 11 i festivalprogrammet. På fredag klockan 18:45-19.30 kommer jag (i sal Euler på institutionen för matematiska vetenskaper, Chalmers) således att tala över ämnet "Vad kan vetenskapen säga om framtiden?".1,2 Jag avser att bjuda på vetenskapsteoretiska och andra insikter, varav mycket kommer att kännas igen från olika bloggposter, men inte riktigt allt, och inramningen kommer att vara ny. Jag råkade på planeringsstadiet nämna för min Chalmerskollega Ulf Persson att jag i föredraget eventuellt kunde väntas undslippa mig något ofördelaktigt om hans husgud Karl Popper.3 Troligen uppfattade han den lätt retsamma glimten i mitt öga, men för säkerhets skull valde han att försäkra sig om snabb replikmöjlighet genom att schemalägga ett eget föredrag rubricerat "Kan vi lita på vetenskapen?" omedelbart efter mitt. Het debatt är möjligen att vänta.

Fotnoter

1) Ur programmet saxar jag följande sammanfattning.
    Ett viktigt inslag i vetenskapliga metoder är att kunna testa sina teorier och resultat. Men om resultaten handlar om vad som ska ske långt in i framtiden uppstår ett problem. Exempelvis kan vi ju inte nu, år 2014, få något facit på hur väl förutsägelser om klimatet år 2100 stämmer. I detta föredrag diskuteras framtidsforskning ur statistiska, vetenskapsteoretiska och filosofiska perspektiv, med exempel hämtade både från klimatforskning och från andra brännande områden.

2) Den vetgirige gör klokt i att infinna sig redan en stund före 18:00, då Serik Sagitov inleder med att tala över ämnet "Matematisk fylogenetik".

3) Till Ulfs nya bok Karl Popper, falsifieringens profet avser jag inom kort återkomma här på bloggen.

söndag 29 juli 2012

Shut up and calculate?

I en bloggpost för snart ett halvår sedan rubricerad Köpenhamnstolkning vs mångs världar uttryckte jag intresse för ett ruskigt besvärligt område: tolkningar av kvantmekaniken. Jag inledde med att konstatera att...
    kvantmekaniken har varit spektakulärt framgångsrik i att ge matematiska förutsägelser som stämmer med observationer. Sämre är det ställt när det gäller sökandet efter en intuitivt tilltalande och begriplig tolkning av ekvationerna. Många tolkningar har föreslagits, inklusive den strikt instrumentalistiska position vilken dömer ut sökandet efter en djupare sanning bortom observationerna och ekvationerna såsom meningslös, men någon lösning som är så tillfredsställande att den förmår samla konsensus bland fysiker tycks inte föreligga
...varefter jag gick vidare med att referera vad ett (starkt personligt färgat) urval av kloka tänkare på området sagt om ett par av de populäraste tolkningarna, nämligen Köpenhamnstolkningen och många världar-tolkningen.

Diskussionen om de olika tolkningarna väckte visst intresse i bloggpostens kommentarsfält. Det var emellertid först i kommentarsfältet till den senare bloggposten Hårdförast bland bloggande matematiker som den tog riktig fart. Ett par av kommentatorerna - Emil Karlsson och Daniel L - tog tydligt ställning för den instrumentalistiska ståndpunkt som ibland lite vanvördigt sammanfattats med orden "Shut up and calculate!".1 Alltså för den position enligt vilken vi bör avstå från metafysiska spekulationer om den verklighet som ligger bakom och beskrivs av kvantmekanikens ekvationer, för att istället nöja oss med ekvationerna och konstatera att de ger förträffliga förutsägelser av våra observationer. Jag hoppas att dessa båda herrar alltjämt har rätt frekvens inrattad (och kanske rentav känner sig manade att på nytt kommentera), när jag nu skall säga några ord om varför jag finner även instrumentalismen en smula problematisk.

Liksom minst en gång tidigare vill jag här hänvisa till Jean Bricmonts och Alan Sokals läsvärda och välavvägda uppsats Defense of a Modest Scientific Realism. De diskuterar intrumentalism mer generellt än bara i frågan om kvantmekanikens tolkningar, och lyfter bland annat fram problemet med vad som menas med "observerbara" storheter. Om intrumentalismen nu bara tillåter oss att befatta oss med sådana, var skall vi då dra gränsen? Får en forskare med nedsatt syn använda glasögon och ändå tala om sådant han ser som reellt existerande objekt? Går det bra med förstoringsglas? Mikroskop? Radar? Radioteleskop? Och om jag klarar mig utan artificiella synhjälpmedel - kan jag då verkligen observera stolen framför mig? Jag har ju knappast någon direktkontakt med der Stuhl an sich, utan det är min hjärna som drar teoretiska slutsatser av förekomsten av en stol på basis bl.a. av den elektromagnetiska strålning som faller in mot mina näthinnor. Bricmont och Sokal går vidare från vardagsföremål till dinosaurier:
    The second, deeper problem with instrumentalism is that the meaning of the words used by scientists goes far beyond what is "observable". To take a simple example, should paleontologists be allowed to speak about dinosaurs? Presumably yes. But in what sense are dinosaurs "observable"? After all, everything we know about them is inferred from fossil data; only the fossils are "observed". These inferences are not, of course, arbitrary: they can be justfied by evidence from biology (that all bones were once part of organisms) and geology (concerning the processes that transform bones into fossils). The point is, simply, that fossil evidence is evidence for the existence of something other than itself: namely, the fossils of dinosaur bones are evidence for the existence (at some time in the past) of dinosaurs. And the meaning of the word "dinosaur" is not easily expressible in a language that would refer only to fossils. For example, assertions about dinosaurs’ eating habits would have to be rephrased as assertions concerning the spatial correlation of certain types of fossils with certain other types of fossils. This seems unhelpful, to put it mildly.
Men måhända säger inte pralallellen med dinosaurier allt om hur vi bör förhålla oss till materiens minsta beståndsdelar och till kvantmekaniska fenomen. Bricmont och Sokal fortsätter:
    Some instrumentalist philosophers of science are prepared to classify dinosaurs as "observable" on the grounds that, though we cannot observe them, they would have been observable to human beings had the the human species existed 100 million years ago. Now, anyone is free to define the word "observable" however he wishes; but there is no guarantee that the word, so defined, has any epistemological significance. In reality, neither dinosaurs nor electrons are ever observed directly; both are inferred from other observations, and the arguments supporting these two inferences are of comparable strength. It seems to us that, either one allows such inferences and accepts the probable reality (in some sense or other) of both dinosaurs and electrons, or else one rejects all such inferences and refuses to talk about either. To be sure, the meaning of "electron" is far murkier than that of "dinosaur": since we can form mental pictures of mid-size objects like dinosaurs, the meaning of the words referring to them is reasonably clear intuitively even if the objects are never directly observed, which is not necessarily the case for entities like electrons. That is why we are careful to assert only that electrons exist "in some sense or other", while admitting frankly our perplexity about what electrons really are.
En konsekvent tillämpad instrumentalistisk vatenskapssyn synes orimlig och ohållbar. Men kanske är den ändå motiverad, i fallet med kvantmekaniken, där alla föreslagna tolkningar framstår som så kontraintuitiva och besynnerliga - rentav som helt crazy? Kanske finns ingen "kvantmekanikens verkliga natur", eller åtminstone ingen sådan inom den mänskliga fattningsförmågan?

Den inställningen kan man ha, och möjligen skulle den rentav kunna vara rätt. Men skall vi verkligen på detta vis ge upp jakten på en djupare förståelse? Med en Popperiansk vetenskapssyn handlar vetenskapen inte bara om empiri - om att observera världen. Ett lika viktigt moment är att formulera de hypoteser och teorier som sedan kan utsättas för empirisk prövning. Innebär inte det instrumentalistiska synsättet att vi ger upp det Popperianska hypotes- och teoribyggandet, och därmed möjligheterna till fortsatta framsteg?

Med åberopandet av Popper följer kravet på hypotesers falsifierbarhet. Om inte vi på empirisk väg kan testa de olika tolkningarna av kvantmekaniken är det för Popperianen inte mycket bevänt med vetenskapligheten i dessa. Men experiment där Köpenhamnstolkningen och många världar-tolkningen ger olika förutsägelser har faktiskt föreslagits. Och även utan sådana förslag på experiment tycker jag att de olika tolkningarna kan ha ett existensberättigande, såsom teorier att bygga vidare på och så småningom komma fram till något mer förfinat som kanske hamnar inom falsifierbarhetens gränser. Jag tror hur som helst att det är på tok för tidigt att ge upp jakten på en djupare förståelse för kvantmekanikens sanna natur.

Fotnot

1) Viss förvirring råder om ursprunget till detta citat. Det har ibland tillskrivits Richard Feynman eller Paul Dirac, men det verkliga upphovsmannen är troligen David Mermin. Googling ger en mängd motstridiga besked.