C&R and Deductive Science
Steve Push
-- among other steps -- create explanatory theories, deduce expected
observations from these theories, and test these expectations against
actual observations. Is C&R possible without those steps? If so, how
does it create new scientific knowledge in the absence of critical
empirical tests?
-- Steve
Elliot Temple
C&R works by guessing ideas and criticizing them. That's how we learn in general. It's the same underlying way that knowledge is created as with evolution.
Sometimes we get a little bit stuck. We have these ideas which are both good explanations. They don't contradict themselves. They aren't vague. They aren't mess of ad hoc complications. They're hard to vary. We don't see anything wrong with them. They both, if true, would solve some problem of interest. And yet they contradict each other regarding the real, empirical, physical world. So we have a *scientific* (empirically relevant) problem on our hands.
What do we do then?
We take that area where they make contradictory predictions and we find a specific *test experiment* we can do. How do we think of a test experiment, by the way? That is C&R and often no scientific testing is needed to do it. We guess some experiment designs and criticize them in terms of issues like whether our theories of interest do or don't make contradictory predictions about the result and what sources of error the experiment may have and how expensive it would be to perform.
So we'll do X for our test, and one theory predicts result Y while the other predicts result Z. So then we do it and at least one of our theories is refuted by the experiment, and so we've gotten unstuck and made some more progress.
(There are also some other possible cases. For example you could have two theories where one is basically a subset of another. Then you could do a test that might refute the more specific theory, or might refute neither. So there's no guaranteed progress from doing the test but you might try it anyway. You would however still be also testing the more specific theory against its negation.)
These steps are always possible when applicable -- when/why wouldn't they be? -- though sometimes it can be hard to think of a relevant experiment we can do with current technology, and sometimes we've had to wait for new technology to be invented in order to carry out some tests.
Where did deduction come in? It really has no notable role here. It's just that ideas have consequences. E.g. if you have the idea that "all loaves of bread are at least 5 lbs", and you have a particular loaf of bread, then you can *deduce* that (according to the idea) it's over 5 lbs. The deduction step in this case is sufficiently trivial it doesn't normally deserve attention or a fancy word. (Actually if you wanted to be picky I'm sure you could point out multiple uses of deduction through the scientific process with getting a scale, weighing it, etc)
Someone who didn't know what "deduction" was could still do this and get it right cause that part -- deducing from "all loaves of bread are at least 5 lbs" that *this one* is at least 5 lbs -- is pretty easy.
-- Elliot Temple
http://curi.us/
Steve Push
On Feb 17, 2:24 am, Elliot Temple <c...@curi.us> wrote:
> On Feb 16, 2012, at 9:29 PM, Steve Push wrote:
>
> > If I understand Popper, for C&R to work (in science, at least) we must
> > -- among other steps -- create explanatory theories, deduce expected
> > observations from these theories, and test these expectations against
> > actual observations. Is C&R possible without those steps? If so, how
> > does it create new scientific knowledge in the absence of critical
> > empirical tests?
>
> C&R works by guessing ideas and criticizing them. That's how we learn in general. It's the same underlying way that knowledge is created as with evolution.
>
> Sometimes we get a little bit stuck. We have these ideas which are both good explanations. They don't contradict themselves. They aren't vague. They aren't mess of ad hoc complications. They're hard to vary. We don't see anything wrong with them. They both, if true, would solve some problem of interest. And yet they contradict each other regarding the real, empirical, physical world. So we have a *scientific* (empirically relevant) problem on our hands.
>
> What do we do then?
>
> We take that area where they make contradictory predictions and we find a specific *test experiment* we can do. How do we think of a test experiment, by the way? That is C&R and often no scientific testing is needed to do it. We guess some experiment designs and criticize them in terms of issues like whether our theories of interest do or don't make contradictory predictions about the result and what sources of error the experiment may have and how expensive it would be to perform.
>
> So we'll do X for our test, and one theory predicts result Y while the other predicts result Z. So then we do it and at least one of our theories is refuted by the experiment, and so we've gotten unstuck and made some more progress.
>
> (There are also some other possible cases. For example you could have two theories where one is basically a subset of another. Then you could do a test that might refute the more specific theory, or might refute neither. So there's no guaranteed progress from doing the test but you might try it anyway. You would however still be also testing the more specific theory against its negation.)
>
> These steps are always possible when applicable -- when/why wouldn't they be? -- though sometimes it can be hard to think of a relevant experiment we can do with current technology, and sometimes we've had to wait for new technology to be invented in order to carry out some tests.
> Where did deduction come in? It really has no notable role here. It's just that ideas have consequences. E.g. if you have the idea that "all loaves of bread are at least 5 lbs", and you have a particular loaf of bread, then you can *deduce* that (according to the idea) it's over 5 lbs. The deduction step in this case is sufficiently trivial it doesn't normally deserve attention or a fancy word. (Actually if you wanted to be picky I'm sure you could point out multiple uses of deduction through the scientific process with getting a scale, weighing it, etc)
>
> Someone who didn't know what "deduction" was could still do this and get it right cause that part -- deducing from "all loaves of bread are at least 5 lbs" that *this one* is at least 5 lbs -- is pretty easy.
on several occasions. I have the impression that the deductive
validity of refutation was essential to his view of scientific method
and his solution to Hume’s problem of induction.
-- Steve
David Deutsch
> Popper seemed to think the deduction is non-trivial. He emphasized it
> on several occasions. I have the impression that the deductive
> validity of refutation was essential to his view of scientific method
> and his solution to Hume’s problem of induction.
It's essential to his explanation of how scientific knowledge is created that the deduction of predictions from theories is 'valid' in the sense that it is
-- Not nonsense; and
-- Actually possible to do (and for others to check that one has done), rather than just vaguely claim one has done.
But it is not 'valid' as a
-- Means of creating new scientific theories (except of course that such theories must not contain logical contradictions); nor
-- As a way of justifying any scientific idea as true or probable.
Or false: As for the 'deductive validity of refutation', that is only important in contrast to the deductive invalidity of confirmation. The 'deductive validity of refutation' is not a means of justifying the falsehood of a theory, because the refutation is never by the observation alone, but only by the observation in the light of background knowledge, which consists of many other theories and assumptions, some of which were are not (yet) explicitly aware of at any given time. Hence the only thing that *logically* follows from the refuting observation is that 'something is wrong with some of our ideas, somewhere' -- which is always true and therefore has no relevant content. Which idea or ideas are false, remains a matter not for deduction but for conjecture in the light of what explanations are available.
-- David Deutsch
Elliot Temple
I thought he emphasized deduction for reasons including:
1) it uncontroversially *is not induction*. Any step that is deduction cannot have any induction hiding there.
2) it uncontroversially works. so any step people agree is deduction they won't be attacking at all.
3) Popper's early work is overly formal in style (though still less than the average in the field) and deduction appeals to such people
4) Popper, more so early on, had some interest in logic and logical systems, and spent time working with them and studying that kind of thing, and understanding their capabilities and limits
However there is no "deductive validity of refutation" except in the following senses:
1) certain sub-steps of refutation are deductively valid
2) refutation as a whole is *not* deductively *invalid* -- it doesn't contradict deduction
But if you want to know if the conclusion of your refutation is true -- e.g. not all loaves of bread are more than 5 lbs -- that conclusion is not a matter or pure deduction and cannot be defended as deductively valid. Reaching that conclusion involved steps other than deduction and you had to get those steps right for the conclusion to be true.
Popper very much knew this and wrote about it as early as in his first well know book, The Logic of Scientific Discovery (LScD). Then he was, bizarrely, criticized for not knowing it and for making a philosophy that is unaware of this issue and doesn't work due to this issue.
The issue has been called the Duhem-Quine problem but would be better called the Duhem-Quine-Popper problem.
It's basically what DD said:
> the refutation is never by the observation alone, but only by the observation in the light of background knowledge, which consists of many other theories and assumptions, some of which were are not (yet) explicitly aware of at any given time. Hence the only thing that *logically* follows from the refuting observation is that 'something is wrong with some of our ideas, somewhere'
Or put another way: refutations are actually *contradictions* and the error could always be on either or both sides (e.g. it could be in one's measuring technique, or his understanding of what he observed, or one's understanding of how to apply a theory he's refuting, as well as it could be in the theory he's refuting).
Logically we have a symmetrical situation: some ideas contradict and at least one is false. So that is an interesting and notable problem.
Popper talks about this in LScD section 30 (and elsewhere). Some things he says include:
- some people would choose the set of ideas (the contradiction is between multiple sets of ideas) which is simplest. he's more interested in the one that has been more severely tested. (note he's focussed on scientific theories here, and it's a book about the logic of *scientific* discovery, but that doesn't mean his worldview/epistemology only deals with testable, scientific theories)
- we shouldn't/can't decide by "experiential justification" of some basic statements (observation statements) nor by logic or deduction (which isn't up to the task)
- "We choose the theory which best holds its own in competition with other theories"
- "it is *decisions* which settle the fate of theories" (this could be read, out of context, as advocating subjectivism or idealism, but is not, what it's doing is emphasizing the human role in thinking, which is necessary, and our results depend on how well we do it. it can't be avoided b/c tools like deduction can't do everything)
- "Science does not rest upon solid bedrock. THe bold structure of its theories rises, as it were, above a swamp. It is like a building erected on piles. The piles are driven down from above into the swamp but not down to any natural or 'given' base; and if we stop driving the piles deeper, it is not because we have reached firm ground. We simply stop when we are satisfied that the piles are firm enough to carry the structure, at least for the time being."
Popper also discusses the issue directly in LScD section 18.
Rafe Champion
it without adding value), I wrote a thesis on the topic, which I thought was
really boring at the time but in retrospect it provides a useful survey of
the issue and the state of play in recent times. Duhem's own response to the
problem was very interesting (and Popperian)!
http://www.the-rathouse.com/Theses/Duhem-QuineIntroPopperians.html
Right now I am preparing a reading guide on The Logic of Scientific
Discovery. The idea is to prepare a series, of reading guides. The first (in
well advanced draft) is The Poverty of Historicism (Popper's shortest book).
http://www.the-rathouse.com/Popper-papers/PH-READER.html
This is some stuff drafted for the LSD Reader.
Why The Logic of Scientific Discovery Matters
The Logic of Scientific Discovery is one of the great �game changing� books
of the 20th century. Originally published in 1935/5 as Logik de Forschung
(The Logic of Scientific Investigation) it emancipated the philosophy of
science from the program of the logical positivists. It also provided
philosophical support for several aspects of Einstein�s methods and his
revolutionary program in physics
The book advances two of the four �turns� that represent Karl Popper�s
lifetime achievement. These are the �conjectural turn� to accept the
inevitably fallible nature of our knowledge and the �social� or
�conventional� turn to take account of the �rules of the game� in scientific
investigation.
Not a good book for beginners
Readers need to be warned about some traps in reading LSD. It is a book that
was written for a time of crisis and it is absolutely essential to
understand, in outline, the state of play in philosophy and physics at the
time. This does not mean that it is necessary to come to grips with the
detailed treatment of probability theory and physics which occupies large
sections of the book. Another aspect of the book that is likely to distract
attention from the major issues on induction and demarcation is the amount
of detail devoted to the �fine print�, the language used to formulate the
arguments. This was Popper�s concession to the �linguistic turn� among the
Logical Positivists.
The essential ideas can probably be gleaned from Part I, less than 60 pages,
and the last chapter on corroboration (30 pages). (90 out of 500 pages)
Background: the philosophy of science in the early 20th century.
Popper�s contribution has to be understood against the background of the
ideas that dominated Anglo-Saxon philosophy in the early 20th century,
largely under the influence of Bertrand Russell and Ludwig Wittgenstein.
This history has largely dropped out of sight among the generations that
grew up with the science wars of the 1960s and 70s under the influence of
Kuhn, Lakatos and Feyerabend.
The philosophy of science was not institutionalised or professionalised
until the 1930s when it became an official movement driven by the logical
positivists on the Continent and later by the logical empiricists in the US.
When Popper first became interested there was only a handful of academics in
that field in the world. The issues that are now addressed by some thousands
of fulltime staff and students around the globe, were in those days the
preserve of small groups of interested people, including working scientists
such as Pierre Duhem, some of them outside the universities, like Charles
Sanders Peirce.
Such was the Vienna Circle of logical positivists who gathered around
Professor Moritz Schlick (1882-1936), Rudolph Carnap (1891-1970) and Otto
Neurath (1883-1945). Their spiritual predecessor was Ernst Mach (1838-1916)
a philosopher-physicist in the strong empiricist tradition of David Hume
whose mission was to purge science of metaphysics and place it on the firm
"positive" foundations of sensation. Few philosophers have had such a deep
and wide-ranging influence. Mach virtually became the official philosopher
of Viennese progressivism through his influence in psychology, physics (the
young Einstein), literature (Robert Musil), and painting (the
Impressionists).
The members of the Circle pursued Mach's positivism, with Russell's
Principia their inspiration and Wittgenstein's Tractatus providing the
program (both these books are probably not being read these days). This was
essentially a war on metaphysics, initially using the strict
"verificationist" definition of meaning. They proposed that statements
should be regarded as literally meaningless if they could not be confirmed
or verified by evidence. The propositions of logic and mathematics were
exempt from the requirement for verification on the understanding that they
are true by definition and they do not pretend to convey information about
the world.
Popper�s Progress
After leaving school (in 1919?) Popper went to lectures at Vienna University
as a non-matriculated student, devoted much effort to politics and social
work and became a qualified cabinet maker.
He contemplated a career in music but instead joined the teacher-training
course established in the University to support the Austrian school reforms
that were under way at the time. He qualified as a teacher in 1928 (check)
after majoring in psychology and writing two theses, one on habit formation
in children and the other on the axioms of the various schools of geometry.
On the way he studied philosophy as an autodidact, with some informal
assistance from Gomperez, Kraft, M Polanyi and some others who were attached
to the Circle. During 1929 he locked onto the twin problems of induction and
demarcation like a heat-seeking missile and followed the logical
positivists/empiricists in a lifelong mission of destruction. This
unfortunately tended to distract attention from other important work that he
did after Logik der Forschung.
A time of crisis
This was a book for a time of crisis in philosophy and physics. The crisis
in philosophy was the failure to provide a justification for the logic of
induction which was supposed to be the distinctive feature of science.
Bertrand Russell described this as the �skeleton in the cupboard� of
rationalism. Another looming disaster for the positivists was the failure of
the verification criterion to provide a workable demarcation between science
and metaphysics. In science there was the problem of understanding Einstein�s
challenge to Newton and the tension between Bohr and Einstein on the
interpretation of quantum theory.
Production of Logik der Forschung
This book is a savagely edited version of a two volume manuscript, one
volume devoted to the problem of induction, the other with the demarcation
of science. The second volume was lost and the first volume appeared in
German in �and in English in 1978 titled � The Two Fundamental Problems etc
The original manuscripts ran over a thousand pages and Logik probably only
contains about a third of the material. The English translation is boosted
to yyy pages with new notes in the text and 150 pages of new appendices.
Overview of the Contents. To be followed by a brief account of the main
arguments about induction, demarcation and the significance of the �rules of
the game� approach which is the real common feature with his social and
political philosophy.
Rafe Champion
Steve Push
On Feb 17, 11:19 am, David Deutsch <david.deut...@qubit.org> wrote:
>
> It's essential to his explanation of how scientific knowledge is created that the deduction of predictions from theories is 'valid' in the sense that it is
>
> -- Not nonsense; and
>
> -- Actually possible to do (and for others to check that one has done), rather than just vaguely claim one has done.
>
> But it is not 'valid' as a
>
> -- Means of creating new scientific theories (except of course that such theories must not contain logical contradictions); nor
>
> -- As a way of justifying any scientific idea as true or probable.
>
> Or false: As for the 'deductive validity of refutation', that is only important in contrast to the deductive invalidity of confirmation. The 'deductive validity of refutation' is not a means of justifying the falsehood of a theory, because the refutation is never by the observation alone, but only by the observation in the light of background knowledge, which consists of many other theories and assumptions, some of which were are not (yet) explicitly aware of at any given time. Hence the only thing that *logically* follows from the refuting observation is that 'something is wrong with some of our ideas, somewhere' -- which is always true and therefore has no relevant content. Which idea or ideas are false, remains a matter not for deduction but for conjecture in the light of what explanations are available.
deducing expected outcomes from explanatory theories in order to
subject them to critical testing. Perhaps the famous experiment
performed during the 1919 eclipse didn’t justify the falsehood of
Newton’s theory, but didn’t it justify the preference for Einstein’s
theory over Newton’s?
-- Steve
David Deutsch
Newton's was refuted. Not 'less preferred'.
More precisely, Newton's theory was already highly problematic before the experiment, because there was no good explanation for how it could be compatible with Maxwell's electrodynamics and with special relativity. Einstein's theory of gravity (general relativity) was already unproblematic in those ways before the experiment, which was why Einstein, and everyone who understood the theory, already expected the experiment to go as it did. The experiment was a promising way to try to discover flaws in it, because it tested this expectation against previous expectations. Had the outcome been consistent with Newton's theory, there would have been no good explanation known at all. Physicists would have started to form new conjectures to try to find one, along the general lines of:
Modify Newton's theory in a different way.
Modify Einstein's theory.
Modify some piece of background knowledge in physics -- say, assumptions about space and time, or about starlight.
Modify the theory that the experiment tested what it was believed to test.
Or some combination of the above.
Note that exactly the same would have happened if the outcome had violated *both* predictions. Einstein's theory had predicted that the sun would deflect light by twice as much as Newton's theory predicted. If the result had instead been a deflection of *four* times as much, and physicists had done the above, and succeeded, and the resulting theory had been a new theory of gravity that then became uncontroversial background knowledge, inductivists would today be retrospectively interpreting the new theory as having been induced from the observation '4'. How could it be otherwise? After all, any way of obtaining a theory that isn't deduction is induction, by definition, right?
-- David Deutsch
Steve Push
based solely on the elimination of deduction as an option. And I
doubt that physicists could have induced a new theory of gravitation.
But in the absence of a theory, might not they have induced that the
deflection measured by Eddington conforms to a law that might have
become part of the background knowledge leading to a new theory?
-- Steve
David Deutsch
> On Feb 19, 9:10 am, David Deutsch <david.deut...@qubit.org> wrote:
>> [In your view], any way of obtaining a theory that isn't deduction is induction, by definition, right?
>
> I don’t think we can conclude that a theory is obtained by induction
> based solely on the elimination of deduction as an option. And I
> doubt that physicists could have induced a new theory of gravitation.
> But in the absence of a theory, might not they have induced that the
> deflection measured by Eddington conforms to a law
A law of physics is an explanation. In the absence of a theory there would have been an absence of any good explanation or law.
They could have *guessed* that there was a new law to be found. They could have (and some did) guess this before, during and after the actual experimental results, and they could (and would have) guessed it before, during and after those hypothetical results too.
They could also have guessed (and many did in fact guess) that there was something wrong with the experiment. In fact, if I recall correctly, the consensus until recently seems to have been that Eddington's experiment *did not* actually obtain any results: its apparent results were experimental errors that conformed to the theory by luck or unconscious selection effects. But more recently, that explanation has been overturned by further analysis and argument.
> that might have
> become part of the background knowledge leading to a new theory?
In your view, whenever a theory was obtained in a way that wasn't deduction, it *might* have been induction?
-- David Deutsch
Steve Push
In my view, theories explain why certain phenomena exist. And
theories can be used to deduce expected observations. If the outcome
of an observation differs from the expectation (assuming the
observations was not a result of error, bias, or fraud), then there
must be something wrong with the theory, which constitutes the
premises of the argument.
But I believe that sometimes observations are used to predict
unobserved phenomena in the absence of a theory. (*Not* in the
absence of background knowledge; just in the absence of a theory that
explains why the observations must be as they are.) In that case
(assuming the observations are not a result of error, bias, or fraud),
the failure of a prediction only requires us to discard the
prediction; the failure does not refute the premises, which are facts.
I believe this second kind of reasoning fits the definition of
induction. But I suppose you could also call it conjecture, because
it is a kind of educated guess. Induction seems like a more precise
term, however, because there are other kinds of conjectures that don’t
fit this definition.
-- Steve
Elliot Temple
So, it's not the absence of a theory. It's the presence of many relevant and necessary theories, and their critical application. Why do you want to describe that as "the absence of a theory"?
> I believe this second kind of reasoning fits the definition of induction.
But we've been over this. Induction is about "inducing" ideas from data (with no answer to the question of which ideas to induce from the logically compatible set). Induction is not interpreting data according to one's existing ideas, finding some problem, conjecturing solutions according to one's existing explanatory knowledge, and then criticizing the conjectures until one is satisfied with a solution. What's going on here is straight out of Popper but is not accurately described in any inductivist book.
Could you cite any inductivist book which isn't substantially wrong or vague/incomplete? You've already conceded enough points that your view is incompatible with the philosophy of induction yet you still defend it.
Steve Push
observations must be as they are.*"
> > I believe this second kind of reasoning fits the definition of induction.
>
> But we've been over this. Induction is about "inducing" ideas from data (with no answer to the question of which ideas to induce from the logically compatible set). Induction is not interpreting data according to one's existing ideas, finding some problem, conjecturing solutions according to one's existing explanatory knowledge, and then criticizing the conjectures until one is satisfied with a solution. What's going on here is straight out of Popper but is not accurately described in any inductivist book.
>
> Could you cite any inductivist book which isn't substantially wrong or vague/incomplete? You've already conceded enough points that your view is incompatible with the philosophy of induction yet you still defend it.
the beginning of this discussion I have wondered whether our
differences were semantic. Perhaps one person’s induction is another
person’s conjecture.
In previous posts, I have cited three recent books by philosophers of
science that define induction and provide examples of its use in
science and everyday life (Okasha, Godfrey-Smith, and Hacking). My
views are influenced by these authors and appear to be compatible with
their views. If I’m mistaken about that, perhaps you could show me
how.
You have said that I have broken with inductivists, but you haven’t
cited specific examples, other than saying that Bacon said we should
empty our minds before observing nature. I’m not familiar enough with
Bacon’s writings to say whether that characterization is correct. I
thought he was arguing against the blind acceptance of ancient ideas
to the exclusion of observation and experiment. (If you could give me
a citation, I’d like to investigate further.) If he did make the
extreme statement you suggest, I would disagree with him. But I think
the contemporary authors I mentioned above would also disagree.
-- Steve
Elliot Temple
I wouldn't go that far. But I think you've accepted some points that contradict induction.
> From
> the beginning of this discussion I have wondered whether our
> differences were semantic. Perhaps one person’s induction is another
> person’s conjecture.
I don't think so.
> In previous posts, I have cited three recent books by philosophers of
> science that define induction and provide examples of its use in
> science and everyday life (Okasha, Godfrey-Smith, and Hacking). My
> views are influenced by these authors and appear to be compatible with
> their views. If I’m mistaken about that, perhaps you could show me
> how.
Which is the *best book* which you have read and think contains *no serious mistakes* in what it says about induction (judged by your *current* understanding)? So that if I point out *one serious mistake* you'll consider your position refuted? And if I point out instances of the book having *poor quality arguments*, you'll be quite surprised and rethink your judgment of inductivists?
If there is no such book, or if you prefer, then could you give a statement of some kind which, if I point out *one serious mistake* you'll consider your position refuted? What specifically do *you* think needs further criticism, and what kind of criticism would you consider decisive?
> You have said that I have broken with inductivists, but you haven’t
> cited specific examples, other than saying that Bacon said we should
> empty our minds before observing nature. I’m not familiar enough with
> Bacon’s writings to say whether that characterization is correct. I
> thought he was arguing against the blind acceptance of ancient ideas
> to the exclusion of observation and experiment. (If you could give me
> a citation, I’d like to investigate further.)
Popper talks about Bacon in Conjectures and Refutations, in the Introduction.
It also talks about Hume, Descartes, and various others.
-- Elliot Temple
http://fallibleideas.com/
Steve Push
Peter Godfrey-Smith (University of Chicago Press, 2003), especially
Chapter 3 (Induction and Confirmation) and Chapter 4 (Popper:
Conjecture and Refutation).
-- Steve