A tacit assumption common in arguments concerning the scientific method, criteria of demarcation, experimental procedures, and so forth, is that scientific methods should be like effective methods for empirical problems.
In logic, an effective method is a procedure for computing the answer to a class of problems. It requires no understanding of the problem, but just symbolic transformations according to mechanical rules. An effective method always give some answer, always gives the right answer, can always be completed in a finite number of steps, and works for all instances of its problem-class (cribbed from Wikipedia). When the truth of a statement can be determined by application of an effective method, it is said to be decidable.
Many people seem to implicitly hold science to similar standards. In this view, each step in the scientific method–from discovering theories to experimental testing–can be specified by a mechanical rule. Scientists might as well be mindless automatons obeying a procedure much like computers apply an effective method to calculate solutions to mathematical queries.
In this view, once we have the right scientific method, we need only input an empirical query into science and wait for the answer. Science will always give some answer, will always give a true (or probable) answer, can always be completed in a finite number of steps, and works for all empirical problems. If science should produce a false (or improbable) answer, then it must be because someone failed to apply the method correctly. In this view, science really might be improved if we replaced humans–biases, quirks, egos, and all–with mindless automatons.
In this context, arguments about science normally revolve around the identity of its problem-class. That is, what kind of statements are decidable by scientific methods? Or what counts as an ’empirical statement’? It is, however, normally taken for granted that whatever the problem-class, the scientific method should, like an effective method, inevitably lead us to true (or probable) conclusions. If there are no methods that can satisfy these aims, then science is considered essentially irrational (or merely instrumental).
Critical rationalists disagree entirely with this way of thinking. It is pre-Darwinian. There is no “effective method” for science. There are good and bad methods, epistemic virtues and vices, but no guarantees of truth (or even probable truth). The scientific method cannot be reduced to a sequence of mechanical rules; it is an inherently creative, unpredictable, fallible, and even artistic endeavour.
Induction was supposed to be a mechanical rule for transforming observations into scientific theories. It is a legacy of this “effective method” approach to science and the growth of knowledge–the quest for an algorithm of scientific discovery. Critical rationalists, however, take their cue from evolution. The method of conjecture and refutation is analogous–and continuous with–mutation and selection, and it bears little resemblance to an effective method.
Suppose, for example, that we proposed an “effective method” for evolution with a rule specifying how to create new mutations. Such a rule must discriminate among possible mutations; it must, therefore, prevent some from ever being realised. However, this puts the cart before the horse. The adaptive fitness of a mutation is revealed by selection pressures, not before. It turns out that we do not really have rule for creating new mutations at all, but for constraining them–it’s just another selection pressure.
Scientific theories are like mutations. We do not need to specify a mechanical rule to create new theories, but merely standards of criticism (selection pressures) to subject them to: logical consistency, falsifiability, problem-solving potential, simplicity, explanatory power, and possibly others. The ecological niche we create for our theories should be one designed to weed out error and falsehood, irrationality and redundancy. Induction, concerned as it is with the origin or source of theories, serves no purpose in such a critical discourse.
Even to this day, most epistemologies and visions of science are essentially Lamarckian. They see adaptive fitness as something induced from a reliable source. The problem, in this view, is to discover the “right” source or foundation for knowledge, and to preserve its purity by shunning guesswork and conjecture. Critical rationalism, in contrast, posits a profoundly Darwinian understanding of science and the growth of knowledge, where rationality is not about justification and “effective methods” but creativity and criticism.
beautifully articulated! I want to argue with it, but I can’t–two thumbs up!
Often when I am asked for a less-than-a-compendium’s answer to the question of what Popper’s scientific epistemology is all about, I’ve simply said: it is that while there are NO SURE bets, and all still-standing theories are FOREVER fallible to one non-zero degree or other and subject to revision/abandonment as future discoveries or data may warrant, at any point in time the Last Theory Standing is the theory to bet on IF INDEED you just HAVE to cast a bet right then (but there are NO guarantees, so postpone betting whenever you can).
But to many folks my answer proves to be less than fully satisfying and not really all that clear; now, I can instead simply direct folks who ask me that to Lee’s post above!
And if that does not do the trick, I can next direct them to a longer essay by Rafe.
And if THAT does not do the trick, the trick likely cannot be done.
Thank you, Frank.
It’s nice to have a response, especially one as glowing as this. The paucity of feedback so far had me wondering if I had expressed myself poorly.
Nice work Lee, I have been distracted lately, no dramas, just busy on seveal fronts, so did not get to the post until today. Sometimes no response just means it is a good piece that can’t be subjected to criticism.
Still recognition of a neat piece of work is always appreciated by the author.
I suppose demarcation both gives a tentative means of separation of science and pseudo-science and also gives one a sort of permission to be fearless in creating conjectures about the world and our place in it. It is the quest for certainty that blurs the role of guessing versus testing. Our hypotheses can arise from the “psychological” experience of observing repetitions, dreaming, watching a movie, running a Bayesian program, casting iChing coins, going for a walk, listening to a song. Some means of generating hypotheses are extremely powerful e.g. the use of Bayesian inference to narrow down the location of U-Boats in the Second World War and computer programs that are now the staple of epidemiological studies, Google, Bing and our virtual and non-virtual life. One can make any number of potentially meaningful and potentially true assertions.
I can’t fault, Lee, how you have articulated the critical overlay that marks out “science”. I suspect that the power of machine “learning” has tended to blind researchers to the conjectural nature of knowledge as has, historically, the sheer power of our own brain and sense systems. This is understandable at a psychological level but as David Miller said in “Critical Rationalism, a Restatement and Defence” “good reasons ought to be renounced” in favour of our conjectures always being open to refutation. Miller also indicated for instance, that a good Bayesian should not be interested in whether a theory is supported or not but rather in whether it can be kept open to the standards of criticism.
When our theories survive testing this means that they have survived testing. It is a survival report. Any inference we make about the future impact of the theory is another guess. Corroboration, as John Sceski says in “Popper, Objectivity and the Growth of Knowledge”, is not logically ampliative like induction. Computers can generate guesses at an ever increasing rate, and of course perform testing at an ever increasing rate. The logical issues for generating conclusions are still intact as I believe you have outlined.
I often wondered why Popper made so much out of demarcation, once you see the point of taking evidence seriously (critically) then it is all about what the theory explains, what it does not explain, tests it has passed, tests it has not passed, problems it reveals, problems it does not solve, how it articulates with theories in other disciplines etc. A bit like a ladder that you can throw away after you use it to climb up some place.
Re-reading Realism and the Aim of Science there is a chaper on demarcation where he shows how it leads into other fundamental problems in the philosophy of science. This may help.
But up to date I looked at the demarcation as a great alternative to the verification criterion, along with rejecting the obsession with induction, so the philosophy of science could have gotten out of the rut where the positivsts and logical empiricists dug themselves in, until Kuhn and Lakatos provided a different hole in the ground (or cul de sac) to explore.
It may be of. Interest that back in the 1950s budding surveyors In London were expected surveyors to take a course in ‘jogic and scientific method’, a course which I taught in the evenins on the strength of having attended Popper and Wisdom’s course of the same title. The surveyors’ institute clearly thought that ghe scientific method was in some sense effective. It probably thought that onduction was effective!
Michael,
I’d be interested in picking your brains about some of that. So you attended courses by Popper and Wisdom and taught another on logic and the scientific method? I’m jealous. It’s often surprising to discover a little about who one is arguing with online.