A massive not-so-hidden benefit of philosophy that’s very rarely mentioned, is the nice way the practitioners write. Is this a nicer perk than, say, an unfortunate necessity to conduct one’s research in Hawaii, as Derek Bickerton is forced to do? I may be an exception, but rrameez does have the pleasant sentences and the calm classy style. In this posting I reply to the suggestion he made in a comment to my earlier posting, that we take a look at that most important of issues, lying at the heart of thinking about science, but which is often decried as not actually of much importance. In fact, some claim that Popperian philosophy of science is not even actually useful in science at all!
RRameez recommended reading Mulkay and Gilbert 1981 “Putting Philosophy to Work: Karl Popper’s Influence on Scientific Practice”, so I did.
The authors think Popperism isn’t actually used in science. They asked 34 scientists, all linked together through their work on “a specific research network in biochemistry” (investigating whether cytochrome oxidase is a proton pump), what they thought of Popperism. Most tended to reply that they thought it was correct and fashionable, but not influential. The authors say P (Popperism) doesn’t guide the researchers, but was used by some of them after the fact, to rationalise their work.
The authors complain that it’s supposed to be prescriptive, i.e., it’s supposed to tell you what to do in certain, or perhaps all, circumstances. Yet, they say, it doesn’t, since whether you’re being P (Popperian) or not, depends on whatever you currently believe about various technicalities. For example, the molecule at hand seemed to be a pump when extracted and made to operate in vitro. Any scientist who believed that was the crucial experiment would, the authors claim, have to believe the “not a pump” theory had been refuted. Yet experiments done on the molecule while it was stuck in its natural context, in vivo, haven’t yet shown any pumping; if you thought that was the crucial experiment, you would be P if you accepted it, but not if you didn’t. It could be that something changed the molecule as it was being extracted prior to the in vitro experiments, or alternatively, some part of the in vivo experiments might be degrading it so it didn’t work properly. But depending on which if any of those you believed, either the pro-pump or the anti-pump theorists would be non-P:
“In other words, assessments of conformity to Popper’s basic rule of scientific method, hinge on scientists’ interpretation of the term ‘falsification’; and the meaning of ‘falsification’ depends entirely on researchers’ technical and scientific judgements.”
Even though experimenters on both sides accused the other of being non-P, there is surely a pretty obvious purely P viewpoint on the whole situation:
We don’t know which is right (there may be further alternatives anyway), but most of all, There is not necessarily a refutation of the theory under consideration here! The result of an experiment may be a refutation, but not necessarily of your original idea!
One big problem is how to deal with those unknown influences which turn the experiment you thought you were doing, into the experiment that actually happened. Ideally, Our Own Dear Theory, about which we are most interested, would be testable on its own:
But we all know that experiments include complicating influences such as extraneous and confounding variables:
Whatever the real situation with the complicating influences, the result of an experiment which seems to refute ‘a’ theory, will (admittedly at a 95% confidence level say) actually be a refutation. But… it will be in the experiment we actually carried out, and we may well not yet know all those complications that stop it being the experiment we hoped we were carrying out. Indeed we never will know 100%. But that doesn’t mean we’ll never get a realistic grasp of the situation; it seems likely that often, after a while, the percentage we do understand is large and adequately representative. The hunt for the truth about Our Own Dear Theory progresses through a sequence of Popperian experiments, until we get the design of the experiment right (or right enough). (And of course Our Own Dear Theory itself might change a bit.)
The growth of knowledge is a universal phenomenon, and occurs while learning a skill, as well as when developing science. Very often, we have to deal with complicating influences we do not know initially, but we assume they are there. It’s a bit like dealing with unknowns in algebra. We may not know them fully at the start, or even ever, but they do not force all our experiments to be non-P. Just because we are dealing with algebraic abstractions, it doesn’t mean we stop doing arithmetic. When learning a physical skill, there are neuronal circuits that adjust, and this shows how natural and simple the process is, and well understood at least by our bodies.
To get back to Mulkay and Gilbert, and whether Our Own Dear Theory, that cytochrome oxidase is a proton pump or not, is true, both sides would have to understand that the big second diagram is the more realistic one. You have to believe it represents the actual experiments, and also that one day you will be able to name, understand and control all those extraneous influence boxes a, b, c etc, as near as dammit. But that is still Popperian. You will progress via sub-theories concerning your experiment, but every step will be P, it will just include often abstract, incompletely understood, extraneous theories/influences. But neither side would assume the initial, simpler diagram would be close enough to the reality for any refutation either way of One’s Own Dear Theory. Even though there was a refutation in the experiment they actually did.
Each new step in the sequence of experiments is formed using induction (the creation of new laws, ideas etc.). This is not the half of science that P is supposed to address. In fact, Popper himself was not only cynical about it but didn’t really understand it. So the fact that every step of the metaprocess often involved in investigating a science is not directed “prescriptively” by the P process is not a fault in the P process, whatever Popper himself might have said.
The Popperian framework is still very much involved in the Testing part of these sequences of experiments with their abstract unknowns, and all the other complications:
This is why Ж 8:2 and Ж 8:3 say:
“…Theories must swim amongst others at various levels, but it’s no good saying “It gets so complex it’s impossible”. Popper-based knowledge engineering works, even with complex uncertainties, but it’s guidelines, not pure mathematics…”
That last bit about guidelines not maths, again deals with Mulkay and Gilbert’s complaint that P doesn’t tell us what to do at every step.
When you buy a kitchen blender it doesn’t come with all the food you’ll ever need. It might not even come with a recipe book. P is a paradigm; I claim it is guidelines. It doesn’t have to be “prescriptive” of the whole process as the authors claim Popper said it would be, even if he did say it. P doesn’t tell you exactly what to do at every step, just the kind of concepts you should deal with. I don’t think any more needs to be said about that. The experimenters in the study knew what needed to be done to investigate further: they had to investigate exactly what is going on in both in vivo and in vitro experiments until they both felt sure they’ve nailed those uncertainties.
I would stress that each experiment does still follow the P pattern. A number of further benefits arise from this:
We can use the underlying P concept of a theory as a model, to draw the incredibly useful conclusion of asymmetry between proof and disproof expressed by Popper and in
“…Proof needs to demonstrate the impossibility of any competing theory, known or not, of equivalent or superior power, whereas disproof merely requires the presentation of evidence inconsistent with the one theory….”
and the whole of Ж 10:
“Evidence is those observations not well explained by a theory, or not as well explained by one theory as by another. Positive but non-comparative evidence (explained by a theory), though useful in the black art of theory generation, merely progresses testing. ‘Positive evidence’ must not be trumpeted as ‘evidence for’ a theory when competing theories also explain it. Papers and journals should ensure the full landscape and pattern of posable theories are reasonably accounted for.”
Whatever the details of the imperfectly known complications, P is still incredibly useful because for example we have not been reduced to saying “this observation agreeing with the theory is as important as that observation which disagrees”. If the adequate accuracy of experiment has been arrived at, there is a big difference. And if you hadn’t hoped before doing the experiment that it would be adequate, presumably you wouldn’t have done it. Those other social issues hinted at the end of Ж 10 are also important, and arise from the P understanding of theorisation.
Another thing that came from the study was the scientists’ view of P as a negative falsification-based description of science. P is the Test part of science which is essentially a process of Generate and Test. Science is both, but P only applies to the testing, since you can come up with your theories any old how, even dream them if you want. Appreciating that would help people see P as part of a required and sensible whole.
Also, the scientists in the study made the old mistake of seriously over-stressing the Testability requirement. Their comments suggested you had to offer a test alongside any theory you present, and that test had to be this and that and obey all sorts of requirements. An important thing to remember about a theory is that it is only untestable if there is proof that it could never leave any kind of trace or implication on the universe. Obviously, it’s incredibly hard to come up with any kind of scientific theory like that. Almost anything will leave messages that subtle statistics on the right things will highlight. We’re as ignorant of what tools might arrive in the future as people were before the invention of carbon-14 dating or geiger counters and all the other weird and wonderful tools science has come up with. And remember, those tools are easier to invent if you know what they have to do! The invalid use of that testability weapon is one of the most destructive in science. That’s why Ж 11 & 12 say:
Ж 11: “An untestable theory is one which is intrinsically logically untestable, not one for which no technique for testing it is yet known to some person, or indeed anyone. Deducing the scope of implications, effects, or influences of a hypothesis (via which it might be tested) can be slow and unending. ‘Untestable’ is a rare category, not to be routinely flung at everyone else’s new theory.”
Ж 12: “Tests a new theory can uniquely pass are best offered, and may be needed for superiority, but not for pseudo-criteria like theory status, false ‘testability’, or truth. Insisting on a mechanism for a theory is a classic error. A theory often inspires the discovery of its mechanisms and special tests.”
Popper stressed testability to help deal with psychoanalysis and historicisms like Marxism (and Nazism of course), all of which glibly accounted for every piece of evidence as being accounted for by the theory. [Note for Marxists (coughChina – not that anyone from China has ever been allowed to access my blog!): a lot of Marxist principles may be a good idea, it’s just that Marx’s portrayal of the inevitable destiny of human future history is rather dodgy, and particularly the way he defended it was too dodgy to count as science. BUT YOU CAN STILL SUPPORT POPPER AND SYMAPATHISE WITH MARX] Those brook no feedback; scientific theories rarely escape feedback on their validity, and can seldom if ever be proved to be immune to feedback-testing tools not yet invented. The whole testability thing really isn’t an issue in real science – but more importantly, must never be treated as if it is.
So I claim again that Popperism is central to science; also, that when Popper handed it to us, it wasn’t yet in it’s ideal form!
A couple more things:
The “Trial and Error and Try Again In A Different Way” aspect central to Popperism, is related to the same sort of thing in evolution, and also of course in human learning, and they all work for the same reasons.
Second, Popperism can be implemented explicitly to make particularly intelligent systems.
For more on Popperism in science, try chapters 2, 14 and 15 in “The Secret Dinobird Story“.