Critical rationalism and its failure to withstand critical scrutiny

Abstract

Observations or experiments can be accepted as supporting a theory (or a hypothesis, or a scientific assertion) only if these observations or experiments are severe tests of the theoryor in other words, only if they result from serious attempts to refute the theory, and especially from trying to find faults where these might be expected in the light of all our knowledge. (Popper, 1994, p. 89) The lack of progress in the neo-Popperian philosophy known as critical rationalism may be traced to its inability to show the acceptability of the fundamental principle underlying the above quote: Severity Principle (SP) Data x count as evidence in support of a hypothesis or claim H, only if x constitute severe tests of Honly if data x (which are in accord with H) result from serious attempts to refute H. This failure seems deeply puzzling, given the intuitive plausibility of SP, as in Poppers exhortation above. The problem is hardly limited to critical rationalists. Something like SP is endorsed far more generally in philosophy as well as in science, and yet it has been notoriously difficult to actually cash out what surviving serious criticism demands, and why Hs surviving the ordeal is good evidence for H. My focus here is on critical rationalists, and in particular on Alan Musgraves recent (1999) attempt. What gives SP its plausible-sounding ring is the supposition that Hs surviving serious criticism is being used in the way it is ordinarily meant: roughly, that H has been put to a scrutiny that would have (or would almost certainly have) uncovered the falsity of (or errors in) H, and yet H emerged unscathed, i.e., that H has survived a highly reliable probe of the ways in which H might be false. However, critical rationalists, as they freely admit, do not have resources to articulate anything like reliable error probes, and even deny the reliability of the method they espouse. Despite exhortations as in the epigraph from Popper, critical rationalists only espouse a weaker, comparativist principle CR: (CR) It is reasonable to adopt or believe a claim or theory P which best survives serious criticism. But without being able to say that surviving the critical rationalists actually affords evidence for P, a best surviving claim may still have been very poorly probed, and thus P may be best tested with x, even though x actually provides scant evidence for P at all. So, while we may (and most of us do) accept the intuitive principle that CR is supposed to capture (namely the severity principle SP), we have yet to be given grounds to accept CR as instantiating the intended severity requirement. To simply declare CR is a reasonable epistemic principle without giving evidence that following it advances any epistemic goals is entirely unsatisfactory, and decidedly un-Popperian in spirit. So it does not help for the Popperian to insist there is no more rational procedure than to prefer a hypothesis that is well-corroborated, i.e., that has withstood serious or severe criticism (Popper 1962, p. 51), without demonstrating the existence of testing methods that are actually severe. Yet, far from demonstrating the existence of severe error probes (or whatever one wishes to call them), the critical rationalist feels bound to deny that tests that are severe in the critical rationalists sense are reliable tools for uncovering errors. The critical rationalist is thus guilty of a kind of bait and switch, getting our nod for plausible sounding exhortations, as in SP, but then serving up, not the robustly severe tests we thought we were getting, but tests incapable of doing their intended job. Granted, Popper invites this problem, due in part to his efforts to distinguish himself from the inductivists of the time. The deductive resources to which Popper limited himself allows neither substantiating a claim to actually have a severe test or error probe, nor to say that the probability of Ps passing test T is low, given P is false. Now that we know so much more about conducting severe testing in experimental practice than was evident through logical-empiricist blinders, one would have expected this weakness to be remedied by Poppers critical rationalist followers. Surprisingly, it has not been. Even so astute a thinker and upholder of common sense as Alan Musgrave (1999) has recently mounted a defence of CR that he openly concedes is circular, admitting, as he does, that such circular defences could likewise be used to argue for principles he himself regards as crazy. (Why is CR a good rule? Because it is a good rule.) As if this were not bad enough, it turns out we cannot even self-referentially apply rule CR, i.e., we cannot show that CR itself is a best tested rule, because it is demonstrably unreliable (and other methods are not). While Musgraves full argument is subtle and clever, these concessions, or so I shall argue, radically undermine his goal, as they render his argument no argument at all. I will expose the series of missteps that have landed the critical rationalist in this untenable position. In Part I, I will argue that the critical rationalist arguments, as urged by Musgrave, themselves rest on the ability to distinguish severe from insevere tests and reliable from unreliable error probes, and thus are self-contradictory when denying the possibility of doing any such thing. In Part II, I will show how to rectify the situation by (a) rejecting the erroneous conceptions of inductive or evidence-transcending inference upon which their sceptical slide is based, and (b) showing how to develop an account with the resources to define and apply severe or reliable error probes. The main points for which I will be arguing are these: 1. An adequate defence of CR must characterize withstanding severe or serious scrutiny, and show it corresponds to classifying claims reliably, which neither Popper nor current day critical rationalists have done. 2. Musgraves argument that all epistemic principles can only be defended circularly, if they are defended at all, is unsound, and confuses self-subsuming methods with self-sealing (circular) methods. 3. In distinguishing crazy and non-crazy methods, Musgrave must assume a reliable classification scheme, which, if drawn out, already goes several steps further than what is alleged by the critical rationalist. 4. Critical rationalists assume falsely that justifying claim P requires either showing it to be true or probable. 5. A satisfactory articulation of withstanding a severe test can achieve the intended goals, without illicit justificationist or metaphysical inductive appeals. © 2006 Springer.