by Charlie Huenemann
Philosophy of science, in its early days, dedicated itself to justifying the ways of Science to Man. One might think this was a strange task to set for itself, for it is not as if in the early and middle 20th century there was widespread doubt about the validity of science. True, science had become deeply weird, with Einstein’s relativity and quantum mechanics. And true, there was irrationalism aplenty, culminating in two world wars and the invention of TV dinners. But societies around the world generally did not hold science in ill repute. If anything, technologically advanced cultures celebrated better imaginary futures through the steady march of scientific progress.
So perhaps the more accurate view is that many philosophers were swept up in the science craze along with so many others, and one way philosophers can demonstrate their excitement for something is by providing book-length justifications for it. Thus did it transpire that philosophers inclined toward logical empiricism tried to show how laws of nature were in fact based on nothing more than sense perceptions and logic — neither of which could anyone dispute. Perceptions P1, P2, … Pn, when conjoined with other perceptions and carefully indexed with respect to time, and then validly generalized into a universal proposition through some logical apparatus, lead indubitably to the conclusion that “undisturbed bodies maintain constant velocities” — you know, that sort of thing.
Alas, the justifications never quite worked. Philosophers are very clever, especially when it comes to exploiting logical loopholes with surprising counterexamples. And so were introduced, alongside the venerable problem of induction, new problems like the raven paradox, the grue problem, and other hijackings of the justifications provided for science. Rescue attempts were made, only to prompt new mutations of the initial problems. It began to look as if logic and sense perceptions may not be quite enough to establish the full rationality of science.
Now a variety of responses to this impasse are possible. First, one might simply declare that science is an irrational undertaking. But this just isn’t so. Of course, there are plenty of instances of irrationality, politicking, data dredging, lying, cheating, and stealing in the annals of science. But it cannot be seriously maintained that science is nothing more than such mischief. When we manage to do science right, it is the greatest example we have of rational activity. We have good reason to be proud of ourselves with regard to this whole science thing (an alternative slogan for the Enlightenment).
Second, one might politely ask philosophers to cease and desist in their fanatical admiration. “Look,” we might say to them, “it is very cool that you love science so much. But maybe a better way to show your love is by joining in the work. Science itself can show the nature of its own justification. Sorting out good inferences from bad is a worthy project to take on, and to do it well you should study statistics, or human psychology, or just enter into one of science’s specialized realms and start digging through the data and reassessing it. We could use your help!” In other words, one could deflate philosophical justification into a naturalistic attitude toward science — which basically consists in the assertion that there really isn’t anything wrong with science lifting itself up by its own bootstraps. (“It’s not a vicious circularity, but a virtuous circularity!” we might say to ourselves, repeating it whenever the occasional doubt creeps in.)
Or, third, we might reconsider the means philosophers were employing in their attempts to justify science. It is clear that sense perceptions and mere logic are not enough. (Seriously, why on earth did anyone think they were?) So why not build a more robust conception of “rationality”? If logic alone is not enough to rule out silly problems like “raven” and “grue”, why not recognize that there is more to being a rational person than mastering an undergraduate course in logic?
It is in the spirit of this third response that I shall now reveal Six Canons of Theoretical Rationality. I just made them up. Of course, there is a wide and deep literature on rationality, and there is a good chance that what I am about to say is ridiculous when put in the context of that rich literature. But what are Monday essays at 3QD for if not throwing out some half-baked ideas in the hopes that further discussion will bake them more thoroughly?
I’ll list the Six Canons, and then explain them a bit.
The Six Canons of Theoretical Rationality
1. Nature follows reliable, knowable patterns.
2. Of two theories otherwise equal, the simpler one is better. (Occam’s Razor)
3. The relative value of two theories can be assessed by balancing how much we value what Theory 1 can explain, and how much we value what it cannot explain, compared to how much we value what Theory 2 can explain, and how much we value what it cannot explain.
4. Other things being equal, a proposal consistent with the “home theory” is better than one not consistent with the home theory. (Theoretical conservativeness)
5. When two theories are equal in value, it is best to be agnostic as to which is better.
6. The evaluative terms in these canons — “reliable”, “knowable”, “simpler”, “value”, etc — are to be understood through one’s “home theory”.
Now for some explanations.
These canons are not necessarily rules for getting at the truth. One might follow them and end up with false beliefs. They are only meant to be rules for arriving rationally at theoretical beliefs. It is quite possible to be thoroughly rational and thoroughly wrong. See Aristotle, and (in his own way) Don Quixote.
The first two canons pretty much accomplish by fiat what early philosophers of science tried to establish by laborious toil. But wouldn’t a rational person, in virtue of being a rational person, expect the patterns observed in the past to continue into the future? Even David Hume thought so, though he could not identify any process of reasoning that justified this expectation. He just decided to call it “custom” and go about his way. Similarly, what rational person would prefer a needlessly complicated theory? Sure, it might end up being the true one; but until it made some sort of difference in our expectations, why would we want to adopt it?
The third canon is a big deal. As Thomas Kuhn taught us, there’s a lot of messiness in any scientific revolution. It is never a wholly rational affair, if by “rational” we mean selecting the theory with the greatest predictive accuracy. Egos, politics, and religion can all complicate the matter. But generally what happens is that, for a variety of reasons, we really like what one theory can do, and we aren’t worried so much about its shortcomings; and we are not any more excited about what the other theory can do, and its shortcomings are irksome. There is plenty of room for arguing here about the advantages of one theory over another, and such arguments absolutely need to happen. That, indeed, is the business of rationality. But in the end, all that a rational person can do is arrive at a judgment about which theory seems, on the whole, to present the greatest value. Again, the rational person might end up being wrong in this judgment. But their erroneous judgment will be a rational one.
The fourth and fifth canons are just matters of efficiency. It makes sense to work within existing theory to the extent one can. This means provisionally assuming its truth, and working toward bringing new results under its scope. One should be wary of results that seem to contradict the “home theory”, since the results might be a fluke, or there might be something deeper going on that is explainable by the home theory. But, for all that, it might turn out that the results are genuine and inexplicable, and one has to break with the home theory, or at least doubt its completeness. But this should be a last resort. And if there are two theories which seem roughly equal in terms of costs and benefits, it is okay to decide not to decide.
The last canon addresses what is probably the most obvious and strongest objection to the whole set. These canons freely make use of murky terms like “reliable”, “knowable”, “simpler”, and “value”. One might wish for more objective criteria which could in principle be coded into an algorithm. But I do not think rationality works this way. We are always stuck in some sort of home theory that informs our judgments about what is valuable, what is simpler, what counts as regular or familiar, what counts as strange or radical, and so on. All we can do is think through these judgments, being willing to negotiate and revise as the need arises. Rationality is not about freeing oneself from such judgments; rationality is about dealing with them like a grown-up, recognizing them for what they are and realizing that other people might disagree with them and not thereby be irrational.
I do think that these six canons go a long way toward justifying our scientific knowledge while allowing for the possibility of scientific progress, change, and revolution. Of course, the success has a lot to do with the shavings of fudge I have sprinkled liberally over them. But this is how rationality works, I believe: it gives us vague principles to follow as we try to make better judgments, and making subjective assessments along the way is unavoidable.
These canons also aim toward establishing the important truth that rational people can disagree and still have a lot to talk about. They can explore one another’s values, and the reasons they have for preferring one theory over another. They can see virtues in positions they themselves do not hold. They can disagree fundamentally, and yet understand where each other is coming from. If that isn’t what we want (and sorely need) from a theory of rationality, then I don’t know what is.