On Philosophy

September 7, 2006

Essay: Knowledge and Stasis

Filed under: Epistemology,Essays — Peter @ 12:12 am

As Jaegwon Kim first detailed, one of the criteria that a successful epistemological theory must meet is to leave most of our intuitions concerning knowledge unchanged. According to Kim, if a theory claimed that most of the knowledge was, in fact, mere belief we would be justified in discarding it [1]. Although Bishop and Trout don’t provide Kim’s specific reasons for making this claim we might guess Kim reasoned that any theory we construct is itself justified by the knowledge that we have, and thus to deny that what we think of as knowledge really would be to undermine the theory itself.

It is just this stasis requirement that gives the Gettier cases, and similar “counter examples” to the traditional approach to epistemology their methodological pull. The existence of such cases reveals that many interpretations of the traditional approach to knowledge, as true justified belief, do not leave our pre-analytic intuitions about knowledge unchanged. In fact the sheer number of cases, plus of course possible variations, indicates that such interpretations may fail in a fair percentage of cases, enough that we might consider rejecting them as a theory of knowledge by the stasis requirement. Thus, because the stasis requirement is accepted, standard analytic epistemology often proceeds by “counterexample philosophy”, in which we attempt to modify existing theories until we find a variation that no counterexample can touch [2].

Bishop and Trout see this stasis requirement as a hindrance to making real progress in epistemology, even though it, and the Gettier cases, cast doubt on some of the traditional epistemological accounts that they oppose. They argue that most advances violate our intuitions about how things work, and thus, if we applied the stasis requirement to other areas of inquiry progress, would grind to a halt. Of course neither do they recommend abandoning some degree of conservatism, simply that we should never be so conservative that we rule out the possibility of progress [3].

But does the stasis requirement really rule out progress? And is it restricted to epistemology or do we use it successfully in other investigations? We can form a more general expression of the stasis requirement as follows: when investigating a concept X our conclusion as to what is and is not X should agree with at least a majority of our pre-analytic judgments about X-ness. Now let’s see how such a requirement could be used into completely different kind of investigation, an investigation into the nature of gold. First, our investigator would accumulate many rocks that he or she thinks of as gold. They would then examine those samples closely for common properties. It is likely that most of them do share a few common features, and those that do not share those features are discarded as being not gold, but something that is simply very similar to gold. Eventually he or she comes up with an explanation of why they all have those properties; because they are made up of the same element, a conclusion that leaves most of their judgments about their initial samples unchanged. Now imagine what could have happened if our investigator wasn’t working with some version of the stasis requirement; they could very well have picked the small amount of pyrite (fool’s gold) in their initial collection of samples as “real gold”, and then defined gold in terms of the properties of what we call pyrite, violating the majority of the researcher’s intuitions as to what gold is. Without something like the stasis requirement there is no reason to claim that pyrite is less qualified to be “gold” than the substance that we currently identify as gold.

This then is the function of the stasis requirement; when beginning with a concept already in use the stasis requirement makes sure that a natural kind we decide to connect with it is the best fit for the meaning of the word as commonly understood [4]. And sometimes this process reveals that there really is nothing that our pre-analytic concept corresponds to, for example when investigations into jade showed it to be really two different kind of minerals, jadeite and nephrite. So, if this process works for investigating many of our concepts, why not knowledge? Ideally we would investigate the common features of the cases we intuitively label as knowledge, discover a natural kind based on some set common features, and then use that natural kind as our definition of knowledge. So, if we adhere to the stasis requirement, we could discover a natural kind that fits the majority of our intuitive judgments, possibly discarding some of the beliefs previously thought to be knowledge, and this certainly is something to recommend it. Of course we may claim not to be interested in connecting the subject of our investigation in any way to our use of “knowledge”, but in that case why call whatever we are investigating “knowledge”?

Even if we accept that the stasis requirement really does have a role to play in investigations into knowledge we aren’t compelled to throw out the work of Bishop and Trout. Certainly their claims about statistical prediction rules would be considered knowledge, especially considering the amount of supporting evidence. And since we accept that statistical prediction rules, properly constructed, are very reliable we should then conclude that a conclusion supported by a reliable SPR is justified by that SPR, and thus may be knowledge. And even if we accept this, SPRs in no way do away with the traditional account of justification, since at the very least we must rely on “traditional” methods to identify which factors are positively correlated with what we are trying to predict, and to identify when situations make a linear SPR impossible [5]. However, accepting SPRs as a kind justification that can meet the standards required for knowledge does not subsume all of the claims made by Bishop and Trout into the standard framework. They also claim that robustness, significance, and the cost-benefits ratio are important when considering whether a reasoning strategy is recommendable. Let us assume that there is good reason to accept their claims about the value of different strategies; even so this doesn’t invalidate the standard analytic approach to knowledge. We can defend the more traditional approach by arguing that the idea of strategies of different positive value is not part of the concept of knowledge that we are trying to capture; knowledge is simply recommendable, and thus strategies that have positive value, but not equal value, both count equally as the justification required to have knowledge. This is not to say that we should ignore the possibility that some strategies are better than others, simply that it is best described as fitting into some more advanced conception, that goes beyond the simpler idea of knowledge that the standard analytic account is attempting to capture.


1. “…it is expected to turn out that according to the criteria of justified belief we come to accept, we know, or are justified in believing pretty much what we reflectively think we know or are entitled to believe.” Kim, Jaegwon, “What is ‘Naturalized Epistemology’?”

2. “For proponents of SAE, the Gettier examples are important because they show that the JTB account can’t be right on the grounds that it does not ‘leave or epistemic situation largely unchanged.’” Bishop, Michael A and Trout J. D., Epistemology and the Psychology of Human Judgment, page10

3. “But while conservatism is fine for excellent theories, it is poison in domains where progress awaits deep and durable changes in method and outlook.” Bishop, Michael A and Trout J. D., Epistemology and the Psychology of Human Judgment, page11

4. The examples, except for one, that Bishop and Trout use as “evidence” that the stasis requirement would hinder progress are all instances where new concepts are being introduced, which is certainly not the case with knowledge. The odd example out is “simultaneity”. Physicists discovered that there was no “natural kind” corresponding to our naïve conception of simultaneity, so they decided to re-use the word to describe another natural kind of temporal relation which has some of the properties that we naïvely think of as simultaneous.

5. For example, if you are attempting to predict the emergence of a certain color when performing a titration there may be no chemical compound among those that you are mixing that correlates positively with the desired color, assuming that the desired color only appears when the chemicals are in a certain ratio. It is true that there is some formula, k*a – j*b, a and b being the amount of the two chemicals, such that the color appears where the result is close to zero, but this isn’t a valid SPR because the color does not appear for every, or even most, values less than, or greater than, zero. Certainly we could form a simple prediction rule for the appearance of the color, but this wouldn’t be one of the very linear SPRs discussed by Bishop and Trout. Additionally, it requires justification that can’t be captured by a linear SPR to determine when a linear SPR can be applied.



  1. Your pyrite example is exactly backwards. Before the existence of sophisticated tests, it might be natural to say, “Well, gold is any yellow metal*, so this rock [which is actually pyrite] must be gold.” By holding to stasis, we’ll say, “Oh, well, the atomic theory is all well and good, but we all know that any yellow rock is gold, so atomic theory doesn’t explain the ‘goldness’ of this other rock.” Similarly, on stasis, any theory of bear-ness would have to admit koala bears/panda bears and any theory of bird-ness would have to admit bats (which are called a kind of bird in the Old Testament). I think it’s better just to say, “Well, we used to call these outliers X as well, but now that we have a better definition of X, they don’t fit under it. We can continue to call them X conversationally, but for technical purposes, we should remember that they’re not ‘really’ X.”

    * Though he no doubt knew better, Kant refers to gold as defined as a yellow metal in his Prolegomena.

    Comment by Carl — September 14, 2006 @ 1:54 am

  2. So pyrite is gold? I have some gold I would like to sell you. What you are missing is that we are searching for the real natural kinds that underlie our judgements, we aren’t simply examining our intuitions about gold. Remember, conceptual analysis is BS.

    Comment by Peter — September 14, 2006 @ 2:01 am

  3. I’m not sure we understand each other.

    I just said that people used to think pyrite is gold. Now, we’ve changed the meaning of gold, and so, pyrite is not gold. I don’t believe in natural kinds, so I can’t say that our new definition must be better because it’s sent from the realm of the forms, but I do think that the elemental definition of gold has more utility for jewelry making, metallurgy, etc., so it’s the one that worth using for practical reasons. The old definition of gold wasn’t contradictory or anything, we just prefer our new one instead because of its usefulness.

    On this analogy, I think it’s quite possible for people to come up with a new, technical definition of knowledge that wipes out a lot of common uses of the word. What definition for knowledge we “should” use depends on what we’re using the word knowledge to do. A lot of the Western project seems to be aimed at removing skeptical doubt. This is kind of a fool’s errand though, since if you want to, you can doubt anything. A more promising mission is one in which we rank things by degrees of doubtableness. That I’ll go to class tomorrow at 3 is pretty likely, but there could be an emergency, so I’ll call that knowledge/doubtfulness group 3. That the sun will rise tomorrow, I pretty much know is going to happen, but I can’t prove that I know it, so I’ll say it’s group 2. Math can be group 1. A is A, can be group 0. And so forth.

    Now from my experience on this site, you seem to want to say that things in group 2 are knowledge, presuming they’re based on Bayesian inferences, statistically valid, scientifically tested, and whatnot. I think that’s fine and all. It doesn’t really matter what you label knowledge, so long as you have a clear definition for it, and you use it consistently. The downside of labeling group 2 the “knowledge” group is that then the skeptics will say, “But do you really know that X will happen in the future,” because they only want group 1 or 0 to count as knowledge, and the common folks will say, “But I do know that I’m going to do X in the future!” because they like group 3 things as well.

    So, call whatever you want knowledge, just be aware of how you’re using it.

    Comment by Carl — September 14, 2006 @ 2:37 am

  4. Ok, well I think you are kinda crazy for not “believing” in natural kinds, since a natural kind is just a common feature in the physical world. For example gold atoms are a natural kind since they all have the name number of protons. We defend / make language useful by arguing that our words describe real things in the world. Since we are naturalists the only real categories out there for language to describe are natural kinds. Therefore if the world gold describes anything it must be a natural kind. Of course gold + pyrite is a natural kind, as is pyrite by itself. But we have to pick one kind for the meaning of gold to line up with if it is supposed to be meaningful. The stasis requirement helps us pick the natural kind that is the best match. As for knowledge, there are a lot of things we describe as knowledge. We investigate. We pick the natural kind that best captures our naive concept of knowledge. We call that natural kind knowledge. There may be better reasoning practices, and they are definitely worth investigating, but they aren’t “knowledge” as they aren’t the kinds of things that our use of the word knowledge describes. As for my investigations into knowledge, well they are totally irrelevant to this foundational issue.

    Comment by Peter — September 14, 2006 @ 2:59 am

  5. Or you could define gold as something with the same number of protons AND neutrons. Or you could define gold as something with the same hardness index. Or you could define gold as something with the same melting point. Or you could define gold as something with the same spectral lines. Or you could define gold as something with the same price on the stock exchange. When I say I don’t believe in natural kinds, I mean that categories are humanly selected, because there are infinite number of ways of categorizing the universe.

    Anyhow, I think we’ve hit a wall here, so I’ll leave it at that.

    Comment by Carl — September 14, 2006 @ 3:25 am

  6. But none of those defintions captures how we use the word gold, and thus are bad. Sure the word “gold” doesn’t magically get pointed to the substance gold, but our use of the word gold means that it is directed at something. And the only thing that something can be said to be is a natural kind. Yes in theory we could re-define words to mean whatever we pleased, but that would make communication impossible, not to mention lead to a lot of bad philosophy (see: postmodernism). By seeing what natural kind underlies the use of words we are improving communication and making language more useful by ensuring that our use of the word is as consistant as possible and points to something that actually exists, without abandoning the use or meaning of the word to as great of an extent is possible, since to abandon the use of the word and to re-define it as something else would be confusing for no good reason.

    Comment by Peter — September 14, 2006 @ 3:31 am

RSS feed for comments on this post.

Create a free website or blog at WordPress.com.

%d bloggers like this: