When Mystery's All That's Left

Our collaborator on this blog, the esteemed Mr.Wolfus Andronicus, has launched a compelling attack on many of the central claims of Jim Brown’s Platonic account of thought experiments. I am very sympathetic to most of the concerns raised in his paper. In particular, I share his worry as to how Brown proposes to arbitrate between (apparently) Platonic TE’s which disagree. I want, however, to draw attention to a point which I think is certainly challengeable. This is the claim that for an explanation to be acceptable, the explanans cannot be a bigger mystery than the explanandum. Mr. Andronicus continues, that by which we explain some phenomenon must be “comparatively more acceptable, palatable, understood, uncontroversial…” (pg.15). This is apparently put forward as a universal regulatory principle by which we can eliminate faulty explanations. It seems to me, however, that it is going to toss out many a good baby with the bathwater. There are extremely fruitful explanations, especially in science, which seem to violate this principle. It seems to me that scientists often work something like in the following way. They begin with some observable phenomenon (the explanandum) which is itself wholly uncontroversial. Then various explanations for this phenomenon are considered. Controversy ensues over which is the correct explanation. More often than not, a consensus is eventually reached. The move here is from an uncontroversial explanandum to an (at least initially) highly controversial explanans. Perhaps it is the terms 'controversial' and 'palatable' that are throwing me here(if so, I'm not sure how crucial Mr. Andronicus takes this point to be.)So what I am suggesting is that sometimes a controversial, initially unpalatable explanans turns out to in fact be the best or only way to make sense of the (uncontroversial) phenomenon to be explained. Likewise, it could in principle turn out that the epistemic power of (platonic) thought experiments could only be accounted for by appeal to some heretofore unknown faculty of platonic perception. The fact that such perception is more mysterious than what it is supposed to explain is thus in my view not enough to rule out the possibility that it is explanatorily indispensable. If it were indispensible in this fashion, it might still not provide a satisfying or unmysterious explanation. But it would be the best we could do. Of course, the mystery of Platonic perception may still provide good reasons to prefer another explanation if such is available.It suggests to me that the Platonic account should be a last resort to turn out to when all other attempts to explain what TE's do fail.

A KANTIAN ACCOUNT?

Since I cannot find a satisfactory Kantian approach (by the way, Buzzoni’s article seems to be a terrible translation from the Italian?!), I thought I would attempt to reconstruct what I think a Kantian account of Thought Experiments (TE’s) would look like. I will not defend this account here, though I am interested in what you all take to be the strengths and weaknesses of such an approach. For my purpose here, I am bracketing the issue of whether or not TE’s are really arguments. Rather, I will try to make sense of Brown’s Platonic Perception (PP).

Kantian Modalities

TE’s are different from real experiments in the sense that they deal with what “might be” the case (possibility), or sometimes what “must be” the case (necessity), but never what “is” the case (though TE’s may very well correspond, in some way, to what is the case). Take, for example, Galileo’s TE, which purports to show the inconsistency in Aristotle’s account of falling bodies. Assuming that the speed of falling bodies in a given medium is proportionate to their weights, a stone half the size of a larger stone will fall at half the rate of the larger stone. Given this, if the smaller stone is attached to the larger, we might expect the smaller, slower stone to retard the larger, faster stone. So, if the two stones are attached, their composite will fall at a slower rate than larger stone alone. Nevertheless, if the speed of falling bodies is proportionate to their weights, then we might expect that the composite will fall faster than the larger stone by itself, since the composite of both the larger and smaller stone is undoubtedly heavier than the the larger stone alone.

Evidently, the composite cannot be both faster and slower than the larger stone alone. Aristotle’s view — that the speed of falling bodies in a given medium is proportionate to their weights — leads to a contradiction, and therefore must be rejected. This discovery corresponds the Kantian modality of necessity or apodictic knowledge, i.e. what must be the case. If we accept the assumptions in Galileo’s TE, we must conclude that Aristotle’s account of falling bodies is mistaken — for how can one and the same thing be faster and not-faster in the same instance?

The first part of Galileo’s TE appears to be uncontroversial (at least when we allow for the assumptions he employs). For Brown, a TE which is able to expose a contradiction in another theory is a paradigm case of a knock-down or “destructive” TE.

The problematic portion of Galileo’s TE, as John Norton and Rachel Cooper both point out, is what Brown calls the “constructive” aspect of the TE. In the second part of the TE, Galileo goes on to draw the conclusion that all masses fall at the same rate. This is the aspect of Galileo’s TE that Brown takes to be an instance of PP. From a Kantian standpoint, this is a clear case of dogmatism. So-called Platonic objects (or forms) are not only unintelligible, but our supposed relation to these objects is also unintelligible. Indeed, for Kant, the status of our relation to any purportedly real object is precarious at best; and, as it stands, there is an absence of any clear criteria for determining whether or not (or how) we have accessed or perceived the Platonic realm. Brown defends PP by saying that our relation to empirical objects is no less mysterious than our relation to the Platonic realm. Kant would agree with this insight — however, this only furthers the Kantian intuition that we can only to talk in terms of appearances.

Aside from the fact the Brown gives us no tools for recognizing PP, he is equally unclear about what exactly PP is. I will say more about this in the next section. For now, the important point is that Galileo need not have reached the conclusion that all bodies fall at the same rate, nor does the conclusion follow naturally for every reader who considers the TE. As Cooper puts it: “Showing that heavy bodies do not fall faster than light bodies is consistent with a multitude of alternative theories, such that red balls fall faster than balls of other colours, that square objects fall faster and so on. Brown thinks that Galileo’s success in picking the right theory can only be explained by his Platonic account. For Brown the thought experiment enables Galileo to perceive the Platonic laws that govern the movement of masses and so see that all masses fall at the same rate.” Norton also exposes the various tacit assumptions that Galileo employs in his TE.

So, we might say that the constructive aspect of Galileo’s TE corresponds to the Kantian modality of possibility. It is possible that all bodies fall at the same rate. Indeed, based on Galileo’s background assumptions, this appears to be the best explanation. At any rate, it is not the only possible explanation. Because Galileo employs certain background assumptions (e.g., that colour, shape, chemical composition, and so on, have no effect on the rate at which a mass falls), the only consistent options left are: (1) that heavy masses fall more quickly than light masses; (2) that light bodies fall more quickly than heavy bodies; or (3) that all masses fall at the same rate. Since Galileo’s reductio showed that (1) and, mutatis mutandis, (2) cannot be the case, (3) is the only option left (again, given the background assumptions). Whether or not Galileo’s TE generated this insight (which may or may not be a contingent matter of fact) does not seem to me all that important. The important point is that — on a Kantian reading at least — the constructive aspect of the TE appears to correspond only to physical or hypothetical necessity (possibility). After all, there is is no contradiction involved in saying that ‘all masses do not fall at the same rate.’ Nor does this judgment appear to be a necessary condition of human experience.

De Dicto and De Re Necessity

In the Kantian framework, there are certain first principles of physics which are absolutely or metaphysically necessary, e.g., the principle of the conservation of matter, Newton’s law of inertia, or the law of the equality of action and reaction. Kant terms these principles synthetic a priori judgments of physics. Of course, these judgments do not fit Kant’s definition of analyticity, and are therefore synthetic. They are also a priori inasmuch as they constitute the form of experience, as opposed to the content of experience. For Kant, these laws are metaphysically necessary insofar as they make experience of a certain kind possible.

Determining which laws are metaphysically necessary or only physically or hypothetically necessary is doubtless a difficult task. Rather than treading through these muddy waters, I will say why both cases are ultimately instances of logical possibility for Kant; and why Brown’s PP reveals nothing more than what logically possible.

It is clear that Kant would consider Galileo’s law that ‘all masses fall at the same rate’ as an a posteriori judgment expressing hypothetical or physical necessity. As Norton points out, there is no mysterious Platonic leap here. Our degree of belief in the conclusion of Galileo’s TE corresponds to our degree of belief in the assumption that the degree of fall of bodies depends only on their weights (which, of course, is not true in instances where the shape of the objects differ). The tacit assumptions Galileo uses are based on experience. Thus, in Kant’s framework, Galileo’s conclusion is synthetic and a posteriori (at most, physically necessary or contingently true).

Whether or not we think that Galileo’s law is physically or metaphysically necessary, it is clear that it is not necessary in a strong sense (i.e. logically necessary or necessary in all possible worlds). Brown treats PP of natural laws as analogous to PP of mathematical truths. But even if we assume that there are mathematical entities existing in some realm outside space and time (i.e. Platonic objects), these entities are in no way analogous to natural laws — which presumably exist within the bounds of space and time.

Thus Brown’s PP, from a Kantian standpoint, is no more than the perception of physically or metaphysically necessary, spatio-temporal laws of nature. Since space and time, for Kant, are transcendentally ideal and empirically real, though not (necessarily) transcendentally real, it follows that the physical laws of nature also lack the necessity that Brown is looking for. While mathematical objects may be Platonic entities insofar as they may obtain in all possible worlds, it is unclear how mathematical entities are analogous to natural laws, since the former may be considered to be non-spatial and non-temporal, while the latter depend on a spatio-temporal framework.

Anyway, I have a lot more to say about how TE’s are successful/unsuccessful; whether or not they are arguments; and whether or not they are equivalent to real experiments. Since this is getting long, I will leave you with my rushed Kantian account of TE’s. The most important point here is that the supposed PP in the Galileo TE is nothing more than the perception of physical necessity based on empirical assumptions. This is in accordance with Cooper's and Norton's accounts. Although, we might ask whether TE’s are able to reveal the forms of experience — i.e. laws which are metaphysically necessary?

Some thoughts on Cooper, Peijnenburg & Atkinson, Nercessian, Brown, and Norton

I've decided to focus my attention on the papers by Cooper (CP), Peijnenburg/Atkinson (PA) and Nercessian (NS) -- along the way, I discuss Brown a little a bit, and spend most of my time defending Norton (NT) against challenges. With some notable exceptions, I'm essentially in agreement with Cohnitz and NT (especially the latter), so I don't discuss them directly here. And as for Buzzoni, well, virtually all of my comments take the form: "What the hell does he mean by that?", so I left his paper out too. (Of course, I'd be happy to hear what you guys thought about it -- hopefully you'll be better able to explain to me what he's up to!)
**Sorry for any typos -- I'm kinda pressed for time today**


CP's "Thought Experiments"


1. CP claims that it isn't clear what NT means by 'argument'. But isn't this painfully obvious? Not only does NT explicitly specify the types of arguments his account allows (i.e. deductive, inductive, abductive/inference-to-the-best-explanation, etc.), but he offers numerous examples of reconstructions of TEs.
2. CP (approvingly, it seems) cites Bishop (1999) as arguing that "[NT]'s account cannot account for cases where people disagree about the results of a [TE]. In such cases the parties reconstruct the [TE] as two different arguments, but they are discussing the same [TE]." This looks like a version of an objection that several philosophers have hurled NT's way (see, e.g., PA and NS for similar worries). In any case, it doesn't seem to carry any weight. Why can't the parties also present arguments for their argument-reconstructions of a TE? Indeed, this is the kind of thing we do in philosophy ad nauseum; it's where we spend most of our time, especially with regard to history of philosophy. It takes a lot of work to defend a particular interpretation of someone's reasoning -- usually, one will have to defend one's interpretation against heavy criticism. But NT's account has no problem with this, since it's still within the epistemically friendly realm of rational argument. Finally, it's worth pointing out that NT never claims that his reconstructions are decisively accurate -- in fact, by arguing against one of NT's reconstructions, we're engaging in precisely the sort of epistemically safe activity NT recommends vis-a-vis TEs.
3. The next failed objection to NT also shows up (in a different guise) in NS's discussion, but I'll briefly mention CP's version of it here. (I spend much more time on NS's version.) CP claims that NT's account can't be right because while TEs are "fun and easy", arguments are (presumably) boring and difficult. That claim itself is disputable, but the objection itself seems beside the point. It's really no more persuasive than the following argument. Let X be "P-->Q. P. So: Q." and Y be "Look dude, if you eat 50 cheeseburgers right now, you'll have a heart attack. Oh my god! You just ate 50 cheeseburgers! You're gonna have a heart attack!" X and Y are both deductive arguments with the same logical structure (i.e. modus ponens). Surely, some (many) people will find the one about cheeseburgers and heart attacks way more interesting and easy to follow (especially if they aren't trained in logic, or if they don't like dealing with abstract structure etc.) than the barebones-Ps-and-Qs. So there's a plausible phenomenological difference here, but it does nothing to cast doubt on the claim that they're both deductive arguments with the same logical form. And if that make sense, why is it any less compelling in response to CP's objection to NT?
4. CP claims that Hume's Missing Shade case is a counterexample to NT's account (i.e. a TE that can't be captured in propositional-argument form). For starters, I don't see how Hume's case can't be reconstructed as an argument. (Why else would we take it to show anything?) Also, BR might characterize it as a psycho-linguistic experiment; and if BR's right, then it's not a TE, in which case it doesn't threaten NT's account. It's also possible that Hume's case is an TE-qua-argument which is best carried out first-personally as a quasi-psycho-linguistic experiment.
5. CP's discussion of BR's account made me revisit my earlier worries about knowledge, perception and causation. Originally, I granted that while knowledge might not require physical causation, perception clearly does. But now I'm doubting the knowledge-side too. After all, unless we're substance dualists, knowledge must get realized at the level of the brain. If so, then all knowledge must involve some kind of causal mediation -- in particular, physical causal mediation. I'm still fuzzy on the details, but I suspect a deep problem for BR's account.
6. CP claims that "in answering the 'what if' questions [of TEs] we predict how imaginary entities will behave in the same way that we predict how real entities will behave", and that "the basic forms of reasoning used to manipulate the model will be the same as those we use to predict occurrences in the real world." This is a defining feature of CP's account, and it seems plausible, but predictions in the 'real world' take the form of inductive inferences. NT's account clearly allows for inductive inferences, so how does CP's account threaten NT's?
7. According to CP, "If the thought experimenter manages to construct an internally consistent model, and thus construct a possible world, then she can conclude that the situation she has imagined is possible." (emphasis added) For starters, this clearly has the dialectical shape of argument, so again, I'm having trouble ascertaining any threat to NT's account. I guess it comes down to what CP takes to be the scope of TEs -- that is, adopting BR's useful terminology, whether TEs are broad or narrow on CP's account. Suppose they're broad in BR's sense. Then any given TE includes not just the construction of an internally consistent model, but also the preliminary inference that the constructed model (therefore) represents a possible world, and followed by the inference that the imagined situation is (therefore) possible. If that's not an argument, I don't know what is.
8. Contra BR, CP claims that TEs showing that a "situation is possible... can indirectly teach us about the world", by showing us "how the world cannot be... and how the world must be." I think she's spot-on here, but why such TEs only do so indirectly? Think of Putnam's Super-Spartans TE against Extreme Philosophical Behaviorism (EPB). According to EPB, pain (e.g.) is nothing but a series of publicly observable behaviors, such as winces and moans. But Super-Spartans, while they still feel pain (e.g. it hurts when you stab them in leg), have trained themselves to suppress all publicly observable behaviors associated with pain-response. If Putnam's TE works, doesn't it directly show that (roughly) there's more to pain than just pain-responses? Meh, this is probably just quibbling...
9. Finally, I'm confused about something... on CP's account, all the magic takes place in constructing internally consistent models, viz., representations of possible worlds or situations. But she claims that some of the best TEs depict physically impossible situations. What's going on here?


PA's "When are Thought Experiments Good Ones?"


1. While I'm in no position to properly discuss the so-called 'definitional' issue, PA's paper got me wondering about a few things. PA, like many others in the literature, don't see any pressing need to give a clear definition of TEs; moreover, they claim that we don't even need a definition in order say "when [TEs] are good [or bad] ones" -- we don't need to know what TEs are in order to know what TEs should do. This just looks bonkers to me. If we don't have a clear idea of what TEs are, then how could we be in any position to justifiably say anything about what they should do? Indeed, aren't these features somewhat interdependent? E.g. part of the definition of a TE should either include or entail what TEs should do, i.e. when they're successful, when they're failures, etc. Of course, it need not be an exact, perfectly crisp definition. 'Either we have an exact, perfectly crisp definition of X or we have no definition of X' is a horrendous false dichotomy. It seems obvious to me that we need at least a substantially clear idea of what something is (even if only via stipulation!) before we can say anything about what counts as a good instance of it, or what role(s) or function(s) it has, and so forth. Anyway, just a thought...
2. PA claim that "a [TE] can only be deemed successful if it induces the same -- true of false -- belief in the majority of people that are exposed to it." On the face of it, this looks like a (fallacious) appeal to popularity: 'Most people believe X after they're exposed to my TE, so it's gotta be good!' How many people need to share this belief? How close must their degree of belief (or credence) correspond? Which people matter? Anyone, or experts in the relevant domain (and if so, how much expertise?), or just fairly intelligent non-experts? (There's a good reason why, e.g., after being exposed to the Shrodinger's Cat TE, a physicist shouldn't put much stock in my snap judgement that it shows X -- namely, I'm not a physicist, which means (among other things) that I'm in no position to make informed judgements about problems in physics.) Also, when people don't share the belief, are they unconvinced by the result (which may be due to strongly held or even dogmatic antecedent convictions), the setup (which may be due to misunderstanding), etc.? Also, how can we reliably determine that underlying cognitive errors aren't at work? (e.g. the 'selection effect', whereby strong beliefs tend to generate their own confirmation, 'confirmation bias', 'cognitive dissonance', etc.) Anyway, this looks very sketchy. While it would nice for TEs to generate collective, near-unanimous conviction, we shouldn't make it a test. 6 billion people could be induced to share the same belief -- and, let's suppose, to exactly the same degree -- on the basis of a bad TE. [Note: In a similar vein, PA claim that good TEs induce a kind of flash of insight to people who encounter them. I agree that it's surely nice if a TE can do that, but is this essential to all good TEs, or is it just happy psychological by-product? (Cf. BR's Mediative TEs.)]
3. PA claim that in cases where two or more philosophers disagree on the outcome of a (philosophical) TE, "there is no way to tell who is right" At best, their disagreement can only pushed down "to another level". This looks unwarranted -- the fact that, presently at least, there seems to be no way adjudicate among competing views regarding a TE does not show that there's no way to do so (much less in principle). Also, to "shunt disagreement to another level" is actually a good thing: it might help us to get clearer as to the source of the disagreement, or give us a better understanding of what's at stake, or even move the debate into more tractable waters. Moreover, why can't philosophers critically evaluate each other's "starting points" (i.e. the "level" that, according to PA, disagreement over the TE gets 'shunted' to)? There are surely cases in which a philosopher disagrees with the outcome of some TE on the basis of an untenable -- or at least highly dubious -- starting point (e.g. Cartesian substance dualism, verificationism, etc.) -- there's nothing dialectically weird about that.
4. The fact that people have intuitions about a TE doesn't imply that it generates "contradictory conclusions". Also, as I mentioned above, disagreement about the status of the TE can be traced to a plethora of factors. In any case, one can and should rationally defend one's intuitions about a TE, and there's a world of difference between rational disagreement about a TE's status, one the one hand, and whether a TE generates "contradictory conclusions", on the other. So PA's first test of whether a TE is a bad one is strange and misleading at best. (As I'll show below, the second test fares even worse.)
5. If PA are right, then philosophy itself (viz., as an area of inquiry), rather than just TEs in philosophy, looks deeply suspect -- or as PA put it, "... doomed to go round on a merry-go-round". After all, philosophers often fundamentally disagree about "starting points", and they rarely reach any clear agreement after attempts to convince one another. Are PA aware of this extreme metaphilosophical skepticism that lurks within their account?
6. The second test for whether a TE is a bad one is whether its "conclusion begs the question". Did you catch that?---whether the conclusion begs the question! That doesn't even make sense. Last time I checked, ONLY PREMISES CAN BEG THE QUESTION. Absolutely ridiculous. Second test = FAIL. :) And that means that PA's account is essentially bunk, since their two tests are untenable.
7. PA claim that one problem with philosophical TEs is that -- unlike many TEs in science (see the EPR stuff) -- "it is unclear how we ever could put [them] to the [empirical] test... ethical considerations aside." That's a total red herring. It's no better than someone who questions the epistemic legitimacy of induction on the ground that an inductive argument can't logically guarantee (i.e. deductively entail) its conclusion. Ugh.
8. Finally, I know I said I wouldn't talk about Cohnitz, but I just wanna note and expand on an excellent point he raises. (Strangely, it's only a footnote!) Basically, on PA's account, it seems we can never know whether any given TE (in science) is bad, since there might be some experiment down the line to vindicate it. (Again, see the stuff on EPR for an example.) That's bad enough, but it seems to apply to TEs in philosophy too. The history of philosophy is replete with examples of once-thought-philosophical-problems-that-became-scientific-problems. So, e.g., while the Chinese seems like a TE in philosophy about a philosophical problem, it might turn out (given, e.g., tremendous advances in cognitive neuroscience) to fall within the ambit of science after all.


NS's "Thought Experimenting as Mental Modeling"

1. NS claims that NT's account requires "soundness...a truth-preserving guarantee". But this is a mischaracterization of NT's account. NT allows not only for deductive arguments (which can provide for soundness and validity [i.e. "a truth-preserving guarantee"]), but for inductive and abductive (i.e. inference to the best explanation) arguments as well. This is a huge mistake on NS's part. Given such fundamental misinterpretations (there are others -- see below), it's hard to imagine how she could be in any position to critically assess NT's account.
2. NS claims that an essential feature of (most/all?) TEs is that they involve a narrative structure, and this is why -- she thinks -- NT's Argument account can't be the whole story. Well, for starters, NT never claims that his account is anything like the whole story. He frankly admits that his account isn't designed to explain every feature of TEs and thought-experimentation. Indeed, his account is an answer or attempted solution to a single problem, namely, the epistemological problem. But there are other things to be worried about here: Why is a narrative structure incompatible with an argument-structure? TEs could still be arguments that can -- and, for various reasons (e.g. rhetorical, pedagogical, facilitative, etc.), probably should -- be converted into a more attractive and widely accessible narrative form, especially to non-experts with respect to some domain. More generally, CP might surely be right that TEs-qua-narratives are (at least generally) easier and more fun than TEs-qua-arguments. (We might put this in Fregean terms: while a TE-qua-narrative and TE-qua-argument surely have distinct senses, they have the same referent. The fact that each has its own 'mode of presentation' does not entail each picks out a distinct referent.) Finally, as NS emphasizes, as they appear in their final "polished... narrative form", TEs are the upshot of much painstaking work, and only appear in their narrative "... after the experimenter has determined that it achieves the desired outcome." This seems to suggest two results: First, if NS is right, then strictly speaking, the TE-qua-narrative is unnecessary -- it's useful, of course (e.g. for rhetorical and pedagogical purposes), but not at all necessary. Second, this further suggests that the TE-qua-argument is really doing all the painstaking work: after all, the experimenter only converts the TE into a narrative form after she's independently established that it achieves the result conveyed or illustrated by the TE-qua-narrative. Indeed, NS peppers her discussion of (what I've been calling) TEs-qua-narratives with remarks and qualifications that are exactly in line with what I've been suggesting -- e.g. TEs being "conveyed through" narratives; TEs-qua-narratives being useful pedagogical tools in "facilitating understanding"; TEs-qua-narratives being "effective rhetorical devices". (I'll provide more examples during our meeting on Tuesday.)
3. There's more I wanted to say about NS's own account (rather than her criticisms of NT's account), but I'll save it for our meeting on Tuesday -- this entry is already far too long!

Problems with Jim's Platonic Account

Much of what I have to say probably reflects the fact that most of Jim's account eludes my understanding. (Perhaps I can't quite see how it works!). Sorry for the length - believe it or not, this is the condensed version of my notes. Of course, I don’t expect anyone to read it all :)

TEs as visualizable

Brown takes it to be a virtue of his account that it can accommodate the phenomenology of thought-experimentation. (I'll say more about this phenomenological stuff later.) In particular, he thinks it nicely captures the idea of TEs being 'visualizable'. He even seems to claim that visualization is a necessary condition of any TE, since he rules out TEs that don't involve "a picturable state of affairs". (LAB, p. 18) Presumably, 'visualization' must mean something like: pictorial-mental representation via the 'mind's eye'. Does this apply to all TEs? To what degree/extent must a TE be visualizable? Does this condition carry over to TEs outside of natural science? Brown's account purports to restrict itself to TEs in natural science, but he often appeals to examples outside of natural science, e.g. Searle's Chinese Room, Thompson's Violinist, Jackson's Mary - and not merely for illustrative purposes. They're each examples of easily visualizable scenarios, so he must think the visualization-condition carries over to TEs in philosophy, too. But that's not clear at all; and it's even less clear why a good TE in philosophy must be visualizable, especially since some of the best TEs in philosophy hinge on making a conceptual point. In what meaningful sense is (e.g.) Spinoza's perspective sub specie aeternitatis or Nagel's 'view from nowhere' (which could surely constitute key elements of a good TE) visualizable? And why is its visualization even relevant? Also, given much of what Jim goes on to say, it's deeply unclear how the visualization-condition is supposed to apply. For instance, he points out that some of the best TEs - e.g. Maxwell's Demon - depict a "contradictory situation." As he puts it: "Though the whole situation is contradictory, the [TE] works..." Borrowing from Norton, it would take an "epistemic miracle" to see or visualize a contradictory (i.e. impossible!) situation. At minimum, we need a clearer account of (a) what we're 'seeing' in a contradictory situation, and (b) the degree to which TEs must be visualizable - one that doesn't imply that we can visualize impossible situations, e.g. a round-squares floating around in Plato's (non-spatial) Heaven. :)

TEs, Thought, and the 'External World'

Jim draws a distinction between a TE and a "psycho-linguistic experiment". (LAB, pp. 15-16) The latter, he thinks, is more accurately characterized as a "real experiment" or "empirical experiment [EE] about thought itself." TEs are about "the external world and thinking is the method of learning something about it." This entails (as Jim admits) that there can be no TEs about thought itself. But this can't be right. Indeed, Jim himself cites a clear counterexample without recognizing it as such. On p. 27, he claims that Searle's Chinese Room is a TE "which has to do with the nature of thought itself"! Searle's Chinese Room is a TE about thought itself which purports to show (roughly) that thought is not (mere) computation. Thus, if he's right about Searle's Chinese Room, his distinction between TEs and psycho-linguistic experiments (and EEs in general) is kaput. Perhaps he means only that TEs in natural science can't be "about thought itself." But that doesn't hold up either. First, unless we're substance dualists, the nature of thought is at least partly within the epistemic ambit of scientific inquiry, and it's surely part of "the external world." If so, nothing rules out the possibility of natural-science-TEs about thought itself. So the distinction is untenable as it stands.

Issues with the Taxonomy and Jim's Critique of Norton's Argument Account.

1. Jim's taxonomy seems to imply that while there are Destructive-Direct TEs (i.e. Platonic TEs), there can't be any Destructive-Mediative TEs. But that's implausible. If Mediative TEs simply "facilitate understanding", or serve as a kind of "midwife" guiding us from a background theory to a new conclusion, or (sometimes) play a more "instrumental role" (which, incidentally, he never explains very clearly), then why must they be only constructive? Ditto for Conjectural TEs. The taxonomy doesn't seem to allow for Destructive-Conjectural TEs, but without justification. Even if we're not starting out with any "well-articulated background theory", by establishing a thought-experimental phenomenon P that requires explanation and conjecturing some explanation E of P, we seem to be ruling out other alternative explanations of P, at least those that are incompatible with E. But this 'ruling out' is clearly destructive. Indeed, we can say more than this: in an important sense, every constructive TE is ipso facto a destructive TE. If I can establish that P via some TE, and if P is incompatible with Q, with the result that Q has been ruled out, isn't this a destructive (as well as constructive) result?
2. Jim claims that Mediative TEs don't do any supportive/evidential work. This is mainly why he thinks that Norton's argument account can't accommodate them. Rather, they play a psychological (rather than justificatory) role, in that they provide (where necessary) the "psychological help" needed to facilitate understanding or acceptance of some theory T. But why can't a Mediative TE provide at least indirect support for T by (e.g.) showing it in a better, clearer, more plausible-looking setting? It seems like a persuasive tool - it's not just psychologically useful, illustrative filler. Indeed, they might function very much like arguments by analogy - or more generally, as appeals to consistency across structurally-similar cases. As Jim says, "a mediative [TE] might illustrate some otherwise highly counterintuitive aspect of the theory thereby making it seem more palatable..." (LAB, p. 35) The idea is that if you accept T in a more plausible-looking or intuition-friendly setting (i.e., the setting as depicted in the mediative TE), then - on pain of inconsistency - you should accept T in a less plausible-looking (etc.) settings, unless there's some relevant substantive/structural difference. Also, if a Mediative TE succeeds in showing something, shouldn't it be possible to show how it shows it, i.e. to reconstruct as an argument?
3. I'm having a hard time making sense of Conjectural TEs, and why Jim thinks they pose a problem for Norton's account. First, he claims that Conjectural TEs don't start with some theory but end with one. What does this mean exactly? Surely any TE, to make sense at all, must presuppose some background theory or another. Second, I think Norton's right in claiming that a Conjectural TE is an inference to the best explanation (or perhaps more accurately, that they end with a inference to the best explanation). After all, it posits some thought-experimental phenomenon and offers an explanation of that phenomenon, i.e. a theory which (purportedly) best explains that phenomenon.
4. One of Jim's objections to Norton's account says that even if it were possible to reconstruct any TE as an argument (and so far, I haven't found any good reason to think it isn't - anyone disagree here?), it doesn't accurately reflect the actual psychology or phenomenology of thought-experimenters. For starters, this does nothing to harm Norton's account of the justification and evaluation of TEs (i.e. in terms of arguments). Second, Norton doesn't seem to be claiming that his view is an account of what's actually going on at the level of conscious awareness whilst thought-experimenters are thought-experimenting. Third, Norton's view is compatible with thought-experimentation involving tacit argumentation, even if it's not explicit during the conscious experience of thought-experimentation. Fourth, Jim's own account of what's going on in no way corresponds to or reflects my own phenomenology vis-a-vis thought-experimentation. Crucially, in developing TEs, I don't experience or take myself to be 'seeing the laws of nature' - at minimum, I take myself to be defending or criticizing some theory via a non-actual but possible case (otherwise, I wouldn't introduce my TEs by saying, "Suppose that...", etc.) which I take to show what I'm trying to show by developing such a case. Of course, as I just pointed in connection with Norton, Jim might be right here: perhaps at a deeper level, some sort of tacit Platonic seeing is going on. Who knows! The point is that if it's unfair to criticize Jim on such grounds (and I think it is), it's unfair to criticize Norton on the same grounds. So, we shouldn't criticize either account on the ground it doesn't match up with one's conscious awareness of thought-experimentation (let alone one's interpretation of such goings-on). Finally, there are clearly several paths to the same destination: even if Jim is right that some (many?) thought-experimenters don't thought-experiment a la Norton, it doesn't follow that no thought-experimenter does - let alone that no thought-experimenter should.

The Mystery of Platonic Perception

Jim admits that Platonic perception ('P-perception' for short) is mysterious, but I don't think he appreciates the extent to which this undermines the plausibility of his Platonic account of TEs. Obviously, a good deal of what I'll say here (which isn't very much) should be taken as tentative/provisional, mainly because I just don't have a clue what I'm being asked to consider (which is also why I can't say very much about it!).

This last point is crucial, and that's where I'll start. If there is such a thing as P-perception, Jim doesn't tell us anything about it. No attempt is made to offer a positive account of exactly what it is, how it works, and how (let alone why) it succeeds/fails. It's frequently - albeit loosely and spuriously - associated with comparatively less mysterious/unpalatable notions, such as "grasping", "intuiting", "understanding" and the like. Not only are these latter notions in no way equivalent (conceptually or ontologically) to "perceiving" or "seeing", Jim doesn't show us how they're connected to (much less how we can understand P-perception on the basis of) such notions. If all he means is (say) intuition, why talk in terms of "perception" at all? To invoke talk of "perceiving" abstract objects borders on a category mistake: Jim might be right that knowledge doesn't require a causal connection to some physical object or property; but perception - at least in the sense that we even remotely understand and use in our literal, philosophically serious talk - clearly does. And that's why I'm at a loss as to why "perception" is invoked at all.

Jim argues that the mystery of P-perception is acceptable because it's analogous to the 'mystery' of "ordinary [i.e. sensory] perception". If we accept it in the case of the latter, we should have no qualms about accepting in the case of the former. I think Norton does a great job criticizing this argument; also, Joey raises some interesting points of his own in addition to summarizing Norton's position on the matter. So I won't rehearse these points in detail - see pp. 358-365 for Norton's arguments (and see also Joey's post!). Instead, I'll try to develop a few points of my own.

Jim claims that ordinary perception and P-perception are equally mysterious: "...the perception of abstract laws of nature is certainly [!] no more mysterious than that [i.e. belief about the object seen, the 'second' part of ordinary perception]." (LAB, 108) I doubt that many cognitive psychologists, neuroscientists, and philosophers of mind would follow Jim in his certainty here. We don't have a complete, fully articulated theory that commands universal consensus, but to claim that it's a "complete mystery" is absolutely incredible. Moreover, at least we understand part of the story; not so for P-perception: we don't understand any part. Nothing whatsoever. Indeed, P-perception is almost 'Meta-Mysterious': it's not even clear what we don't have any understanding of.

But let's suppose Jim's right about both types being 'equally mysterious'. What could we conclude from this? We couldn't conclude that this provides any positive support for, or even makes room for rational and principled acceptance of, P-perception. If the 'equal mystery' claim holds, we should deeply suspicious of both. Salve double-negation, two wrongs don't make a right (and not just in ethics). We appreciate this fact in other contexts. Consider, e.g., one of Plantinga's (bad) arguments for the rational acceptability of Christian faith: Do we have conclusive grounds for believing in the existence of minds other than our own? Suppose our grounds for the one (belief in God) were rationally/evidentially equivalent to our grounds for the other (belief in other minds). What could we conclude from this? Could we conclude that Christian faith is thereby vindicated, because it's just as bad - rationally/evidentially - as belief in other minds? Not at all. You can't show that one thing is good by showing that something else is just as bad as what you claim is good. In the same way, if ordinary perception is just as mysterious as P-perception, then so much the worse for both. Fortunately, there are good reasons to reject the 'equal-mystery' thesis.

Norton does a good job in showing why it's completely mysterious not only what P-perception is, but how/why it succeeds/fails. But I want to point something out about justification and evaluation in particular. Suppose Jim is right that in (successful) thought-experimentation, Platonic Seers can see the laws of nature. What if I can't see what they see? Suppose I'm lame, uninsightful - 'Platonically challenged', so to speak. I just don't have the 'gift'. Why should I trust the alleged Seer? It seems to me that I'm perfectly entitled to say:

Good for you! Now show it in a way that's accessible to all of us. Surely, if what you claim is true; if what you claim accurately represents how the world is, it must be possible to give reasons for why this is so. It must be possible, in other words, to give us a sound or inductively strong argument that shows whatever it is you claim that you have 'seen' with your Platonic 'eyes'. After all, how else can we rationally assess (let alone accept) your claim - we who do not See as you do, O Great Seer?

Another way to put the point: when the Seer comes back down in the Cave, he shouldn't expect us to merely take his word for it. If he can't do what we ask of him, why should we believe anything that he says?

An Argument for P-Perception?

In closing, let me gesture at a possible defense of P-perception. I'm not sure that it works, but it makes at least the possibility of P-perception more digestible. In Knowledge and the Flow of Information, Fred Dretske defends a distinction between sensory perception (i.e. the perception of objects) and cognitive perception (i.e. the perception of facts). Very roughly (omitting details), while one S-perceives such objects as cats and mats, one C-perceives that the cat is on the mat - the latter involves not merely objects, but a state of affairs (i.e. the cat's being on the mat) and a relation (i.e. x is on y [or something similar]), which constitute a fact (i.e. that the cat is on the mat). Again, ignore the details and just reflect on whether C-perception is plausible. If it is, then it somehow seems to involve, at least in part, the perception of abstracta. Why? Well, plausibly, states of affairs and relations (or at least the latter) are abstracta. If so, then C-perception is (at least partly) the perception of abstracta. But if C-perception makes sense, then - contra my claim above about the inextricable link between physical causation and perception - this opens the (metaphysical) door to P-perception. Moreover, it seems very plausible that the candidate objects of P-perception - namely laws of nature - are facts of some kind.

A Few Notes on Platonic Perception

I want to note straight out that I, like Jim Brown, am deeply sympathetic to Platonism in mathematics. I want to be able to say that we can have a priori knowledge of mathematical truths, and it strikes me as tempting to say that we attain this knowledge through a process that is somehow analogous to perception. How would we come to know mathematical truths that cannot be given in experience, except by somehow ‘glimpsing’ them with another faculty? Brown thinks this is precisely what is going on, even if no account of what this faculty of mathematical perception is or how it works can be given. Now, on Brown’s view we can sometimes ‘glimpse’ the laws of nature in much the same way that we ‘glimpse’ that 2 and 2 are 4. And indeed it is not readily apparent why we should find Platonism a plausible view in the mathematical case but not in other cases of (apparent) a priori knowledge.

Now, one might object that we just cannot understand what it is to see abstract entities such as laws, numbers, sets etc. And Brown is more or less willing to concede that we really don’t understand how this sort of platonic perception works. Thus the analogy between platonic and ordinary perception breaks down. However, Brown doubts that this can really stand as an objection to his view. He points out that we really don’t have an adequate understanding of how everyday perception works either. So critics who demand a full account of how we make epistemic contact with abstract entities are simply demanding too much from the Platonist. Moreover, the mere fact that we don’t have such an understanding is not by itself evidence against the existence of this kind of perception (doubtless a lot of things exist of which we don’t have an adequate understanding).

John Norton thinks this is an unsatisfactory response to the worries of the empirically minded. Brown’s account allows that platonic thought experiments (even very good ones) can sometimes lead us into error. Such thought experiments afford apparent glimpses of the true laws of nature, but in some cases this is an illusory appearance. What appears to be an instance of a priori knowledge of natural laws can turn out to be no knowledge at all. Brown suggests that we need not be troubled by the fallibility of platonic perception any more than we are troubled by the fallibility of ordinary perception. The argument from illusion is hardly taken seriously in the latter case, so why should it be in the former.

Norton thinks this overlooks a serious disanology between the two cases, which he summarizes as below:

(3A) Ordinary perception is well understood insofar as it governed by an array of familiar regularities. They make the success or failure of ordinary perception controllable and intelligible in all but exceptional cases.

(3B) The mechanism of perception of Platonic laws is essentially completely mysterious. We have no understanding of why this perception succeeds when it succeeds and fails when it fails. ( "Are Thought Experiments Just What You Thought" 359-60)

I think it is difficult to deny 3A, without opening ourselves up to a somewhat unsavory skepticism about sense perception. We may not ultimately understand all the mechanisms at play in ordinary perception, but surely we know something about the conditions under which it is reasonable to trust our senses and those under which their deliverances are particularly liable to be deceptive. We have whole catalogues of perceptual illusions of which we are aware, and we can and do correct for them. This itself seems to indicate some understanding of the mechanisms in play, even if it is imperfect. The upshot is that sensory perception is familiar enough to us that we have a kind of system (not perfect but a system nonetheless) by which to gauge its reliability in particular cases. It may be true that in no particular instance can we be absolutely certain that our senses are not deceiving us. But if we’ve agreed not to take the argument from illusion seriously, this is no problem.

However, if 3B gives an accurate picture of platonic perception, this does strike me as a serious problem. Norton’s argument view can say quite simply that thought experiments fail when the argument of which they are but a picturesque presentation is either unsound or invalid. We can choose between thought experiments with different conclusions by weighing the comparative strength of the arguments they can be reformulated as. This is all very neat and convenient. But if platonic thought experiments are not arguments, but perceptions of abstract entities, and this perception is fallible, how are we to choose between them when their ‘findings’ conflict? It seems to me that if the Platonist view is to be persuasive, we need to be able to say that the working of platonic perception is not just mysterious, but that it is governed by something like the sort of regularities governing ordinary sense perception. Further, we need to be able to say something about these regularities, so as to ascertain the conditions under which a platonic thought experiment is most likely to be a genuine source of a priori knowledge. I am not ready to concede that the Platonist cannot meet these demands, but doing so would seem to require a more robust account of platonic perception than what Brown gives us.