16 February 2009

Thompson Puzzles

From "Naive Action Theory", footnote 33:

It is interesting that the examples through which Anscombe attempts to illustrate the idea of "many descriptions of the same" do not actually illustrate it: it is the rare act of moving an arm that can be classified as a replenishment of a house water-supply.
I have no idea what's supposed to be wrong with Anscombe's example (from ss23 of "Intention"). It looks to me like a perfectly fine case of redescribing an action in multiple ways. The rest of the footnote seems to imply that Thompson wants to say that "replenishing the house's water-supply" has "moving my arm" as a proper part, rather than this being something it's identical to. But this doesn't seem right. Moving my arm, in these circumstances, just is my replenishing the house's water-supply. It's not a part of something else I do; the only thing I do to replenish the house's water supply is move my arm (while holding the pump-handle). Once the event that is my moving my arm is in view, there's nowhere else to look to see me replenishing the house's water supply. (Though you might need to look elsewhere to get the necessary background in view to recognize that I am in fact replenishing the house's water supply by moving my arm. Taking in this background is not taking in more actions performed by me. It's seeing that the house has functioning plumbing, and that my hand is holding the pump-handle.)

The preceding footnote also mystifies me:
Mention of this great paper [Davidson's "Agency"], in the present context, invites the remark that its account of the concept of agency fails to take proper account of actions with parts. Surely it will be "agency" in the sense Davidson means to capture if the agent sinks the Bismark, or ruins her finances, by doing A, B and C, each of them intentionally. But the events falling under the descriptions A, B and C need not fall severally under the description "a sinking of the Bismark" or "a ruining of her finances", as the case may be; none by itself, we may suppose, adds up to that. And so it might be that nothing done intentionally falls under that description, and thus that something "done", in the emphatic sense Davidson means to elucidate, isn't done intentionally under any description.
I can't see how this argument is supposed to work. Suppose {A,B,C} are things like "moved her arm towards the firing button", "moved it even closer to the firing button" and "pushed the firing button". The third would be identical to a sinking of the Bismarck (in the sense of "sinking" which people do, as opposed to what boats do). Standard "pulling-the-trigger-is-killing-Jones" sort of example. But that doesn't fit Thompson's criteria. He needs items {A,B,C} such that none of them is identical to a sinking of the Bismarck. So, none of them can be anything like "pushing the button that launches the torpedo that sinks the Bismarck". Suppose we change C to "moved her arm closer still to the firing button" -- now we have intentional actions {A,B,C} such that none of them is a sinking of the Bismarck, but each of them are done as part of the agent's attempt to sink the Bismarck. But this doesn't work, either, since she doesn't sink the Bismarck by doing A, B, & C; she sinks the Bismarck by doing A, B, C, and D (where D gets her the rest of the way to the button and has her pushing it). I'm not able to think of any {A,B,C} that actually fit Thompson's criteria.

The "ruining her finances" bit seems like it might work a little better: Sue can do A, B, and C such that any two of them would not ruin her, but all three bankrupt her if done together. But then it seems like whichever happens to be the third will be identical with her ruining her finances -- as the straw that breaks the camel's back. Thompson's "none by itself adds up to that" seems to obscure this feature of Davidson's theory of action: what casual relations an event ends up standing in can determine what descriptions are true of it. "By itself" an event doesn't tell you what descriptions are true of it. Other events are relevant for that.

Another thought I had was: it's a gestalt thing. Whole is more than its parts. Maybe nothing she did was, by itself, "ruining her finances", but she just ended up with ruined finances after doing various things. But here it looks like either there is no action of hers which is a ruining of her finances, or else the thing she did which is a ruining is composed of the other things she did. In the former case, there is no action to account for; in the latter case, the fact that no part of the action is describable as a ruining doesn't cut against Davidson, that I can see. For there is an action which is describable as a ruining: the event composed of A, B, and C (or whatever went it to the relevant gestalt in this case). If there are gestalt-y actions, then I'm not seeing why Davidson has any problems with that.

Also: saying "this is no worse than the Sorite's paradox" does not seem like much of a help when trying to defend a doctrine. But this seems to be what Thompson says here:
Such an appeal to the idea of vagueness carries with it a number of theoretical difficulties, but supposing them handled, the same vagueness would no doubt then be found to infect the division of our nested classes of descriptions into "pre-intentional" and intentional. In that case, my conjecture ── viz., "Acts of moving something somewhere intentionally always have an initial segment which is also an act of moving something somewhere intentionally" ── could again be sustained, if only it were given the sort of construction that an adequate theory of vagueness would supply for such sentence as "No one is made bald by the loss of a single hair".
I'm not sure how such a construction could work for Thompson's conjecture. Avoiding the Sorite's requires mathematical induction breaking down at some point (or else removing hairs one at a time until they're all gone does not give you a bald head). It seems plausible that vague terms are like this. Induction with them only works for a bit, then it shades off into not working anymore. You take a hirsute head, remove a hair (not bald), remove a hair (not bald), and then eventually he's bald. (What you say in the details of this are, as far as I can tell, generally still up in the air.) It doesn't strike me as very plausible that Thompson's terminology her employs vague terms. "Initial segments of acts" is not the sort of phrase that you pick up from ordinary language (the usual home of vagueness); it's a technical term in the theory of action. More to the point, I don't know how the vagueness story would go, here. You identify an initial segment of an action (which gives you another action), and then an initial segment for that action, and then at some point you can't? (This also seems to cut against Thompson's "quasi-Kantian" defense of his claim, which rests on the "Axioms of Intuition"'s claim that what is given in intuition is always an extensive magnitude. Thompson seems to offer two different and incompatible defenses for his weird view.)


Ben Wolfson said...

Moving my arm, in these circumstances, just is my replenishing the house's water-supply.

And, indeed, Anscombe says explicitly that "more circumstances" are required for an arm-moving to be a water-supplying than for it just to be an arm-moving, in I think section 26.

Shawn said...

This doesn't respond to the post at all, but there's a nice review of Thompson's book on NDPR now.

Daniel Lindquist said...

It is indeed a nice review.

The comment at the end about "relative rationalization" has come up in Ford's class. The stuff at the end of "Naive Action Theory" (the "Myth of Jones" parallel) is supposed to address that, I think. The "sophisticated" (non-naive) explanations of action are actually dependent on the naive. They grow out of it. So sophistication doesn't actually make any progress; it just seems to (due to dualisms like inner/outer, spirit/nature, mind/body, speculates Thompson).

jon said...

Found this post when going through 1000+ unread posts in google reader.

Ben's right-- I think the key is the "in these circumstances." There's nothing about button-pushing, trigger-pulling, or arm moving *as such* that makes these amount to Bismark-sinking, Jones-killing, or cistern-filling, respectively. That's Thompson's point.

What permits the argument against Davidson to work is his inattention to the way in which the circumstances figure in to the description of what's going on under which we've got an intentional action that's truly the sum of its parts. Davidson's way of building the "other events" in won't do it-- my button-pressing may be Bismark-sinking, but if I don't know this (perhaps I think the button launches a torpedo at the George Washington, rather than the Bismark) then I won't be intentionally sinking the Bismark. Davidson might want to bring in other events, but I think the right ones to bring in will often be in the past-- perhaps my having learned the target at which the torpedoes are aimed by checking the periscope. And my button-pressing certainly doesn't cause me to do make these observations, with the (I hope) reasonable assumption that there ain't no backwards causation.

The thing which unifies the parts is the "cognition requirement" described in the paper. The action is unified by nothing less and nothing more than the agent knowing what she's doing: which, depending on a host of relevant circumstances, might be further explicable in terms of her knowing either or both how or why she's doing it.