tag:blogger.com,1999:blog-7612775512249380628.post6892222311089782456..comments2023-10-20T03:22:48.388-07:00Comments on Common Currencies: Draft: Philosophers should be interested in ‘common currency’ claims in the cognitive and behavioural sciencesDoctor Spurthttp://www.blogger.com/profile/16403355179680558182noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-7612775512249380628.post-32385280542457664682014-07-05T06:59:03.840-07:002014-07-05T06:59:03.840-07:00Thanks again, John. A partial response for now:
E...Thanks again, John. A partial response for now:<br /><br />Elimination by aspects is a useful case. It could provide an example of what I meant by an incompletely ordered value representation. Suppose that I was choosing a car, and wanted more top speed, more acceleration, more fuel economy, and more rear leg room. But I was also bad at expressing (and perhaps just didn’t have) a determinate sense of how I’d trade those off if required to. Then I might choose by eliminating aspects, but my partial valuations (of separate aspects) would be inputs to that process.<br /><br />But maybe I’m choosing a novel as a gift, and I want a title less than one year old, by a writer who has won at least one prize before, where the story is set in Ireland, but also involves animals. Then my elimination by aspects process will be more procedural - I might not ever have to deal with relative valuations at all - especially if I’m happy to accept the first candidate that satisfies all the conditions.<br /><br />Two things about this, though:<br /><br />(1) We need to be careful to hold constant the fact that the behaviour produced is supposed to be consistent. This matters because the issue (in the rather cryptic paragraph you commented on) is competing explanations of consistency, and whether they refer to proximal mechanisms that consume representations, including representations of value.<br /><br />So returning to elimination by aspects, compare the two illustrations I came up with. In the one case it seems possible that elimination by aspects could sometimes produce consistent choice (that is, consistent with a definite ‘exchange rate’ between quantities of the various aspects, which might be discoverable via something like conjoint analysis). But it’s hard to see that it would generally or always.<br /><br />But in the second case, it seems far less likely that choice would be consistent. Elimination by aspects transforms intractable (or prohibitively expensively tractable) problems into more tractable ones that produce decent enough solutions often enough. <br /><br />(2) I think there’s more than one way of a value representation being incompletely ordered. Here’s one: It might be ‘locally complete’ for a number of separate categories of valuation (food, mate selection, nest location) but instead of all of those having determinate exchange values, the overall ‘mode’ of activity could be determined by semi-procedural systems. (For example, ones that simply interrupt or over-ride foraging when mating prospects are identified.) <br /><br />Here’s another: It could be made of a series of locally complete orderings, but which are organised into a hierarchy with discontinuities, such that no amount, no matter how large, of any good from a ‘lower’ level was worth as much as any amount, no matter how small, of one from a ‘higher’ level. (People sometimes say something like this about so-called ‘sacred’ or ‘moral’ values.)<br /><br />I’ll keep on thinking about other parts of your comment.Doctor Spurthttps://www.blogger.com/profile/16403355179680558182noreply@blogger.comtag:blogger.com,1999:blog-7612775512249380628.post-73289702516271394012014-07-02T20:49:52.539-07:002014-07-02T20:49:52.539-07:00Meant to reply a while ago. This'll be a littl...Meant to reply a while ago. This'll be a little scattered. I think that's a useful list of the possibilities. I'm not sure though that your description of 2b and its gloss in parentheses match up, although I'm also not exactly sure what 'incompletely orderly' means. In particular, I'm not sure whether it means that you can assign a value to anything but these values induce, for example, transitivity violations or whether it means that there are some things you cannot/do not assign a value to. I'd like to keep as separate possibilities systems which exclusive take in values but these values do not induce consistent preferences and systems which do not exclusively take in values.<br /><br />I'd guess that humans are examples of the gloss of 2b.<br /><br />Something like 2c is what 'rule-based decision making' was pointing to, although my own view obviously isn't that people exclusively do this. To flesh out my thought a little, the kind of idea I had in mind was something like elimination by aspects (chosen as an example only because it's classic). People are choosing from some consideration set, they go through this procedure, and they eliminate items giving rise to some set of preferences. I initially thought of this as a case where it doesn't seem like you should necessarily posit a system which takes in values, but perhaps proponents would want to posit some kind of ordinal one.<br /><br />With respect to your question re how much consistency a rule-based system might exhibit, I think that's an interesting question, but that half the work is probably making the question itself precise enough to get a handle on :) I think that could well be a useful thing to try and do. The question makes me realise that I'm not sure the extent to which you want to conflate representations which stand for preferences and representations which stand for values. <br /><br />John<br /><br />Unknownhttps://www.blogger.com/profile/16866021538789823841noreply@blogger.comtag:blogger.com,1999:blog-7612775512249380628.post-38973509719839602014-06-26T13:42:55.088-07:002014-06-26T13:42:55.088-07:00Hey John,
Thanks for the comment/question.
Space...Hey John,<br /><br />Thanks for the comment/question.<br /><br />Space constraints did indeed force me to be rather cryptic in that paper - the conference proceedings maximum length was about half the size of my working draft. (A slightly revised version of the text above is now in press, but I'm also still tinkering on the longer version.)<br /><br />That said, no, you’re not misreading me. But I agree that you’ve described a possibility that my ‘because’ elided. So here’s a longer (not exhaustive) list of options, given order/consistency in behaviour:<br /><br />1. There’s a mechanism and it consumes an orderly value representation. (Inferring from ultimate to proximal, as some think we must.)<br />2. There’s a mechanism, but it doesn’t consume an orderly value representation:<br />2a. It’s some kind of architecture with no representations at all. (One class of examples would be subsumption architectures, but other distributed architectures might also be candidates.) <br />2b. It’s some kind of architecture with an incompletely orderly value representation. (A mix of valuations and procedures, or valuations and subsumption, or …)<br />2c. It’s some kind of architecture that consumes representations, but not value representations.<br /><br />I reckon, by the way, that humans are instances of (2b).<br /><br />I take it that (2c) is what you were driving at when you referred to “rule-based decision making”. And yes, that’s a perfectly coherent option. (One could also hypothesise a large enough lookup table, which might count as representational but for all that not involving rules in any interesting sense.)<br /><br />I wonder how much consistency (with some behaviourally determined preference ordering) it would be possible to get a rule-based system to exhibit without it processing representations that stood for preferences. Do you have any thoughts about this?Doctor Spurthttps://www.blogger.com/profile/16403355179680558182noreply@blogger.comtag:blogger.com,1999:blog-7612775512249380628.post-87695791731326984992014-06-26T12:15:41.273-07:002014-06-26T12:15:41.273-07:00Hey,
I enjoyably read parts of your common curren...Hey,<br /><br />I enjoyably read parts of your common currencies blog a couple days ago - it's a fun read. <br /><br />I was a little puzzled that it seemed like you don't leave much room for the idea of agents with consistent behavior, without a proximal currency but with representations. For example, in the why philosophers should be interested paper: "Not everyone who thinks that behaviour is consistent, and that there is a mechanical process explaining behaviour selection, is committed to a proximal common currency. This is because not all views about how behaviour is caused involve representations, including value representations."<br /><br />Perhaps you're not really committed to this view and just expressed it this way because of space constraints or perhaps 'mechanical process' is doing some work for you here, but otherwise the 'This is because' seems wrong to me. I'd claim that the 'interesting' reasons people may hold the view in the first sentence mostly have nothing to do with denying representations but rather with models which posit processes like rule-based decision making. Another way of putting this is that it seems to me that there are interesting questions about the correctness of views which claim that process models should posit a common currency, but the opposition to these views worth grappling with isn't solely (or even mostly) views which deny representations but rather views which deny that the way all decisions are made is by assigning a value to each option and then choosing the option with the highest value.<br /><br />But perhaps I'm misreading you and just missing something obvious!<br /><br />JohnUnknownhttps://www.blogger.com/profile/16866021538789823841noreply@blogger.com