Links

I don't really have time to write a proper post right now, so instead I'll link to some interesting discussions where I've recently offered comments:

At Parablemania, What is Race?
From Fake Barn Country we have A Puzzle about Obligation, The Impossibility of Believing at Will, and Anti-Consequentialism and a God's-eye View

Truth Machines

I've been thinking about James L. Halperin's book, The Truth Machine. It portrays a near-future transformed by the invention of a 100% accurate lie-detector. Really interesting stuff. Halperin's vision of the consequences is very optimistic - 'utopian' even - but he manages to describe it all in a very compelling and plausible way.

Apart from all the usual ponderings of whether such an invention would be a "good thing" or not, last night I began to wonder about how such a machine would actually work. Presumably the idea is that our brain/nervous system behaves in some predictably different way when engaging in deceitful, as opposed to honest, behaviour (particularly involving the vocal expression of believed vs disbelieved propositions). But we cannot expect the 'symptoms of deceit' to be easily detectable (modern polygraphs seem to mostly just detect anxiety etc, which isn't quite good enough!). Responses may also vary from individual to individual.

Anyway, what's required here seems to be some sort of pattern-recognition - a job ideally suited to parallel processing, or artificial connectionist networks. As I understand it, connectionist algorithms are not 'programmed', but rather, are trained. You start off with a basic network, with all the inputs connected to the output(s), according to some base strength. Then to "train" it, you simply run through a large sample of known data as input, tweaking all the connection strengths accordingly so that the correct output results. Do this enough times, and it will eventually 'evolve' the ability to recognise the pattern, whatever it may be (possibly unknown even to the trainers!)... thus giving reliable results even when faced with entirely new input data.

I think that this sort of approach might have a lot of potential in psychological analysis, and particularly lie detection. It would certainly be interesting to try, anyway... Start off with a basic network, which takes a person's neural activity as inputs, and a single "lie detector" output. Train it by getting the person to say a large sample of propositions - known to be either true or false - thus 'training' the network to recognize whatever patterns there may be (if there are any), even if we are unaware of them. There's the added advantage that networks could be individually trained to respond to the patterns of a particular individual.

If there are any (patterns). That's the key, I think. If there are, then I think this approach should (at least in theory) be capable of picking up on them. If not, then reliable lie detection is presumably impossible.

So that's my key idea. If reliable lie detection is possible, then 'connectionist' algorithms are one way of achieving it.

Question time (please feel free to respond in the comments to any or all of them!):
1) Do you know whether anyone else has thought of this before?
2) Do you think it would work?

and the fun questions:
3) Do you think the invention of a 100% (or close) reliable lie-detector would be a good thing?
4) Would it be (practically) possible to restrict its use (e.g. to approved law enforcement/security purposes only) for privacy reasons?
5) Would such a restriction be a good thing?

Or anything else you can think of... Fascinating topic anyway, I reckon.

Update: The Economist has an interesting article suggesting brain scans (rather than merely examining physiological symptoms) may indeed have a lot of potential for detecting lies. The approach described uses more traditional computational techniques, rather than connectionist ones. I've still never heard of anyone investigating the use of the latter for this purpose. But then, I haven't looked very hard.

Fictional Worlds

The recent discussion of fictional worlds at Fake Barn Country (fast becoming my favourite blog) has brought to my attention several aspects of my initial semantic contextualism theory which are in need of improvement.

The basic problem is this: I had conceived of fictional worlds as being simply reducible to real-world objects (such as ink on paper) - a sort of 'naive account'. But this is clearly inadequate once you realise all the implicit assumptions that we seem justified in making about fictional worlds, e.g. that Grisham's world has the same US Constitution as ours (where did I see this example? Please mention the link in the 'comments' if you know!).

Apparently David Lewis has a theory relating fictional worlds to possible worlds, which solves this problem. As Joshua explained it: The rough version of Lewis' theory is that a statement S is true in a fiction F iff S is true in all those worlds where F is told as known fact.
Surely that "all" is far too restrictive though, since it would allow us to only make explicit, rather than implicit, inferences from the text. (There is some distant possible world where all the facts of Grisham's stories are true, and told as such, except that Aliens have invaded Earth and rescinded the US Constitution.) More likely it is supposed to be "in all those closest possible worlds where F is told as fact".

I wonder if we can generalise this? To put it into the terminology of my previous posts, let us say that a proposition P is true of a 'sub-world' SW (eg a fiction story, a dream, whatever) iff P is true in the closest possible world (talking about OWs here) consistent with the set of facts S [all the explicit (i.e. 'reduced') facts of the SW].

(I've effectively replaced "where F is told as fact" with "where the set (S) of facts which are told of F, are all true. I wouldn't think this should cause any problems, but do let me know if you disagree!)

Now this is getting a bit mangled, since I'm using the terminology 'sub-world' to refer to some - possibly very indirect - aspect of reality (such as "the world described by fictional story X"), whereas counterfactual "possible worlds" do not refer to anything real, but rather, to different ways the OW (objective world) could, possibly, have been. So this is merging two very different concepts (though both have been discussed in - different - recent posts), with unfortunately similar vocabulary.

Simple example: John Grisham's books are all based on Earth as we know it, where various non-factual events occur (big trials, runaway juries, and whatnot). The closest possible worlds consistent with S, then, will be the worlds most identical to reality in all other respects, except for those (relatively few) facts which must be altered to make all his counterfactual propositions (S1, S2, ... Sn) true and consistent. So the US does indeed have just the same constitution as it does in reality, unless something in the stories (explicitly or implicitly) contradicts this.

I'm not sure whether the generalisation beyond fiction works at all. It may be that the sorts of inferences appropriate to fictional stories are totally inappropriate to various other sub-worlds, such as dreams. I'm really not sure.

I'm not even sure this works for all fiction... what about fantasy worlds, which are entirely divorced from reality? You presumably could have a world "closer" to reality consistent with S, which nevertheless seems less true to the spirit of the story? I feel that an important factor here is reader expectations, shared cultural backgrounds (especially mythologies, etc). This seems a very serious point, and I think it must be from me misinterpreting Lewis' theory (perhaps I shouldn't have added in all that "closest possible worlds" stuff after all? But it does seem necessary in the Grisham case!)

A further problem concerns what the "close possible OWs" are close to. You see, if we are really Brains in Vats, then the possible OWs closest to our own OW consistent with story S would be entirely different from the OWs closest to the world of our common experience (CW) consistent with story S. But we want the latter, rather than the former - our understanding of [stories, dreams, etc] is based on our undersanding of the CW as we know it, rather than the abstract OW, whatever it may be. So we should clarify that by "closest possible OW", we mean the OW which is closest to our CW (i.e. the OW which is closest to [the OW where our CW is fact]). But I guess that's a somewhat pedantic point, and can be avoided if you're happy to assume that we're really not BIVs (or any other similar skeptical hypothesis).

I'm gonna have to think about all this a bit more... it certainly seems no easy task to reconcile the problem of background/implicit assumptions into my previously proposed theory of 'semantic contextualism' with its fully reducible sub-worlds. Bugger.

Update: See the "comments" for a very helpful explanation from Joshua.
I should also clarify that the basic ideas of semantic contextualism (that truth is relative to the 'world of reference', and that what world is relevant will vary according to context) aren't really affected by any of this - the whole 'reducibility' idea helped to keep the theory nice and simple, but it isn't in any way essential to it. But I plan to write a follow-up post sometime soon, which will discuss these matters further.

Frozen Photons, explained

I recently asked how light/photons could get into the present, given that they supposedly don't move through time. I then proceeded to ramble incoherently. This post is to make amends for that. I came up with the following metaphor which may, I hope, make the solution intuitively understandable.

Suppose we imagine an ethereal sphere encapsulating an object, as a visual representation of its 'reference frame'. If the sphere - and all its contents - are moving very quickly (relative to us), then we now say that it appears (to us) that time within the sphere is passing in slow motion. But the sphere itself is unaffected, and still moves just as fast (and, of course, carrying along all its contents with it).

To apply this to the problem at hand: We picture a sphere surrounding the photon, and this sphere travels at the speed of light, and passes through time normally just like everything else (taking its contents - the photon - with it). But time is frozen within the sphere, so that the photon doesn't age.

Problem solved! (Many thanks to Gemma Mason for explaining the basic solution to me.)

"The Real Worlds" - Semantic Contextualism

This post is built on the idea that we can look at the world in many different ways, using a coaser- or finer-grained focus.

Take the ultimate macrocosm of the ultimate reality, whatever it might be. Let's call it the "Objective World" (henceforth, OW). Now, the OW is of such a depth that we can identify and focus on various microcosms, little worlds in themselves. Each of these microcosms will in turn contain further mini-worlds, ad infinitum. Human reflection has an infinite zoom, and can choose to focus on any particular "world" - we are not restricted to the OW. All of these worlds are (at least in a certain respect) real. They are simply different ways of looking at a single reality. A different frame of reference, if you will.

My central theses are that:

• truth is relative to the 'world of reference' [It should be noted, however, that all micro-worlds are ultimately reducible to the OW... but such a reduction will often require significant semantic alterations in order to preserve the truth values of attributed propositions.]


• The 'world of reference' will vary according to context.

That was a rather abstract introduction. So let's consider a simple example: a fictional story. I'm suggesting that the imaginary world communicated by the story is (in a sense) real. One can say things like "Frodo went to Mt Doom to destroy the One Ring", and it is true within the context of Tolkien's world, despite its apparent literal falsity. The method of reduction in this case should be obvious: propositions about Tolkien's world simply need to be 'translated' into propositions about what Tolkien wrote in his famous books. Fictional worlds are reducible to real-world text on real-world pages.

The example I borrowed from Patrick may help you to get an intuitive grasp of this theory:
Suppose I say to you “In my dream last night, I was walking down the street-”, it is clearly inappropriate to respond “Liar! You were lying in bed asleep!”. Instead, we implicitly recognise that the phrase “In my dream last night” alerts the listener to a shift of semantic context: the truth of the subsequent proposition is to be evaluated in terms of the dream world, not the real world.

So that should give you the basic idea. Hopefully this theory of metaphysical contextualism will strike you as intuitively plausible - perhaps even obvious. What really makes it useful, however, is the response it allows us to give to the Skeptic. I discussed this in my post on Skepticism & the Matrix (but note that what I previously referred to as the "Real World (RW)", is what I'm now calling the "OW", since the microcosmic worlds are - in a sense - no less real).

The Skeptic highlights the distinction between the world we commonly assume to be reality (CW), and what is in fact the ultimate OW (which we cannot possibly know). A common scenario is to suggest that you might be a brain in a vat (BIV), being ‘fed’ experiences by a super-computer which electrically stimulates your brain, causing lucid hallucinations which you have mistaken for reality. In this case, our CW is reducible to the rules of the computer program that determines the experiences fed to a BIV in any given situation.

I am suggesting here that even though you are a (handless) BIV in the OW, it is nevertheless true that you have two hands - within the context of the CW. The semantic revision I propose to achieve this is roughly similar to that of Hilary Putnam (in Brains in a Vat) who suggests that a BIV saying “there is a tree in front of me” likely speaks the truth, given what “tree” and “in front of” mean in Vat-English. I suggest that when someone says "I have two hands", what they really mean is "I have two hands in the CW", rather than "I have two hands in the OW". We can know the former proposition to be true, despite our ignorance of the latter.

But enough about Skepticism. My main purpose here was to expand the theory from the mere delineation of CW from OW, to instead allow us to choose from infinitely many reference frames. I can talk about the world of a dream, of a story, of my conscious experiences, of a computer simulation, of interstellar interactions, or of anthills. Some propositions will be true in some of these worlds, but not others. Nevertheless all those worlds are in some sense real, because they can all be 'reduced' to the OW. The microworlds supervene on the macroworld. They just look at reality from different perspectives.

Update: Some helpful folks at Ephilosopher inform me that the name "semantic contextualism" is already used to refer to a certain variant of epistemological contextualism.  To emphasise the idea that there are multiple 'worlds' or perspectives on reality, I'll rename my theory here as "Metaphysical contextualism" instead.

Skepticism and Possible Worlds

This post will explain Robert Nozick's Conditional Theory of Knowledge, how it solves the problem of skepticism, problems with the theory, and how Keith DeRose develops it so as to overcome these problems (and provide a better answer to the skeptic).

Nozick's theory is quite cool. A simple version of it is as follows...
S knows that P iff:

• P is true


• S believes that P


• If P were false, then S would not believe that P


• If P were true, then S would believe that P

The subjunctive conditionals might look a bit odd - especially that last one - but I think they make a lot more sense if interpreted in terms of close possible worlds.

To briefly explain 'possible worlds': We live in one possible world, let's call it "actuality". But you can imagine that things could have been different, in many (indeed, infinitely many) ways. When we conceive of an alternative possibility (say, that I'm sitting watching TV now instead of typing this), then that is an alternative "possible world". The greater the discrepancy from reality, the more "distant" that possible world is from actuality.

So, to explain Nozick's subjunctive conditionals:
a) in the closest possible worlds where P is false (unlike actuality), S no longer believes that P.
b) in all other close possible worlds where P is also true, S does believe P. (How close? Nozick suggests all those worlds that are closer than the first not-P world.)

Note that we are not concerned with extremely distant worlds - that is, it is possible to "know" something even if you haven't ruled out some (extremely unlikely) alternative possibilities.

Now, when this theory is applied to skeptical scenarios, we get the following results:
1) I know that I have two hands
but 2) I don't know that I'm not a (handless) brain-in-a-vat (BIV).

To explain why:
1) I have a true belief that I have two hands. In the closest possible worlds where I do not have two hands, it is because I lost them in some sort of accident, which I of course would be well aware of, and so I would then not believe that I have two hands. Furthermore, in all close worlds where I do have two hands, I also believe that I do. All the conditions are fulfilled, so I have knowledge of this fact.

2) I have a true belief that I am not a BIV. But in the closest possible world where this is false (i.e. where I am a BIV), I nevertheless would still believe that I was not one (since I would have had all the same conscious experiences as I have now). The third condition for knowledge is NOT met, so I do NOT know that I'm not a BIV.

Nozick embraces these results, but surely the contradictory nature of these abominable conjunctions counts against the theory. Fortunately, DeRose found a way to iron out the inconsistencies, by combining Nozick's theory with Contextualism.

Contextualism says that the epistemic standards required for knowledge will vary, and are context-dependent. The basic idea is that according to normal standards we know that we have hands (and thus DO know that we're not BIVs), but according to higher standards we don't (and don't know that we're not BIVs).

DeRose's key insight was to embrace Nozick's conditional theory and its 'possible worlds', but to drop the strict requirements, and instead allow the number of possible worlds under consideration to vary according to context. The strength of S's epistemic position (with regard to belief P) depends on how well S's belief in close possible worlds matches the truth in those respective worlds. The more possible worlds there are where S would have an accurate belief about P, the stronger S's epistemic position is (here and now) with regard to P.

He offers a visual metaphor here which I found illustrative: Imagine our world ("actuality") surrounded by all the close and distant possible worlds, modelled in a sort of 3-d space. Now imagine a sphere ballooning out from the centre (i.e. the actual world), but stopping the moment it comes to a world where S fails one of the two subjunctive conditionals previously mentioned (i.e. #a- S believes P when it is false, or #b- S doesn't believe it when it is true). Equivalently: S believes that P iff P is true in that possible world.

Now, the size of this sphere describes the strength of S's epistemic position. If small, then S can only claim to know P according to very low standards, whereas if the sphere is very large, then S's knowledge is correspondingly strong (after all, it means that S would be very unlikely to have formed a false belief here... things would have had to go very differently (since it is a very distant world) for S to have been mistaken). By restricting comparative knowledge claims to similar epistemic contexts (i.e. sphere-sizes), you avoid the contradictions entailed by Nozick's theory.

So, to apply this new & improved theory to the skeptical scenarios:
1) I have very strong knowledge that I have two hands (because you would have to go out to a VERY distant world before I had a mistaken belief about this!)

likewise, 2) I have very strong knowledge that I am not a BIV. How is this? Well, once again, you have to go out a very long way before you can find a possible world where I would have a mistaken belief about it.

So why do we find the skeptic's argument so convincing? Why do we feel that we don't know that we're not BIVs?

DeRose's answer is that mentioning skeptical scenarios raises the standards of knowledge so high that we cannot meet them. So (according to those extreme standards), the skeptic is right... our "sphere of knowledge" is not large enough to extend all the way out to those (extremely distant) possible worlds.

Just one thing remains to be explained: how does this 'raising of standards' occur?
DeRose posits the Rule of Sensitivity, which effectively says that when knowledge about P is mentioned, we tend to raise the standards of knowledge (if need be, and for an appropriate duration) so that S's belief in that particular P becomes sensitive. That is, the sphere of knowledge must extend all the way to the first possible world where P is false (like Nozick originally suggested).

That means that whenever anyone mentions BIVs, suddenly the standards for knowledge are raised to this absurdly large sphere, which must extend out all the way to that (extremely distant) possible world where S is a brain in a vat! It is no surprise that we do not have knowledge according to these excessive standards.

But, thankfully, our knowledge in more normal contexts is preserved. For most of our beliefs which we think we 'know', our sphere of knowledge extends out quite a fair distance, so we do have knowledge, according to contexts where the required standards are more moderate (i.e. smaller than the size of our sphere).

Thus the Skeptic is refuted, without any bizarre side-effects. If anyone knows of (or can think of) any objections to this solution, I would be very curious to hear them...?

Race, Generalisations and Profiling

Today I came across a couple of very thought-provoking (and controversial!) articles on race issues, one about empirical studies suggesting affirmative action has significant negative consequences, and the other tackles the vexing issue of (especially racial) profiling:

Thus, as Aristotle would have it, if we really want justice, there must be not only rules, but also a discretion-based system to correct for the inevitable errors that the strict application of the rules will produce.

Indeed, the course of history can be written as a tug of war between rule-based and discretion-based methods of organizing and controlling human behavior. Lately, rules are crowding out discretion, and Schauer is all in favor of that. [...]

Schauer's core argument in defense of his position is this: We all use generalizations all the time, whether we acknowledge them or not. So we will all be better off if we are honest about the generalizations we are using, and if we give close consideration to the propriety of the constituent elements of the analytical shortcuts we have created for ourselves. [...]

Here is the important point: Back before the profile was written down, the Customs officials at the borders were already employing rules of thumbs and generalizations.

The only thing that has changed is that there is now transparency in the process. And that is a good thing.

If Transparent Profiling Is Desirable, Can One of the Factors Be Racial?

Once the factors used are made explicit, they can be examined: Do people up to no good disproportionately travel on one-way tickets? The answer is yes.

Do they disproportionately by those tickets with cash? Yes again.

Are they disproportionately young men? For sure.

Are they disproportionately minorities? That one is not so clear, but let's assume the answer -- empirically, as a matter of research -- is yes.

We can then say these four factors have an empirical basis for being in our profile.

Schauer's next step is to ask whether they should be included, and on race, he essentially fudges, which is probably the right thing to do. [...]

Schauer's Point: Transparency Allows Conscious Choice Among Profiling Systems

The important thing for Schauer is not that we decide the racial profiling question one way or another, but simply that we ask that tough question and answer it as a society.

Race issues aside, I think the argument for transparency is a very good one.
But to tackle the really tricky point... should profilers make use of racial statistics and generalisations?
I don't know. But whatever the answer is, it should be the same as whether the profiler is allowed to target individuals based on any other generalisations (eg gender-based ones).

I just don't see any relevant difference between racial and other stereotypes. Men are more violent. Young people are more violent. These are common stereotypes, and if a particular individual who happens to be young and male is targeted by security personnel because of this, then that is no less prejudice than if they targeted him because he is black.

But is prejudice necessarily a problem? Is it wrong to extrapolate from data about a group, to (tentative) conclusions about an individual member of that group? Not intrinsically wrong, I don't think. But there is certainly room for abuse. In particular, it must always be remembered that such extrapolations are extremely unreliable. Doubt should be the default epistemic position.

It seems to me that there is a very real sense in which such stereotypes are dehumanizing, and insulting to the individual who is so targeted. Generalisations deny individuality. They say to us that we are nothing more than the groups to which we belong. I'm not a person, I'm a young white male. Behave accordingly - there are expectations to live up to.

But then again, security personnel have a job to do; one which necessarily demotes everyone to the status of "potential threat". If some sort of sampling is required, then surely we might as well use the most efficient sampling techniques available to us? If given a choice between random sampling, and targeted sampling (such as is more likely to be successful), then isn't it simply common sense to choose the latter?

Well, perhaps, but it still makes me feel decidedly uncomfortable.

Suicide and Traffic Jams

Jonathan Ichikawa links to an article which begins as follows:

More than Iraq, the presidential election or the start of baseball season, one topic is dominating talk radio and water-cooler conversation in the Bay Area this week: the would-be suicide jumper who tied up traffic on the Bay Bridge and surrounding freeways for 13 1/2 hours last Friday.

The opinions fall into two camps.

One says you do whatever it takes for however long it takes to prevent the suicide. Saving a life is worth more than the inconvenience and costly ramifications of a traffic jam, even one that brings a wide slice of the Bay Area to a standstill. To believe otherwise, this camp says, is to abandon a core societal belief in the value of a person's life.

The other side says there ought to be a time limit for negotiating with a jumper -- say an hour or two -- then the authorities should remove him. The resistant jumper might be hurt or even fall to his death in the process, but since he put himself in such a dangerous position, he is ultimately responsible for the result. The rest of us should not be held hostage, the argument goes, to a narcissist -- even a mentally ill narcissist -- who wants to be the center of attention for 13 hours.

Farhad Ajir is surely not the last suicidal person who will find himself or herself paralyzed with fear, doubt or confusion at the edge of the Bay Bridge. So how do we figure out the ethical and social calculus to resolve the debate about handling such situations in the future?

I'm not sure that I agree with either side. Or at least, I don't think a (justified) decision can be made without further information. Of particular importance, I think, is the 'risk factor' - what is the probability that forcably "removing" the jumper will result in his injury or death? Also, what is the base-rate to which we are comparing this, i.e. what are the chances that the guy will jump even after several hours of negotiations?

It seems to me fairly clear that if the risk factor is either a) quite low, OR b) comparable to the base-rate, then there is nothing wrong with forcably "removing" him.

Contrarily, if the risk factor is moderate-to-high (i.e. if the removal is likely to kill the would-be-jumper), then such a removal seems quite blatantly immoral. You can't go around killing people just to ease a traffic jam... it's obscene!

But it's not that simple, as Kat (see Ichikawa's comments) points out:

Considering there are thousands of people stuck in the traffic jam, it seems very possible that one of those people is a diabetic who will die if she has to wait 13 hours for her insulin shot; or what if an ambulance carrying a critical patient has to get across the bridge? Certainly there could be more than just time wasted in this situation.

So there are further empirical facts we need to know in order to answer this question, namely, what are the likely (harmful) consequences of having the bridge jammed up for a long period of time?
Perhaps risking the life of an already suicidal person would be necessary to save other lives? Though I would expect that there must be better alternatives.

So the real dilemma here seems to arise when the risk-factor is on the low side of 'moderate', or if there is a genuine threat to others posed by the traffic jam. In that case, I'm not sure what to do. I guess it would depend on the exact details of the situation (compare the relative probabilities of serious harms being done to either party, etc).

Either way, I think this issue really highlights the importance of empirical facts in moral decision-making. For either camp to just blindly assert an absolutist moral judgement, independent of the contingent facts of this particular situation, strikes me as a serious mistake.

This reminds me of an old article about bridge-jumpers:

Dr. Seiden’s study, "Where Are They Now?", published in 1978, followed up on five hundred and fifteen people who were prevented from attempting suicide at the bridge between 1937 and 1971. After, on average, more than twenty-six years, ninety-four per cent of the would-be suicides were either still alive or had died of natural causes. "The findings confirm previous observations that suicidal behavior is crisis-oriented and acute in nature," Seiden concluded; if you can get a suicidal person through his crisis—Seiden put the high-risk period at ninety days—chances are extremely good that he won’t kill himself later.

It was also discussed at Crooked Timber a while back:

What I found terrifying to contemplate: many of those who survived the jump (not that there are many, I believe the article cites 26 survivors) recount that they immediately regretted the action. Awful to think of the hundreds more who didn’t survive, who probably also regretted the action even as they were plunging to their deaths.

I'm reminded of that saying about how the measure of a civilization is how well it treats its worst-off members. Surely you can't get much worse than those contemplating suicide.

From my idealistic armchair, I, for one, want to live in a society which does what it can (within reason) to help these people. But I guess if I was stuck in a traffic jam for 13 hours I might change my mind...

Institutional Evil

Matt Weiner has an interesting post about employers stealing from their employees by doctoring the electronic time records (for those paid an hourly wage).

How are we to allocate the blame for this? The managers may not be under orders to cheat their underlings, and I think the morally required action is to quit (and blow the whistle) rather than do so, but the blame certainly doesn't stop with them. Upper-level management who put pressure on lower-level management to achieve impossible results surely bear some blame, but they didn't order the theft.

What's going on here is an institutional evil, and I think that's a category that's much underdiscussed in philosophical ethics.* The institutions seem to be set up to put pressure on underpaid district managers, to make cheating easy, and to make it easy for the corporations to turn a blind eye to what's going on. The culpability of the whole is greater than the sum of the culpabilities of the parts. It's worth noting here that institutional practices can make a difference; note the contrast between Wal-Mart and McDonald's, which gives employees printouts of hours worked and doesn't have time shaving problems.

In this specific case I'm more inclined than Matt to blame the managers who are directly responsible. However, more generally, I do think the idea he's getting at about "institutional evil" is a very important one.

Actually... isn't this sort of thing a central focus of political philosophy? Asking what sorts of social/political/economic institutions are best for a society to have? How a society's institutions influence the behaviour (and particularly moral behaviour) of its citizens? I don't know enough about the field to say. Perhaps these are not considered to be crucial questions. But I really do think they should be.

Though perhaps Matt is more thinking of smaller-scale issues (as suggested by his Wal-Mart vs McDonalds comparison), which are arguably nothing to do with government or society-at-large. Instead, we're more concerned with the few people responsible for the institution in question being what it is (assuming it's possible to divorce the institution in this way from the wider society of which it is part). If you focus on individuals in this way, it does look to be more a concern of ethics than politics. But there's surely some cross-over, so I wonder if political philosophy might still have something to offer here...?

Where's Einstein when you need him?

I recently read Brian Greene's The Elegant Universe (a really good introduction to modern physics, by the way, I highly recommend it). But there was one particular problem concerning special relativity which I don't understand: if light/photons don't move through time, then how did they get to the present? Shouldn't they all be frozen in that moment of time when they were formed?

Background: The key idea here is that any object is always moving (relative to some other object, I suppose) through space-time at the speed of light. For objects at rest (e.g. me, relative to my chair), all of that motion is going through the time dimension. However, for moving objects, some of that motion is "spent" on spatial dimensions, and so less remains to go through time. Just like if you travel north-east for 100m, you won't get as far north as if you travelled due north for that distance. I find this analogy quite powerful, because it explains (in a reasonably understandable sort of way) why it is that time slows down as objects approach the speed of light, and also why nothing can ever move through space at faster than lightspeed.

But what about light itself? A photon moves through space at the speed of light, so there is no motion leftover for it to move through the time dimension. Greene explicitly says that photons are no "older" now than they were at the birth of the universe. So how did those photons get into the present, if not by passing through time like everything else?

Some musings: I really haven't got much of a clue, so I'm hoping a more knowledgable reader will be able to enlighten me here. But I'll offer some thoughts anyway...
I'm guessing the problem must be due to me having too much of a naive-absolutist conception of space and time. Perhaps the apparent paradox can be solved by expressing space and time in purely relational terms.

I can't yet see how that helps, but to get a feel for it, let's consider the constancy of the speed of light:
If two cars, each travelling at 50 km/h (relative to an observer sitting on the footpath) are travelling in the same direction, then they are stationary relative to each other. If moving in opposite directions (e.g. about to have a head-on collision), then their relative speed is 100 km/h. All nice Newtonian stuff so far. But light is different. No matter how fast you're moving, in whatever direction, photons are always travelling at a constant speed c (the speed of light) relative to you (and everything else for that matter).

To apply the 4-d space-time stuff to the above examples:

• The observer sees all the cars as going at the same speed (50 km/h), so they all move through time at the same rate relative to him (which is very close to full-speed, but not quite).


• The same-direction cars are relatively stationary, so they move through time at full speed (i.e. slightly quicker than from the observer's perspective).


• The head-on cars are going relatively fast, so each moves through time relatively slowly (though the difference at this scale, compared to the speed of light, is so tiny as to be indistinguishable to humans).

So "how fast is the car moving?" is a nonsensical question unless you include the frame of reference (i.e. "how fast relative to X..."). Moreover, the car is moving through time at different rates, depending on who the observer is.

So what about photons then? No matter the frame of reference, their speed is always the same: c. So no matter the frame of reference, their movement through time is always the same: zero.

Well, that was no help.
Any ideas... anyone?

Update: see here for the answer.

Joogle

The Crooked-Timberites are Google-bombing for a good cause. Apparently the #1 google hit for "jew" is currently an anti-semitic site. If enough sites join in the linking to the wikipedia entry for Jew, this should soon be corrected...

Illusions and Zombies

Joshua at Blogosophy links to a really cool optical illusion, and offers some interesting commentary besides:

I think this kind of thing really undermines the plausibility argument for "zombies" (creatures indistinguishable from humans, but who lack qualia)--and plausibility is the only argument there is that it's a coherent notion. We think that we can imagine seeing the color red but having a completely different (in some hard-to-specify sense) feeling of what it's like to see red than someone else who also sees that color and tells us "yep, that's red." From this the argument goes that it's conceivable that someone could say "that's red" whenever we would, but who has nothing that it feels like to see red; there's just a bunch of nerves firing and chemicals changing, but nothing subjective going on. And if it's conceivable, so it is said, then it's possible.

I think, though, that it's harder (maybe even impossible) to imagine a zombie being fooled by this optical illusion; the illusion is, after all, precisely that you're losing the sense of what it is to feel like you're seeing one or more of the yellow dots, even though physically the photons are still hitting your eyes, the nerves are firing, etc. The lights are on, but you're intermittently not home to Mr. Yellow-dot Qualia. Is it conceivable that the zombie sees but isn't really aware of seeing the yellow dot, and also sees but isn't really aware of not seeing the yellow dot, and yet somehow still can distinguish objectively between the two states (so it can describe the illusion) just like someone with a mind? Or does your brain seize up in a kind of concept induced blindness trying to picture it?

I'm not entirely sure that I find his line of reasoning convincing however. It seems to me that any qualia-less 'zombie' would still have a sort of conscious/sub-conscious division to his mind. But by 'conscious' here it is only meant to refer to what information is available (i.e. directly accessible) to the central processing module (to use a computing analogy) of his mind. It is important to note that this is a distinct matter from whether phenomenological 'conscious experiences' (qualia) occur.

Given this distinction, it strikes me as entirely plausible that a 'zombie' could have sensory appartus detect the yellow dots, and yet have this information restricted to the lower sub-modules of his mind, so his central module would be entirely unaware of it. Thus a zombie could be just as fooled by the illusion as us, without any need for qualia.

P.S. For any readers with a basic knowledge of computer science, you may find it easier to follow this argument if you imagine a robot, with a specific function responsible for the parsing of perceptual input. This function would pick up all the raw data (including the light from the yellow dots), but the complicated parsing algorithm - fooled by the illusion - may fail to correctly interpret the raw data. Thus, when a different function (say, the speech one, if we want the robot to tell us what it sees) refers to this interpreted data, it will be unaware of any yellow dots.

So, yeah... it seems entirely plausible that a simple robot (let alone a full-blown zombie!) could be fooled by such illusions.

The 2-envelopes paradox

A really interesting math paradox was discussed at our university's "philosophy retreat" (at a lodge up in the mountains, it was very cool) a month or so ago. Crooked Timber posted a variation of it late last year. Anyway, the basic formulation is as follows:

Imagine you are offered a choice of two envelopes, and all you know is that one of them has twice as much money in as the other. So you pick one at random, but just as you are about to open it, you are given the option to swap envelopes, if you want.

Now, it should be obvious that there is nothing to be gained by swapping to the other random envelope. But the more you think about it, the more you begin to doubt this. After all, however much money you have at the moment, the other envelope either has half of that, or else double. Each possibility is equally likely, so by averaging them [ (2 + 1/2) /2 = 1.25] you realise that by swapping envelopes you will, on average, make a 25% profit!
This answer is clearly incorrect... but what is wrong with the math used to work it out?

Here's the solution I came up with...

First, formalise the scenario:
We have two envelopes, with values x and 2x (for some unknown x).
You take one at random. Call whatever value is in this envelope "y", and whatever is in the other one "z".

There is a 50% chance that you have the lesser envelope (y = x, z = 2x), so swapping in this case would double your money (z = 2y).
The other 50% chance is that you have the greater envelope (y = 2x, z = x), so swapping will halve your money (z = y/2).

It seems that based on the variable y, the expected value of swapping (to the other random envelope z) would yield a net increase of y/4 [E(z) = .5 * 2y + .5 * y/2 = 5/4 * y]. Yet this is clearly ridiculous (especially since you can do the same calculations on z to reach the opposite conclusion).

What has gone wrong?
The problem is that, within the E(z) calculation, the y in "2y" is different from the y in "y/2". That is, the variable y is being used to simultaneously represent two different values (x in one place, and 2x in the other).

Of course there is no problem with using a variable to represent some (single) unknown value (such is their purpose). BUT in this case the expected profit is calculated using two different instances of y simultaneously, whilst (deceptively) treating it as though it were a single variable (with a single value).

This inconsistency causes the bizarre results. The y that gets doubled is different from the y that gets halved. It is easy to lose sight of this crucial fact.

Note that there are only 2 unique numerical values involved in the entire scenario: x and 2x (all other variables (y and z) are equal to one or the other value).
Contrast this to sort of scenario insinuated by talking about having one envelope containing y (or x, or whatever you want to call it), and another which contains either half or double this... here you have the base case x, but also two other possible values, x/2 AND 2x.

This demonstrates a crucial difference between the original scenario, and this other (deceptively similar) implied scenario. To highlight this difference, consider the range of possible values:

1) in the original scenario, possible values range from x to 2x (so the maximum is double the minimum).

2) in the implied scenario, possible values range from x/2 to 2x (so the maximum is QUADRUPLE the minimum).

So how should we solve it instead?
There are several alternatives, but I think the most illuminating are probably the following two methods:

1) Substitute in x instead of y. Notice that when calculating the expected value of z, the "y/2" part is using y = 2x (so "y/2" can be replaced with "x"), whereas the "2y" bit is using y = x (so "2y" is to be replaced with "2x"). Substituting this gives us [E(z) = .5 * 2x + .5 * x = 1.5x], which is a zero profit, as expected.

2) Use "y" consistently.
To do this, we must calculate the expected initial value of y, before you work out the profits. This is easy enough: E(y) = 1.5x. Notice that the other envelope z does NOT possess "either half or double" of the expected value of y. The wording of the problem sort of implies this lie, which is why the problem sounds so confusing at first.

By using y consistently (i.e. by calculating its expected value), we can expose this flaw in the so-called "paradox". Solving for z gives E(z) = 1.5x also, so there is no gain to be made in swapping.

The math works... one just needs to be careful to apply it properly ;)

Update: I've just noticed that Matt Weiner posted about this paradox several times in January. A quick skim shows his answers to be a lot more complicated than mine... I'm not sure whether that means my solution here is insufficient, or if he's just missed the easy answer. Any ideas?

2nd Update: Crooked Timber is back into it!

Inclination vs Duty - which is better?

I came across the following problem on Jonathan Ichikawa's blog:

Here are three people, all of whom end up perfoming the same action. Which is, morally speaking, the best? A: Shopkeeper A is motivated solely by making money. He reasons that if he treats his customers fairly and is nice to them, they'll become repeat customers, recommend friends, etc. So he treats them fairly and is nice to them. B: Shopkeeper B just gets a kick out of making people happy. It makes him feel good to make other people feel good, so he treats his customers fairly and is nice to them. C: Shopkeeper C hates people. Also, he likes money, and is constantly tempted to cheat his customers. And maybe to kick them, too, because he'd like it if they experienced pain. But he knows that this would be morally wrong. So he treats his customers fairly and is nice to them, because it's his duty. I take it everyone will agree that A is less good than either B or C. But which of B and C is better?

The essence of the question is whether it is better to do good from inclination or from duty. Now, I don't know a huge amount about the history of ethics (so I'm probably grossly oversimplifying things, and I could be just plain wrong), but it seems to me that this is the essential divide between two distinct ethical traditions:

On the one hand, you have the Aristotlean notion of the "Good life", taken up by modern Virtue Ethicists, which focusses on moral character. By this conception (which I share), Shopkeeper B - who genuinely enjoys helping people - is the best of the three.

Alternatively, the Christian/Kantian tradition focusses on abstract obligations, and obedience to moral "rules" or "duties". This sort of view (which I detest) would imply that Shopkeeper C is the most moral, due to his selfless adherence to that which is morally required.

This scenario reminds me of one James Rachels described, which forcefully advocates inclination over duty. Imagine you are sick, and lying in hospital, when a friend comes by to visit. His visit cheers you up, and you thank him for coming, but he replies "oh, I'm just doing my duty" (perhaps he is a utilitarian and worked out that he could do nothing more productive with his time). At first you think he's just being modest, but further probing suggests that he really means it - he doesn't particularly want to see you, he just felt compelled to visit out of an abstract sense of obligation. Surely in this scenario, we feel that the visit of the "friend" has lost all its value?

Rather than being a perfectly virtuous moral agent, it seems to me that an uncaring, robot-like obeyer of duty is morally bankrupt. (I even have doubts as to whether Shopkeeper C is any better than A.) Having genuinely good desires strikes me as far better than merely acting morally out of duty. I'd choose Shopkeeper B any day.

Ideal Agent Theories

Jonathan Ichikawa wrote:

This is from Peter Railton, discussed yesterday in Jamie Dreier's metaethics class. Railton is looking to define a person's best interests. He suggests something like this: "It is in A's interest to do x in situation S just in case it would be in A+'s subjective interest to do x in situation S, where A+ is what A would be if he had unlimited cognitive and imaginative abilities."
...
So we attempt to patch up the account: "It is in A's interest to do x in situation S just in case it would be in A+'s subjective interest for A to do x in situation S." I'm not sure how to make sense of this formulation. The following looks like it might be a counterexample:
Bubba's IQ is 95. Bubba is confronted with the choice of whether or not to press the big red button. The big red button, if pressed, would have the following result: God will empty the bank account of every person whose IQ is higher than 100, and give the money to those whose IQ is lower than 100.
...
Jamie's response to this kind of suggestion was to push the idea that Bubba+ is the same person as Bubba. Well, ok... but he's still different in an important way. And it still seems like it would be in Bubba's interest to press the button, but not in Bubba+'s. So that looks like a problem. But I recognize that I'm pretty confused about the argument.

I've just been reading a bit of Railton myself, and it seems that Jonathan (or Jamie) has crucially misinterpreted what Railton was saying.

Railton emphasises that we are not interested in A+'s direct subjective interests (for precisely the reasons that Jonathan identifies) - that is, we do not ask what A+ would do in situation S, nor what action of A's is in A+'s interest. Both of these possibilities make the mistake of envisaging A+ as continuing to be a distinct individual from A, even after the decision is made. Instead, we ask A+ what he would want his non-idealised self A to seek were he (A+) to find himself in the actual condition and circumstances of A.

One way to think of this would be to consider A as temporarily gaining full cognitive powers (i.e. turning into A+), and being frozen in a moment of time until he makes a decision, whilst knowing that the moment the decision is made, he will be turned back into A. This ensures that A+ has motivation to seek what is in A's genuine interest, even in those cases when the apparent interests of A and A+ would otherwise diverge. Hopefully this way of conceptualising the situation may help to make sense of Jamie's suggestion that we think of A and A+ as being the same person.

Since A+ knows all of A's desires, and knows how to best realise them, and furthermore knows that he himself will soon share those desires and lose his own divergent ones (because he is about to turn back into A), it seems that this conception of the 'ideal agent' overcomes the problems Jonathan raises.

So to apply this to Jonathan's counterexamples:

A should still talk to other philosophers, the fact that A+ would gain nothing from it is irrelevant, for in making the decision, A+ knows that he will afterwards lose his omniscience, and so learning from others is still worthwhile.

Bubba should press the button. Bubba+ would choose this, because by the time the button is pressed, Bubba+ will have lost his omniscience (by turning back into Bubba), and so will reap the rewards of having an IQ less than 95.

I should probably clarify that I don't particularly like Ideal Agent theories. They can be useful as a sort of intuitive heuristic, but ultimately the reason for doing X is not merely that "A+ would choose X". Instead, surely, the reasons are those that are behind A+'s choice.

I think Railton actually agrees with me on this point, too. For he goes on to say that the "objectified subjective interest" of an agent (as identified through the Ideal Agent heuristic) can be 'reduced' to purely descriptive facts (eg about the agent, his underlying desires, and how best to realise them). It is then these facts, this 'reduced form', that constitute what Railton identifies as an Agent's "objective interest".

Things decided, not discovered

To return to moral philosophy, but in a slightly roundabout way, I first want to draw attention to that class of concepts which are ultimately decided upon by the relevant community, rather than discovered by scientists or philosophers.

The classic example is that of Theseus' ship, regarding the concept of identity:

There is an ancient puzzle concerning the ship of Theseus, a sacred wooden vessel of the Athenians. As the years passed, repairs saw more and more of the ship replaced by new components: one year half a dozen planks, the year before that a mast, and so on. Eventually a time was reached when the ship contained none of the original components. These originals, the tale goes, had been assiduously collected over the years, and as soon as the collection was complete the parts were reassembled into a (leaky) ship.

Which of the two, if either, is the original ship?
Suppose that when the ship was launched, she was granted free mooring rights at her home port in perpetuity: to which of the two vessels would this concession apply?

There is no absolute "right answer" - we can know all the facts, and still be at a loss as to which ship deserves the identity "Theseus' old ship". Ultimately, we (the linguistic community) simply need to decide for ourselves which definition would be most useful to us.

Prof. Jack Copeland discusses this in his book on Artifical Intelligence:

Our decision must not be arbitrary... in the case of the ship, the decision must take account of whether the purposes for which we need the concept same ship (eg for long-term contracts) are best served by taking the reconstructed ship, the repaired ship, or neither, to be the same ship as the one originally launched.

He goes on to observe that the question Can a machine think? poses a similar problem. The concept of 'intelligence' is as man-made, artificial and relative as the concept of identity, so we simply need to decide for ourselves what definition of 'intelligence' is most useful for our purposes.

Now, I'm beginning to think that morality may be another one of these artificial concepts. The universe itself doesn't care about what we do. So there are no absolute/objective moral truths just waiting out there for us to discover. Instead, we (collectively, not individually) have to decide for ourselves what we want moral concepts to refer to. That doesn't mean it's arbitrary, of course. The previous discussion of 'ought' suggests that any concepts which aspire to prescriptivity (as surely morality does) need to be closely linked to the fulfilment of human desires. And as discussed in Words & Meanings, terms become meaningless if used to refer to entirely different concepts than what the rest of us are using those words to talk about. Instead, as with Theseus' ship, we need to ask ourselves what understanding of moral concepts would be most useful to us.

This now raises the question: For what purposes do we need moral concepts?

Bridging the IS/OUGHT gap

To pick up from where I left off a couple of posts ago: What does 'ought' mean?
If we assume metaphysical naturalism, then nothing exists outside the world of 'is' (i.e. that which is open to empirical investigation). This implies that either normative concepts (such as 'ought') can be reduced to a purely descriptive form, or else those concepts do not refer to anything real.

Now this poses something of a problem for us, since the phenomenology of our experience suggests to us that normative facts (eg "you should not lie") are of an entirely different nature than descriptive facts (eg "snow is white"). However, Alonzo Fyfe suggested that this is simply because "ought" is a particular subset of "is", with noticable distinctions which set it apart from the typical subsets. Rather like whale is a type of mammal, despite being of an entirely different nature from, say, dogs.

So what type of real-world entity does "ought" refer to? From where does it get its apparent prescriptivity? What makes it different from normal descriptive facts like 'snow is white'? The easiest way to approach these questions is probably to forget about morality for a moment, and just consider "prudential oughts", or hypothetical imperatives.

Consider the following examples:

• John wants to watch The Simpsons, so what channel should ('ought') he set his TV to?
• You should leave now if you want to get to the movie on time.
• I've got a big exam tomorrow, so I really should study.

It should be clear that in each of these cases, the 'ought' (or 'should') is alerting us to what must be done in order to fulfill a particular desire. Sometimes (as in the final example) the desire may be implicit, but it nevertheless is crucial: unless I want to do well in the exam, there is no reason for me to study - it would no longer be true that I "ought" to.

In effect, "If you want Y, then you ought to do X", is just another way of saying "Doing X is such as to fulfill the desires in question" (where the desires in question are 'Y'). They mean the exact same thing (assuming a truth-conditional theory of meaning). Yet the second sentence is purely descriptive - it makes no mention of 'ought', it is purely within the world of 'is'. Thus we have successfully bridged the is/ought gap, at least for prudential 'oughts'.

Now we are in a position to answer where prescriptivity comes from - in a world of 'is', how is 'ought' even possible? The answer is found in human desires. The proposition "You ought to X" has no prescriptive properties in and of itself (the words are purely descriptive, as the translation in the previous paragraph demonstrates). Any prescriptivity is attached to the proposition by us, and the source of the prescriptivity is our desire-set.

The most widely accepted theory of human action is BDI theory, or Belief-Desire psychology. The core of this theory states that people always act so as to fulfill their (present) desires, according to their beliefs (about how to best achieve this). So if we gain a new belief about how to best fulfill our (present) desires, then we will (as a brute fact of human nature) act on that belief. To pre-empt the obvious counterexamples: If I continue to smoke despite knowing that it is bad for my health, that simply shows that my current desire to smoke is stronger than my current desire for future health benefits.

The end of all human action is the fulfillment of our desires. We use the word/concept 'ought' to bring attention to possible courses of action which (it is believed) will help attain that end. Desires are real, actions are real, and the relationship between them (i.e. "ought") is real. It is thus intrinsically descriptive in nature - the prescription gets attached only when human nature is added into the equation. Our desire set is precisely the list of prescriptions which govern our actions.

You might wonder if this makes the notion of 'ought' redundant: after all, if we always act to fulfill our desires anyway, what do we need this extra notion of 'ought' for? But note that we only act according to our current desires, whereas the concept of 'ought' can refer to the fulfilment of any desire (or set of desires). Thus in the above example of the smoker, we can sensibly say that although I might in fact smoke (because it fulfills my present desires), I 'ought' not to (with regard to the fulfilment of my future desires).

So prudential 'oughts' are part of the world of 'is' (the real world), and can be described purely in terms of real-word entities - such as desires and actions. Specifically, I claim that "You (prudentially) ought to do X", is truth-functionally equivalent to "Doing X is such as to fulfill the desires in question".

But what about morality? The moral 'ought' is somewhat different, since it is considered to (somehow) be binding on agents, regardless of their personal desires. The moral objectivist posits a notion of 'ought' which is supposedly independent of desires. Yet according to the above analysis, offering an 'ought' without a corresponding desire set (for the ought to be 'in regard to'), seems to be as incoherent as offering a location without a corresponding reference location (eg "the keys are above the table" makes sense, whereas "the keys are above" does not!).

This problem will be further examined in future posts...

[Update: this post is seriously flawed -- see my more recent posts on 'Empty Definitions' and 'Is Normativity Just Semantics?']