Puzzles for Physicalists

Link post

The fol­low­ing is a list of puz­zles that are hard to an­swer within a broadly-phys­i­cal­ist, ob­jec­tive paradigm. I be­lieve crit­i­cal agen­tial­ism can an­swer these bet­ter than com­pet­ing frame­works; in­deed, I de­vel­oped it through con­tem­pla­tion on these puz­zles, among oth­ers. This post will fo­cus on the ques­tions, though, rather than the an­swers. (Some of the an­swers can be found in the linked post)

In a sense what I have done is lo­cated “anoma­lies” rel­a­tive to stan­dard ac­counts, and con­cen­trated more at­ten­tion on these anoma­lies, at­tempt­ing to pro­duce a the­ory that ex­plains them, with­out rul­ing out its abil­ity to ex­plain those things the stan­dard ac­count already ex­plains well.


(This sec­tion would be philo­soph­i­cal pla­gia­rism if I didn’t cite On the Ori­gin of Ob­jects.)

In­dex­i­cals are phrases whose in­ter­pre­ta­tion de­pends on the speaker’s stand­point, such as “my phone” or “the dog over there”. It is of­ten nor­mal to treat in­dex­i­cals as a kind of short­hand: “my phone” is short­hand for “the phone be­long­ing to Jes­sica Tay­lor”, and “the dog over there” is short­hand for “the dog ex­ist­ing at co­or­di­nates 37.856570, −122.284176″. This ex­pan­sion al­lows in­dex­i­cals to be ac­counted for within an ob­jec­tive, stand­point-in­de­pen­dent frame.

How­ever, even these ex­panded refer­ences aren’t uni­ver­sally unique. In a very large uni­verse, there may be a twin Earth which also has a dog at co­or­di­nates 37.856570, −122.284176. As com­puter sci­en­tists will find ob­vi­ous, spec­i­fy­ing spa­cial co­or­di­nates re­quires a num­ber of bits log­a­r­ith­mic in the amount of space ad­dressed. Th­ese globally unique iden­ti­fiers get more and more un­wieldy the more space is ad­dressed.

Since we don’t ex­pand out refer­ences enough to be sure they’re globally unique, our use of them couldn’t de­pend on such global unique­ness. An ac­count­ing of how we re­fer to things, there­fore, can­not posit any causally-effec­tive stand­point-in­de­pen­dent frame that as­signs se­man­tics.

In­deed, the trou­ble of globally unique refer­ences can also be seen by study­ing physics it­self. Phys­i­cal causal­ity is spa­cially lo­cal; a par­ti­cle af­fects nearby par­ti­cles, and there’s a speed-of-light limi­ta­tion. For spa­cial refer­ences to be effec­tive (e.g. to con­nect to ob­ser­va­tion and ac­tion), they have to them­selves “move through” lo­cal space-and-time.

This is a bit like the prob­lem of hav­ing a com­puter re­fer to it­self. A com­puter may ad­dress com­put­ers by IP ad­dress. The IP ad­dress “” always refers to this com­puter. Th­ese refer­ences can be re­solved even with­out an In­ter­net con­nec­tion. It would be to­tally un­nec­es­sary and un­wieldy for a com­puter to re­fer to it­self (e.g. for the pur­pose of ac­cess­ing files) through a globally-unique IP ad­dress, re­solved through In­ter­net rout­ing.

Study­ing enough ex­am­ples like these (real and hy­po­thet­i­cal) leads to the con­clu­sion that in­dex­i­cal­ity (and more speci­fi­cally, deixis) are fun­da­men­tal, and that even spa­cial refer­ences that ap­pear to be globally unique are re­solved de­ic­ti­cally.

How does this re­late to physics? It means refer­ences to “the ob­jec­tive world” or “the phys­i­cal world” must also be re­solved in­dex­i­cally, from some stand­point. Pay­ing at­ten­tion to how these refer­ences are re­solved is crit­i­cal.

The ex­per­i­men­tal re­sults you see are the ones in front of you. You can’t see ex­per­i­men­tal re­sults that don’t, through spa­cio-tem­po­ral in­for­ma­tion flows, make it to you. Thus, refer­ences to the phys­i­cal which go through dis­cussing “the thing caus­ing ex­per­i­men­tal pre­dic­tions” or “the things ex­per­i­ments failed to falsify” are re­solved in a stand­point-de­pen­dent way.

It could be ar­gued that phys­i­cal law is stand­point-in­de­pen­dent, be­cause it is, sym­met­ri­cally, true at each point in space-time. How­ever, this ex­cludes vir­tual stand­points (e.g. ex­ist­ing in a com­puter simu­la­tion), and ad­di­tion­ally, this only means the laws are stand­point-in­de­pen­dent, not the con­tents of the world, the things de­scribed by the laws.

Pre-re­duc­tion references

(For pre­vi­ous work, see “Re­duc­tive Refr­er­ence”.)

In­dex­i­cal­ity by it­self un­der­mines view-from-nowhere mythol­ogy, but per­haps not phys­i­cal­ism it­self. What pre­sents a greater challenge for phys­i­cal­ism is the prob­lem of pre-re­duced refer­ences (which are them­selves de­ic­tic).

Let’s go back to the twin Earth thought ex­per­i­ment. Sup­pose we are in pre-chem­istry times. We still know about wa­ter. We know wa­ter through our in­ter­ac­tions with it. Later, chem­istry will find that wa­ter has a par­tic­u­lar chem­i­cal for­mula.

In pre-chem­istry times, it can­not be known whether the for­mula is H2O, XYZ, etc, and these for­mu­lae are barely sym­bol­i­cally mean­ingful. If we dis­cover that wa­ter is H2O, we will, af­ter-the-fact, define “wa­ter” to mean H2O; if we dis­cover that wa­ter is XYZ, we will, af­ter-the-fact, define “wa­ter” to mean XYZ.

Look­ing back, it’s clear that “wa­ter” has to be H2O, but this couldn’t have been clear at the time. Pre-chem­istry, “wa­ter” doesn’t yet have a phys­i­cal defi­ni­tion; a phys­i­cal defi­ni­tion is as­signed later, which ra­tio­nal­izes pre­vi­ous use of the word “wa­ter” into a phys­i­cal­ist paradigm.

A philo­soph­i­cal ac­count of re­duc­tion­ism needs to be able to dis­cuss how this hap­pens. To do this, it needs to be able to dis­cuss the on­tolog­i­cal sta­tus of en­tities such as “wa­ter” (pre-chem­istry) that do not yet have a phys­i­cal defi­ni­tion. In this in­ter­me­di­ate state, the philos­o­phy is talk­ing about two en­tities, pre-re­duced en­tities and physics, and con­sid­er­ing var­i­ous bridg­ings be­tween them. So the in­ter­me­di­ate state needs to con­tain en­tities that are not yet con­cep­tu­al­ized phys­i­cally.

A pos­si­ble phys­i­cal­ist ob­jec­tion is that, while it may be a pro­vi­sional truth that wa­ter is defi­ni­tion­ally the com­mon drink­able liquid found in rivers and so on, it is ul­ti­mately true that wa­ter is H20, and so phys­i­cal­ism is ul­ti­mately true. (This is very similar to the two truths doc­trine in Bud­dhism).

Now, ex­pand­ing out this ac­count needs to provide an ac­count of the re­la­tion be­tween pro­vi­sional and ul­ti­mate truth. Even if such an ac­count could be pro­vided, it would ap­pear that, in our cur­rent state, we must ac­cept it as pro­vi­sion­ally true that some men­tal en­tities (e.g. imag­i­na­tion) do not have phys­i­cal defi­ni­tions, since a good-enough ac­count has not yet been pro­vided. And we must have a philos­o­phy that can grap­ple with this pro­vi­sional state of af­fairs, and judge pos­si­ble bridg­ings as fit­ting/​un­fit­ting.

More­over, there has never been a time with­out pro­vi­sional defi­ni­tion. So this idea of ul­ti­mate truth func­tions as a sort of utopia, which is ei­ther never achieved, or is only achieved af­ter very great ad­vances in philos­o­phy, sci­ence, and so on. The jour­ney is, then, more im­por­tant than the des­ti­na­tion, and to even ap­proach the des­ti­na­tion, we need an on­tol­ogy that can de­scribe and us­ably func­tion within the jour­ney­ing pro­cess; this on­tol­ogy will con­tain pro­vi­sional defi­ni­tions.

The broader point here is that, even if we have the idea of “ul­ti­mate truth”, that idea isn’t mean­ingful (in terms of ob­ser­va­tions, ac­tions, imag­i­na­tions, etc) to a pro­vi­sional per­spec­tive, un­less some­how the pro­vi­sional per­spec­tive can con­cep­tu­al­ize the re­la­tion be­tween it­self and the ul­ti­mate truth. And, if the ul­ti­mate truth con­tains all pro­vi­sional truths (as is true if for­get­ting is not epistem­i­cally nor­ma­tive), the ul­ti­mate truth needs to con­cep­tu­al­ize this as well.

Epistemic sta­tus of physics

Con­sider the ques­tion: “Why should I be­lieve in physics?”. The con­ven­tional an­swer is: “Be­cause it pre­dicts ex­per­i­men­tal re­sults.” Some­one who can ob­serve these ex­per­i­men­tal re­sults can, thus, have epistemic jus­tifi­ca­tion for be­lief in physics.

This jus­tifi­ca­tory chain im­plies that there are cog­ni­tive ac­tors (such as per­sons or so­cial pro­cesses) that can do ex­per­i­ments and see ob­ser­va­tions. Th­ese ac­tors are there­fore, in a sense, agents.

A phys­i­cal­ist philo­soph­i­cal paradigm should be able to ac­count for epistemic jus­tifi­ca­tions of physics, else fails to self-rat­ify. So the paradigm needs to ac­count for ob­servers (and per­haps speci­fi­cally ac­tive ob­servers), who are the ones hav­ing epistemic jus­tifi­ca­tion for be­lief in physics.

Believ­ing in ob­servers leads to the typ­i­cal mind-body prob­lems. Dis­be­liev­ing in ob­servers fails to self-rat­ify. (When­ever a phys­i­cal­ist says “an ob­ser­va­tion is X phys­i­cal en­tity”, it can be asked why X counts as an ob­ser­va­tion of the sort that is epistem­i­cally com­pel­ling; the an­swer to this ques­tion must bridge the men­tal and the phys­i­cal, e.g. by say­ing the brain is where epistemic cog­ni­tion hap­pens. And say­ing “you know your ob­ser­va­tions are the things pro­cessed in this brain re­gion be­cause of physics” is cir­cu­lar.)

What mind-body prob­lems? There are plenty.


The an­thropic prin­ci­ple states, roughly, that epistemic agents must be­lieve that the uni­verse con­tains epistemic agents. Else, they would be­lieve them­selves not to ex­ist.

The lan­guage of physics, on its own, doesn’t have the ma­chin­ery to say what an ob­server is. Hence, an­throp­ics is a philo­soph­i­cal prob­lem.

The stan­dard way of think­ing about an­throp­ics (e.g. SSA/​SIA) is to con­sider the uni­verse from a view-from-nowhere, and then as­sume that “my” body is in some way sam­pled “ran­domly” from this viewed-from-nowhere uni­verse, such that I pro­ceed to get ob­ser­va­tions (e.g. vi­sual) from this body.

This is already pretty wonky. In­dex­i­cal­ity makes the view-from-nowhere prob­le­matic. And the idea that “I” am “ran­domly” placed into a body is a rather strange meta­physics (when and where does this event hap­pen?).

But per­haps the most crit­i­cal is­sue is that the phys­i­cal­ist an­thropic paradigm as­sumes it’s pos­si­ble to take a phys­i­cal de­scrip­tion of the uni­verse (e.g. as an equa­tion) and lo­cate ob­servers in it.

There are mul­ti­ple ways of con­sid­er­ing do­ing so, and per­haps the best is func­tion­al­ism, which will be dis­cussed later. How­ever, I’ll note that a sub­jec­tivist paradigm can eas­ily find at least one ob­server: I’m right here right now.

This re­quires some ex­plain­ing. Say you’re lost in an amuse­ment park. There are about two ways of think­ing about this:

  1. You don’t know where you are, but you know where the en­trance is.

  2. You don’t know where the en­trance is, but you know where you are.

Rel­a­tively speak­ing, 1 is an “ob­jec­tive” (rel­a­tively stand­point-in­de­pen­dent) an­swer, and 2 is a “sub­jec­tive” (rel­a­tively stand­point-de­pen­dent) an­swer.

2 has the in­tu­itive ad­van­tage that you can point to your­self, but not to the en­trance. This is be­cause point­ing is de­ic­tic.

Even while be­ing lost, you can still find your way around lo­cally. You might know where the Fer­ris wheel is, or the food stand, or your back­pack. And so you can make a lo­cal map, which has not been placed rel­a­tive to the en­trance. This map is us­able de­spite its dis­con­nec­tion from a global refer­ence frame.

An­throp­ics seems to be say­ing some­thing similar to (1). The idea is that I, ini­tially, don’t know “where I am” in the uni­verse. But, the de­ic­tic cri­tique ap­plies to an­throp­ics as it ap­plies to the amuse­ment park case. I know where I am, I’m right here. I know where the Earth is, it’s un­der me. And so on.

This way of lo­cat­ing (at least one) ob­server works in­de­pen­dent of abil­ity to pick out ob­servers given a phys­i­cal de­scrip­tion of the uni­verse. Rather than find­ing my­self rel­a­tive to physics, I find physics rel­a­tive to me.

Of course, the sub­jec­tivist frame­work has its own prob­lems, such as difficulty find­ing other ob­servers. So there is a puz­zle here.

Tool use and functionalism

Func­tion­al­ism is per­haps the cur­rent best an­swer as to how to lo­cate ob­servers in physics. Be­fore dis­cussing func­tion­al­ism, though, I’ll dis­cuss tools.

What’s a ham­mer? It’s a thing you can swing to ap­ply lots of force to some­thing at once. Ham­mers can be made of many phys­i­cal ma­te­ri­als, such as stone, iron, or wood. It’s about the func­tion, not the sub­stance.

The defi­ni­tion I gave refers to a “you” who can swing the ham­mer. Who is the “you”? Well, that’s stand­point-de­pen­dent. Some­one with­out arms can’t use a con­ven­tional ham­mer to ap­ply lots of force. The defi­ni­tion rel­a­tivizes to the po­ten­tial user. (Yes, a per­son with­out arms may say con­ven­tional ham­mers are ham­mers due to so­cial con­ven­tion, but this so­cial con­ven­tion is there be­cause con­ven­tional ham­mers work for most peo­ple, so it still rel­a­tivizes to a pop­u­la­tion.)

Let’s talk about func­tion­al­ism now. Func­tion­al­ism is based on the idea of mul­ti­ple re­al­iz­abil­ity: that a mind can be im­ple­mented on many differ­ent sub­strates. A mind is defined by its func­tions rather than its sub­strate. This idea is very fa­mil­iar to com­puter pro­gram­mers, who can hide im­ple­men­ta­tion de­tails be­hind an in­ter­face, and don’t need to care about hard­ware ar­chi­tec­ture for the most part.

This brings us back to tools. The defi­ni­tion I gave of “ham­mer” is an in­ter­face: it says how it can be used (and what effects it should cre­ate upon be­ing used).

What sort of func­tions does a mind have? Ob­ser­va­tion, pre­dic­tion, plan­ning, mod­el­ing, act­ing, and so on. Now, the mil­lion-dol­lar ques­tion: Who is (ac­tu­ally or po­ten­tially) us­ing it for these func­tions?

There are about three differ­ent an­swers to this:

  1. The mind it­self. I use my mind for func­tions in­clud­ing plan­ning and ob­ser­va­tion. It func­tions as a mind as long as I can use it this way.

  2. Some­one or some­thing else. A cor­po­ra­tion, a boss, a cus­tomer, the gov­ern­ment. Some­one or some­thing who wants to use an­other mind for some pur­pose.

  3. It’s ob­jec­tive. Things have func­tions or not in­de­pen­dent of the stand­point.

I’ll note that 1 and 2 are both stand­point-de­pen­dent, thus sub­jec­tivist. They can’t be used to lo­cate minds in physics; there would have to be some start­ing point, of hav­ing some­one/​some­thing in­tend­ing to use a mind for some­thing.

3 is in­ter­est­ing. How­ever, we now have a dis­anal­ogy from the ham­mer case, where we could iden­tify some po­ten­tial user. It’s also rather the­olog­i­cal, in say­ing the world has an ob­server-in­de­pen­dent telos. I find the the­olog­i­cal im­pli­ca­tions of func­tion­al­ism to be quite in­ter­est­ing and even in­spiring, but that still doesn’t help phys­i­cal­ism, be­cause phys­i­cal­ist on­tol­ogy doesn’t con­tain stand­point-in­de­pen­dent telos. We could, per­haps, say that phys­i­cal­ism plus the­ism yields ob­jec­tive func­tion­al­ism. And this re­quires adding a com­po­nent be­yond the phys­i­cal equa­tion of the uni­verse, if we wish to find ob­servers in it.

Causal­ity ver­sus logic

Causal­ity con­tains the idea that things “could” go one way or an­other. Else, causal claims re­duce to claims about state; there wouldn’t be a differ­ence be­tween “if X, then Y” and “X causes Y”.

Pear­lian causal­ity makes this ex­plicit; causal re­la­tions are defined in terms of in­ter­ven­tions, which come from out­side the causal net­work it­self.

The on­tol­ogy of physics it­self is causal. It is as­serted, not just that some state will definitely fol­low some pre­vi­ous state, but that there are dy­nam­ics that push pre­vi­ous states to new states, in a nec­es­sary way. (This is clear in the case of dy­nam­i­cal sys­tems)

In­deed, since ex­per­i­ments may be thought of as in­ter­ven­tions, it is en­tirely sen­si­ble that a phys­i­cal the­ory that pre­dicts the re­sults of these in­ter­ven­tions must be causal.

Th­ese “coulds” have a difficult sta­tus in re­la­tion to logic. Some­one who already knows the ini­tial state of a sys­tem can log­i­cally de­duce its even­tual state. To them, there is in­evita­bil­ity, and no log­i­cally pos­si­ble al­ter­na­tive.

It ap­pears that, while “could”s ex­ist from the stand­point of an ex­per­i­menter, they do not ex­ist from the stand­point of some­one ca­pa­ble of pre­dict­ing the ex­per­i­menter, such as Laplace’s de­mon.

This is not much of a prob­lem if we’ve already ac­cepted fun­da­men­tal deixis and re­jected the view-from-nowhere. But it is a prob­lem for those who haven’t.

Try­ing to de­rive de­ci­sion-the­o­retic causal­ity from phys­i­cal causal­ity re­sults in causal de­ci­sion the­ory, which is known to have a num­ber of bugs, due to its re­li­ance on hy­po­thet­i­cal ex­tra-phys­i­cal in­ter­ven­tions.

An al­ter­na­tive is to try to de­velop a the­ory of “log­i­cal causal­ity”, by which some log­i­cal facts (such as “the out­put of my de­ci­sion pro­cess”, as­sum­ing you know your source code) can cause oth­ers. How­ever, this is oxy­moronic, be­cause logic does not con­tain the af­for­dance for in­ter­ven­tion. Logic con­tains the af­for­dance for con­struct­ing and check­ing proofs. It does not con­tain the af­for­dance for caus­ing 3+4 to equal 8. A suffi­ciently good rea­soner can im­me­di­ately see that “3+4=8” runs into con­tra­dic­tion; there is no way to con­struct a pos­si­ble world in which 3+4=8.

Hence, it is hard to say that “coulds” ex­ist in a stand­point-in­de­pen­dent way. We may, then, ac­cept stand­point-de­pen­dence of cau­sa­tion (as I do), or re­ject cau­sa­tion en­tirely.


My claim isn’t that phys­i­cal­ism is false, or that there don’t ex­ist phys­i­cal­ist an­swers to these puz­zles. My claim, rather, is that these puz­zles are at least some­what difficult, and that suffi­cient con­tem­pla­tion on them will desta­bi­lize many forms of phys­i­cal­ism. The cur­rent way I an­swer these puz­zles is through a crit­i­cal agen­tial frame­work, but other ways of an­swer­ing them are pos­si­ble as well.