I worked at SF in an admin role a while back. Everyone liked Jim. He paid for employee vacations, spoke at employee meetings, greeted you in the elevator. You knew at some level you worked for him, not your boss. It was almost like playing. It was interesting and fun. I always described him to friends as the one good billionaire. Unfortunately new leadership seems determined to port over the worst most calcified features of the academy- like a trauma response. It's not fun anymore. It's risk management all the way down. RIP Jim.
Thank you, no stake or interest in these people or things whatsoever, but it just proves any good profile of any interesting person is itself interesting (and good for me, since I've been writing multiple entries about one such person and interview, myself).
"He too was intellectually active into his 80s, still robust, still contributing. The kind of man who didn’t treasure his body in the effete way elites do now, so afraid to smoke, so afraid to drink, so afraid of not taking the right supplements. The older class both men belonged to viewed the body as a thing to be used up, and used up well, by the end of it all."
I've been thinking about this sort of thing a lot. I'm about to turn 40, most younger people still seem to think I'm around 30. Shunryu Suzuki once said something like, "In your practice, you should consume yourself like a flame until it's all used up."
I think it's more about not becoming trapped in your (we'd say in Zen), "small-Mind," version of yourself--don't NOT take care of yourself and avoid unnecessary risks, but the idea of preserving every aspect of yourself totally deludes yourself into thinking of yourself as a separate being with no connection to the totality of a society and an ecosystem, not to mention the great numbers of people breathing in massive amounts of air pollution every day, or children in Asia and Africa scourging piles of garbage for heavy metals to sell on a daily basis. Another grotesque inequality.
The Universal Approximation Theorem history (fascinating) reminds me of Dan Dennett (RIP dear thinker I’ve been out touting): “don’t mistake a failure of imagination for an insight into necessity” BOOOOOM 🤯
You have touched on something I am curious about. I paraphrase your belief as follows: "The macro-scale description contains information that is lacking from the micro-scale fundamental physics description."
I have a question and a comment.
1) Would you agree with my paraphrase? It seems to me to be trivial to show it is false since the micro-scale description is sufficient to calculate the macro-scale description so it is wrong to say the micro-scale description lacks any information. I assume there is some subtlety I am missing.
2) Your belief reminds me of many of Douglas Hofstadter's writings. I am thinking in particular of his thought experiment where he builds a computer out of dominos that can take a number as input and as an output tell you if that number was prime. Let's say you input "641", encoded as a certain configuration of dominos, topple the dominos to set off the reaction, and in the end, indeed find that the domino which corresponds to "prime" has fallen. Hofstadter asks "Why did that last domino fall?" The micro-scale explanation is that this domino hit that domino which then hit that domino... and so on. On a more macro-scale, though, the explanation is simply "Because 641 is prime". This is much simpler and seems to be, in some deep way, the truer explanation.
And finally, to sum up my comment in general, I think this comes down to some Godelian nature of reality wherein there is no fundamental level and our job is to intuit the best level of explanation for a given phenomena.
Since you're asking specifically, I actually wouldn't paraphrase it like that. In the cases of examples we give, we universally assume supervenience holds. That's just a fancy way of saying we assume that you can always derive the macroscale from the microscale, if you have full information about the microscale - exactly what you're saying is true. We take that as a background assumption that you could always do such a calculation. But the causal relationships are still stronger/better/more reliable at the macroscale (whatever terminology is appropriate to the measure of causation you're using, since there's a bunch). You can still calculate the macroscale out from the microscale (by, e.g., making a macroscale model) but that model you make will have less errors/uncertainty in the causal relationships. Agree it's related to that Hofstadter idea, btw, that's a nice reference.
Wow! That’s really helpful. When you write about emergence, I was always thinking of something similar to harry’s paraphrase. Does it make sense to say that causal relationships are more reliable at the macro scale because there’s too much information to process at the microscale? I’m obviously not an academic.
Thank you for the lucid explanation and thank you for cluing me into the technical term "supervenience" as that will help me cognitively in my investigations in this arena, e.g consciousness, reductionism, etc.
I have been very much enjoying the Intrinsic Perspective. A moment of context about me: I am an animist.
People largely talk about the EMH with Simons. But he also evaded the trap which says data mining is dangerous. Finance professors have dismissed evidence of profitable trades as spurious results from torturing data going back to the 1970s. I work in academic finance and the dominant view is still that data mining doesn't work. Simons proves that wrong too, on top of proving wrong the EMH.
All of this is ironic because in the end it seems like much of academic finance is naive data mining. In a recent paper we show that the out-of-sample performance of academic and data-mined strategies are eerily similar: https://arxiv.org/pdf/2212.10317
Hi Eric, I haven't read your paper, and I intend to, but I thought I'd better act quick and ask you, before you move on to another post: What do you think of Terrence Deacon's hypothesis that what makes something emergent are the constraints on the parts (both spatial and temporal constraints). These can't be reduced, since they are about something not there at the microscale. So in your case, a macro system is constrained by some average (coarse grained) quantity like energy, pressure, volume, density, or some lower dimensional attractor in phase space, and describing it using those quantities is not just more parsimonious, but also more predictive (since the microscale description does not include the constraints) than describing it without the constraints. It's not tautological, constraints are not always present, at least not in terms of average quantities, and only when they are present does emergence happen.
So I read Deacon's Incomplete Nature, and I do think he has some interesting ideas. With that said, I also think those ideas are very difficult to understand because (a) there's a lot of jargon, and (b) he uses very complex examples from physics without actually working them out. E.g., the kind of "I'm in a coffeeshop and the waitress pours me a coffee and there's a pattern in the foam, etc etc." I never use examples like that, because they are impossible to work through and easy to get confused about. He's not the only one guilty of this, it's basically everyone who has anything to say about emergence, to be honest, with only a couple exceptions (and it's why the philosophy literature is a mess, it's nowhere near rigorous and mathematical enough). But in Deacon's case I think the combination of jargon and no simple modeling makes the clarity less than ideal, not just for an outsider trying to understand, but also for Deacon himself. Although I like your summary of it!
One example of a confusion I found regularly in Incomplete Nature: a distinction has to be made about whether he means parts vs. wholes or macro vs. micro. These are a bit different. For instance, let's say you have an XOR gate. You can't tell the future of the XOR gate without knowing both inputs. So they constrain it in a way irreducible to the individual parts, since if you only knew one input, you would have zero information about the XOR's next state. But this is just a matter of joint causation, which people often confuse for emergence. Joint causation is much better understood, I think integrated information theory is probably the best take on it but there are plenty of others, like the partial information decomposition or so on. Emergence, which is asking what's special about macroscale descriptions (which are in turn just dimension reductions of microscale descriptions) is less well-accepted and more difficult to understand.
However, if there is a connection between Deacon's views and mine, it is this: if you want, you can rephrase causation as constraint. So rather than saying A causes B to shift into state X, you can say that A constrains the future of B into one state, X. So you could use the framework of causal emergence to say that the macroscale constrains the future to a greater degree than the microscale of the system. This wouldn't necessarily involve any particular physical quantities, however, you can get the result in just like Markov chains. I'm unaware of any other research on the causation/constraint connection than my own, however.
Thanks so much for that! I will think about it and perhaps have some more questions that follow. I love your paper also. If you don't mind, a few more more Deacon-related questions:
What do you think about his attempt to extend information theory beyond Shannon to semantics (signifiers, interpretation, meaning and all that)? Are effective IT and Bayesian theories missing that? Or is all of that just a trivial coupling of Shannon with evolutionary algorithms (or some other utility-maximizing algorithm like Levin and Watson's "natural induction")?
The exclusion argument seems like obvious nonsense to me, which makes me suspect I'm misunderstanding it. Wouldn't it equally prove the pointlessness of studying aerodynamics, since fluid flow is a bulk phenomenon and doesn't really exist when you consider individual molecules?
So it's not that some macroscale property P is non-existent in the *individual* microscale parts (so in this case, just to go along with your example, the fluid flow of individual molecules considered alone or by themselves isn't a sensible quality you could calculate). It's rather that the same property (here I guess, fluid flow) could be described in terms of the molecules collectively together... without ever describing the fluid at some macroscale (like zooming out and dimension-reducing the individual molecules into some larger thing by not mentioning specific configurations of molecules, etc).
As so delightfully often in your commentary, Eric, there are far more numerous and interesting theses touched on in your post than can be addressed in a single comment box.
As someone who works seriously with concepts of emergence in society and historical process at the large scale, and with such concepts intellectually in models of physics and neurocognitive organization, I'm drawn to your struggles with the nullity of even much of informed theorization on emergent processes, which you touch on here. Your theory of causal emergence strikes me as a good one. I would tend to say any macroscale organization tends to mode lock activity to its particulars, with a stability gain resulting at that particular scale of order which is not necessarily implied by organization at any other scale of order. 'Error correction' is one result, though that is to give the process an eschatological reading which is perhaps not necessary to the basic behavioral outcome (where 'error' and 'correction' may not be inherently meaningful terms for the resultant organization).
I would say that there is a more fundamental logical and evidentiary failure to the Causal Exclusion Argument. This proposition requires as an a priori that there be some scale of organization which is formally basic . . . but "there ain't no such animal." An argument that some microscale organization predicates all macroscale organization _requires_ that elements at that microscale be irreducible, whether atoms, quarks, cells, genes, human decisional minds, individual market transactions, or so on. There are no such 'basic orders.' All, when examined closely, have levels of behavior of finer composition yet. Now, organization at whatever scale of order one likes may be highly stable and generally predictable (except of course for the numerous and evident anomalies which said predictions cannot explain and by bodily revulsion ignore). If there is no 'basic scale of order,' there can be no microscale supervenience that is provable, nor in fact likely any that exists at all. It's "emergence all the way down" in my view, a perspective which I've pursued myself in multiple lines of theoretical argument.
The absence of formally basic order is so profoundly evident in any actual field of research investigation that causal exclusion should be laughed off the podium. It isn't, and I would posit isn't for the simple reason that most minds from their own psychological necessity _demand_ that there exist some basic elements "which are true." It's not the putative probabilistic nature of reality which gets them, it's the approximate coherence of anything and everything. Most disciplines in philosophy demand some basic set of a prioris . . . and impose one where it does not exist, effectively if often surreptitiously falsifying themselves from the outset. There are a few lines of argument in philosophy which argue conversely for absolute indeterminism, but that's not much fun as it means you can't really 'prove' anything, and so be smarter than the next simian. One beauty of emergence as a concept is that it bridges that false dichotomy. 'Facts' are order- and organization-specific, not universals from this perspective. At the same time, mutability and inconstancy are inherent to systemic behavior, yet they are seldom truly and pervasively random as opposed to being shot through with organization, thresholds, and boundary constraints (with organizational components which are order-specific stable to a degree). Emergence DEMANDS an entirely new approach to philosophy, but for that matter to science and history as well, not least because of how emergence reformulates what 'causality' can and does mean, in my view.
You touch on interesting points regarding the declining complexity of poetry, Eric, and gatekeeping hostility disguised as indifference cloaked in 'objectivity' to any kind of original thinking at all. Just in reading the application guidelines for 'genius grant' foundations, it has struck me that their process is profoundly antithetical to any truly interesting theory or modeling, much as you found. From that perspective, Jim Simon would never have secured a grant from his own foundation either, other than perhaps by personal charisma in the interview or prior reputation (because that is what drives something like 70% of the yes in granting, as I see it). Perhaps I'll find time to opine on these issues as well.
Great comment. The argument you're making against the Causal Exclusion Argument is definitely a good one. Ned Block also makes it in "Do causal powers drain away?" with the idea that, if it really is "bottomless" in that you can keep dividing things smaller and smaller, all notion of causation vanishes, since it's forever "draining away" at each microscale. I think it's an interesting way out of it, but at the same time requires that assumption that there is no basement scale to physics (many physicists might disagree). Additionally, like any reductio, there's always the option of just biting the bullet. In other words, it does away with the Causal Exclusion Argument, but you're also giving up causation in general. I think maybe there's a way to save that (it sounds like you'd be sympathetic to this) which is to say there are no "privileged scales" in the situation where everything is forever draining away. But this path, at most, nets you a sort of relativism about what causes what, rather than clear-cut cases of emergence vs. reduction.
I'm sure my conception here is not original or unique to me, and it is interesting to hear how others approach Causal Exclusion. I have spent literally years working to build up a frame of reference for causal process in history where emergence is intrinsic. I have so much work to do already, I'm beginning to think I won't live long enough to get to that particular, treasured, book.
I would tend to say that 'causation is local.' There is no need to see causation as draining away: what fails is any _global_ causation. Causation is local to scale of order, and often if to lesser extent specific to system of organization.
I in fact have a toy model of general physics in which order scales in a profound way, with causal relationships and order specific to varying shells-tiers of dimensional organization and complexity. I'd love to see someone competent mathematize it---I can't---because I think it's the way out of the dead end into which physics has particle-ized itself. Yes, most physicists would and will hate it, but it's a way around the sense that there is no causal validity at all, just as its a way out of the trap that there is only a single, universal causal relationship set as opposed to a scaled set of relations only some of which are accessible at a given scale of order.
I'd love to discuss some of this with you, Eric, and other issues as well. I've wanted to send you drafts of my text as I think aspects of it would interest you. If that's to your liking, please let me know a path to contact you outside of this. Should you be inclined to check out my approach first, I comment over at The Jagged Spend.
Interesting piece on an interesting person. In regards to what you share around causal emergence, and the notion of it acting like an error correction at the macro scale. I have a question and a though:
Q: There seems to be some inconsistency with the idea that a fully known micro-state (ie. "If everything is accounted for in the microscale atomic-level descriptions") would not also include, theoretically, the causal effects without uncertainty and probability-based approximation. It seems like it still comes down to whether, at the most fundamental level, there is a deterministic or stochastic system at play in the universe. This conflates the idea that while we may theoretically be able to predict future states in chaotic systems, we do not (currently or possibly ever) have enough precision on the current state at the micro level to do so. I think this lead to a dual interpretation of your work to suggest that there is a stochastic system at the micro level, and we then perceive a deterministic system when we operate at any macro level.
Thought: Your analogy of this concept seems like it would equally suit a description of moving between an analog system (real world) and a discrete one (computers). Of course, that is also based on the underlying aspects of information theory as well, but might be a little more intuitive to people. What see in the real world and what you get in the pixels of a photograph is common interface for ADC.
I really enjoyed your use of error correction analogy and information theory.
Very interesting and well written piece as always. It gave me a lot to think about.
I will say however, that the efficient market hypothesis is one of those outdated bits of junk economics that just never died in the public consciousness. It's analogous to those little experiments where you place a jar on the table and have a room guess how many jelly beans are inside. In this case it often turns out that the average guess is more accurate than most or any individual guess, and if you keep repeating this experiment, no one would ever beat the average over time.
Well, this is exactly how the market works except for one thing... when emotion is involved, it knocks the accuracy completely out of whack. The crowd is quite irrational, as are the individuals that make it up, so a well tempered and logical person has a decent chance of beating it. The market is not driven by a cold assessment of available information, as the EMH assumes, but by fear and greed. Unfortunately, it is extremely difficult to distinguish oneself from the fervour of the masses.
‘Later, I think it was Simon DeDeo who presented a talk to all of us, including Jim, on the fall of complexity in poetry. It turns out that if you measure the entropy of published poems they’ve been declining as far back as back goes…’
Yup, that's what I remember from the talk. I forget what measure they used, it was some sort of bag-of-words calculation. But yeah, poetry is supposedly more predictable than it used to be in the past. Another way of saying it uses a smaller vocabulary, by the way, just to give some intuitions (it's easier to guess the next word if there are fewer words used in general).
Coming from the poetry trenches, I would tend to agree without seeing the analysis. Lots of people that sound like a single fairly anonymous collective poet. I tend to think it's related to the increasing prevalence of MFA programs and the academic professionalization of poets, at least in the US. Or maybe I'm hoping that it's that, and not some basic erosion of culture and creativity in the species.
I would love to read the methodology used by Simon DeDeo, and especially the timeframe of observation. It seems to me I've run across similar conclusions before, and I agree with the overall argument: complexity in poetry has at least a modern vector of continuous decline. As someone who has long written complex poetry, I can attest that a) it is impossible to other than self-publish it, and b) published poetry is vastly less complex now even than when I began.
I think there are three reasons for this trend of result:
1) The Well-trodden Path. When few write or publish poetry, the possibility of divergent results are greater, and hence aggregate complexity is inherently likely to be higher. When many publish, even if they are in principle original, the probability of convergence of expression goes up, and aggregate complexity necessarily declines.
2) Conformity Bias. Originality is always possible, but social groups have strong expectations of conformity, or at least of not standing out. As more publish, the numbers who herd together in similar vocabulary, diction, and construction go up, by psychological osmotic pressure, as it were, even more than by intention. In that context, aggregate complexity is simply bound to decline, even if outliers in expression are as frequent as ever (which they are not).
3) Suffocation Under the Bell Curve. Outliers in capacity are few. The numbers (and hence the money) are in the fat part of the bell curve. Low complexity is simply more probable in the middle of the distribution, so that as the total numbers increase, complexity is inherently driven down, even if everyone is 'original' in their creative result.
I would love it if the mass popular production of 'poetry' we see now yielded the beauty and the complexity of the ages . . . but it does nothing of the sort, and IS nothing of the sort. To say more than that would be to run off on a rant; so I shan't.
I made a note to self to google this. Is it not published? Fascinating. To my pale understanding, information entropy is the distance from uniformity of a probability distribution. Poetry has become closer to Random words? By the way I must tout w3w.com at showing how 3 words are like a personal space upon our vast Earth. My office is at quietly.divide.raindrops and my parents are buried near spare.person.parts. Coincidence? Guaranteed, yet it sticks in the brain.
I’ve never heard of the exclusion argument in the philosophy of mind being used to argue against the significance of macro-scale behavior, that would be bizarre indeed! As far as I know, it’s always used as an argument in support of physicalism. The argument being that since the macroscale behavior of neurons supervenes on the microscale, you can’t change the behavior of neurons through some non-physical effect without contradicting physics. That’s what the wiki article also suggests in the link you provided.
I would be very surprised if any serious physicalist argued that macroscale neuronal behavior is not an appropriate level of scale to study, or that cognitive science is not meaningful since it can’t contribute anything beyond physics. But perhaps you’re familiar with a section of the community that I’m not?
It's an inevitable consequence of the argument that it generalizes. This is the crux of Ned Block's reply to the causal exclusion argument, "Do causal powers drain away?" which is probably the most popular piece of the literature besides the argument itself (if it were only about minds, causation wouldn't "drain away"). There's plenty of other admittances of this in the literature, e.g., like Bontly's "The supervenience argument generalizes."
I wasn’t disputing whether causal exclusion generalizes to other cases. I was saying that even in the case of the mental, the notion that causal exclusion entails that there is no mental causation seems to me false. Rather it entails that there is no mental causation iff we adopt a non-physicalist theory of mind. The argument as I understand is just that causation doesn’t go over and above its causal parts. Whether we want to say that this rules out higher-level causation depends on your terminology, and whether you think of higher-level causation as needing to be something non-reductive. In other words, I take Jaegwon Kim’s argument to be saying that we either have to accept epiphenomenalism or reductionism about the mental (mental states just are physical states).
Hence Kim’s argument is compatible with mental causation, provided we accept reductionism. It’s purely an argument against interactionism. If this is true, then it doesn’t matter whether the argument generalizes, since even in the case of the mental it doesn’t say that there is no mental causation.
I read Block’s piece but am unfamiliar with Bontly. Block gives the interesting argument that if there is no fundamental level of physics, then were would be no causation according to the CEP, since every causal action is grounded in its parts. But even if he’s right and the possibility of no fundamental level argues against CEP, I don’t see how that justifies your assertion in this article that proponents of CEP think there is no mental causation, or that thinking about macro-scale causation is a waste of time since everything happens at the micro scale anyway.
“I don’t see how that justifies your assertion in this article that proponents of CEP think there is no mental causation”
To be clear I meant that I don’t see how it justifies your assertion that CEP proponents think there is no higher level causal emergent behavior, which of course would exclude mental causation. If even in the mental case CEP proponents think there can be mental causation on reductionism, then obviously they will think there can be higher level causation in general cases as well.
Unless of course you’ve been talking about strong emergence all along, where the lower level behavior of atoms is not actually exhausted by the laws of physics. But if I understand you correctly, that’s not what you mean.
Just to see if it clears anything up: The issues have nothing to do with mental causation at all, if that helps. That’s where some of the literature originated but I’d set all mental causation stuff aside entirely. And understanding of “every causal action is grounded in its parts” is not how I’d phrase it, exactly. I’d phrase it as about different levels of description for the causation that does occur (this turns out to be different than “it’s in the parts,” e.g., a macroscale can have parts, they’re just macroscale parts).
Thanks. I agree that your levels characterization is better. Do you agree that a micro scale quantum description of some event won’t miss anything that a macro scale description would? In other words, that the behavior entailed by the emergent efficiencies (among other things) which pop up at macro scales would be predicted by the micro scale description? If no, then wouldn’t this just plainly contradict physics? If yes, then I’m not sure what you think CEP proponents are disputing when you say they think the macroscale doesn’t ‘add’ anything.
The notion of emergence manifesting somewhere between the micro- and macroscales is most interesting, particularly if we think about it in the context of artificial intelligence. It strikes me that, intuitively if not formally, a great number of the most temporally optimistic theories of AI progress (i.e. those that believe we’re closer to AGI rather than further) seem to hope-against-hope that scale will deliver the manifestation you speak of, wherein a non-intelligent ‘micro’ function turns by some miracle of aggregation into an intelligent and cohesive ‘macro’ function. Puts a lot of AI scientists in league with acolytes of the Upanishads and adherents of panpsychism, as far as I can see, relying on the aforementioned ‘magic’ to conjure the emergent property. I can imagine, however, the potential of a kind of science of ‘emergence dynamics’ to determine how multifarious interactions between different elements at the microscale not only produce the adjustments you speak of but fundamentally attune such various inputs as combine in extremely complex processes like the generation of consciousness.
Also interesting your experience applying to be a Simons fellow – it seems as per your account that the people in charge of assessing your application were an almost perfect dispositional inverse of the fellowship’s namesake (or perhaps it’s truer to say, if they seemed clocked out during your presentation and Jim was also prone to nod off during an address, they might in fact have emulated him too thoroughly!). Knowing how to build an institution in the image of a unique founding figure really does seem like the most delicate art in the world. Do you know if the foundation still has any comparable/loosely-comparable figures of genius/high-interest doing work there now?
I worked at SF in an admin role a while back. Everyone liked Jim. He paid for employee vacations, spoke at employee meetings, greeted you in the elevator. You knew at some level you worked for him, not your boss. It was almost like playing. It was interesting and fun. I always described him to friends as the one good billionaire. Unfortunately new leadership seems determined to port over the worst most calcified features of the academy- like a trauma response. It's not fun anymore. It's risk management all the way down. RIP Jim.
A nice insight into the actual workings, thank you. It will be interesting to see if the historic returns continue without Jim.
Thank you, no stake or interest in these people or things whatsoever, but it just proves any good profile of any interesting person is itself interesting (and good for me, since I've been writing multiple entries about one such person and interview, myself).
"He too was intellectually active into his 80s, still robust, still contributing. The kind of man who didn’t treasure his body in the effete way elites do now, so afraid to smoke, so afraid to drink, so afraid of not taking the right supplements. The older class both men belonged to viewed the body as a thing to be used up, and used up well, by the end of it all."
I've been thinking about this sort of thing a lot. I'm about to turn 40, most younger people still seem to think I'm around 30. Shunryu Suzuki once said something like, "In your practice, you should consume yourself like a flame until it's all used up."
I think it's more about not becoming trapped in your (we'd say in Zen), "small-Mind," version of yourself--don't NOT take care of yourself and avoid unnecessary risks, but the idea of preserving every aspect of yourself totally deludes yourself into thinking of yourself as a separate being with no connection to the totality of a society and an ecosystem, not to mention the great numbers of people breathing in massive amounts of air pollution every day, or children in Asia and Africa scourging piles of garbage for heavy metals to sell on a daily basis. Another grotesque inequality.
Markets are not efficient; they are made efficient by people like him. See the “Grossman Stiglizt paradox”
The Universal Approximation Theorem history (fascinating) reminds me of Dan Dennett (RIP dear thinker I’ve been out touting): “don’t mistake a failure of imagination for an insight into necessity” BOOOOOM 🤯
You have touched on something I am curious about. I paraphrase your belief as follows: "The macro-scale description contains information that is lacking from the micro-scale fundamental physics description."
I have a question and a comment.
1) Would you agree with my paraphrase? It seems to me to be trivial to show it is false since the micro-scale description is sufficient to calculate the macro-scale description so it is wrong to say the micro-scale description lacks any information. I assume there is some subtlety I am missing.
2) Your belief reminds me of many of Douglas Hofstadter's writings. I am thinking in particular of his thought experiment where he builds a computer out of dominos that can take a number as input and as an output tell you if that number was prime. Let's say you input "641", encoded as a certain configuration of dominos, topple the dominos to set off the reaction, and in the end, indeed find that the domino which corresponds to "prime" has fallen. Hofstadter asks "Why did that last domino fall?" The micro-scale explanation is that this domino hit that domino which then hit that domino... and so on. On a more macro-scale, though, the explanation is simply "Because 641 is prime". This is much simpler and seems to be, in some deep way, the truer explanation.
And finally, to sum up my comment in general, I think this comes down to some Godelian nature of reality wherein there is no fundamental level and our job is to intuit the best level of explanation for a given phenomena.
Since you're asking specifically, I actually wouldn't paraphrase it like that. In the cases of examples we give, we universally assume supervenience holds. That's just a fancy way of saying we assume that you can always derive the macroscale from the microscale, if you have full information about the microscale - exactly what you're saying is true. We take that as a background assumption that you could always do such a calculation. But the causal relationships are still stronger/better/more reliable at the macroscale (whatever terminology is appropriate to the measure of causation you're using, since there's a bunch). You can still calculate the macroscale out from the microscale (by, e.g., making a macroscale model) but that model you make will have less errors/uncertainty in the causal relationships. Agree it's related to that Hofstadter idea, btw, that's a nice reference.
Wow! That’s really helpful. When you write about emergence, I was always thinking of something similar to harry’s paraphrase. Does it make sense to say that causal relationships are more reliable at the macro scale because there’s too much information to process at the microscale? I’m obviously not an academic.
Oh wow this caused something to click for me that I’ve been struggling with for months. Thank you!!!
Thank you for the lucid explanation and thank you for cluing me into the technical term "supervenience" as that will help me cognitively in my investigations in this arena, e.g consciousness, reductionism, etc.
I have been very much enjoying the Intrinsic Perspective. A moment of context about me: I am an animist.
People largely talk about the EMH with Simons. But he also evaded the trap which says data mining is dangerous. Finance professors have dismissed evidence of profitable trades as spurious results from torturing data going back to the 1970s. I work in academic finance and the dominant view is still that data mining doesn't work. Simons proves that wrong too, on top of proving wrong the EMH.
All of this is ironic because in the end it seems like much of academic finance is naive data mining. In a recent paper we show that the out-of-sample performance of academic and data-mined strategies are eerily similar: https://arxiv.org/pdf/2212.10317
Hi Eric, I haven't read your paper, and I intend to, but I thought I'd better act quick and ask you, before you move on to another post: What do you think of Terrence Deacon's hypothesis that what makes something emergent are the constraints on the parts (both spatial and temporal constraints). These can't be reduced, since they are about something not there at the microscale. So in your case, a macro system is constrained by some average (coarse grained) quantity like energy, pressure, volume, density, or some lower dimensional attractor in phase space, and describing it using those quantities is not just more parsimonious, but also more predictive (since the microscale description does not include the constraints) than describing it without the constraints. It's not tautological, constraints are not always present, at least not in terms of average quantities, and only when they are present does emergence happen.
So I read Deacon's Incomplete Nature, and I do think he has some interesting ideas. With that said, I also think those ideas are very difficult to understand because (a) there's a lot of jargon, and (b) he uses very complex examples from physics without actually working them out. E.g., the kind of "I'm in a coffeeshop and the waitress pours me a coffee and there's a pattern in the foam, etc etc." I never use examples like that, because they are impossible to work through and easy to get confused about. He's not the only one guilty of this, it's basically everyone who has anything to say about emergence, to be honest, with only a couple exceptions (and it's why the philosophy literature is a mess, it's nowhere near rigorous and mathematical enough). But in Deacon's case I think the combination of jargon and no simple modeling makes the clarity less than ideal, not just for an outsider trying to understand, but also for Deacon himself. Although I like your summary of it!
One example of a confusion I found regularly in Incomplete Nature: a distinction has to be made about whether he means parts vs. wholes or macro vs. micro. These are a bit different. For instance, let's say you have an XOR gate. You can't tell the future of the XOR gate without knowing both inputs. So they constrain it in a way irreducible to the individual parts, since if you only knew one input, you would have zero information about the XOR's next state. But this is just a matter of joint causation, which people often confuse for emergence. Joint causation is much better understood, I think integrated information theory is probably the best take on it but there are plenty of others, like the partial information decomposition or so on. Emergence, which is asking what's special about macroscale descriptions (which are in turn just dimension reductions of microscale descriptions) is less well-accepted and more difficult to understand.
However, if there is a connection between Deacon's views and mine, it is this: if you want, you can rephrase causation as constraint. So rather than saying A causes B to shift into state X, you can say that A constrains the future of B into one state, X. So you could use the framework of causal emergence to say that the macroscale constrains the future to a greater degree than the microscale of the system. This wouldn't necessarily involve any particular physical quantities, however, you can get the result in just like Markov chains. I'm unaware of any other research on the causation/constraint connection than my own, however.
Thanks so much for that! I will think about it and perhaps have some more questions that follow. I love your paper also. If you don't mind, a few more more Deacon-related questions:
What do you think about his attempt to extend information theory beyond Shannon to semantics (signifiers, interpretation, meaning and all that)? Are effective IT and Bayesian theories missing that? Or is all of that just a trivial coupling of Shannon with evolutionary algorithms (or some other utility-maximizing algorithm like Levin and Watson's "natural induction")?
Erik, deep appreciation for you and your writing - such a joy to get your articles in the inbox first thing to start the day.
And I really like your emergence comments.
Look forward to keeping reading your work, Alex
The exclusion argument seems like obvious nonsense to me, which makes me suspect I'm misunderstanding it. Wouldn't it equally prove the pointlessness of studying aerodynamics, since fluid flow is a bulk phenomenon and doesn't really exist when you consider individual molecules?
So it's not that some macroscale property P is non-existent in the *individual* microscale parts (so in this case, just to go along with your example, the fluid flow of individual molecules considered alone or by themselves isn't a sensible quality you could calculate). It's rather that the same property (here I guess, fluid flow) could be described in terms of the molecules collectively together... without ever describing the fluid at some macroscale (like zooming out and dimension-reducing the individual molecules into some larger thing by not mentioning specific configurations of molecules, etc).
As so delightfully often in your commentary, Eric, there are far more numerous and interesting theses touched on in your post than can be addressed in a single comment box.
As someone who works seriously with concepts of emergence in society and historical process at the large scale, and with such concepts intellectually in models of physics and neurocognitive organization, I'm drawn to your struggles with the nullity of even much of informed theorization on emergent processes, which you touch on here. Your theory of causal emergence strikes me as a good one. I would tend to say any macroscale organization tends to mode lock activity to its particulars, with a stability gain resulting at that particular scale of order which is not necessarily implied by organization at any other scale of order. 'Error correction' is one result, though that is to give the process an eschatological reading which is perhaps not necessary to the basic behavioral outcome (where 'error' and 'correction' may not be inherently meaningful terms for the resultant organization).
I would say that there is a more fundamental logical and evidentiary failure to the Causal Exclusion Argument. This proposition requires as an a priori that there be some scale of organization which is formally basic . . . but "there ain't no such animal." An argument that some microscale organization predicates all macroscale organization _requires_ that elements at that microscale be irreducible, whether atoms, quarks, cells, genes, human decisional minds, individual market transactions, or so on. There are no such 'basic orders.' All, when examined closely, have levels of behavior of finer composition yet. Now, organization at whatever scale of order one likes may be highly stable and generally predictable (except of course for the numerous and evident anomalies which said predictions cannot explain and by bodily revulsion ignore). If there is no 'basic scale of order,' there can be no microscale supervenience that is provable, nor in fact likely any that exists at all. It's "emergence all the way down" in my view, a perspective which I've pursued myself in multiple lines of theoretical argument.
The absence of formally basic order is so profoundly evident in any actual field of research investigation that causal exclusion should be laughed off the podium. It isn't, and I would posit isn't for the simple reason that most minds from their own psychological necessity _demand_ that there exist some basic elements "which are true." It's not the putative probabilistic nature of reality which gets them, it's the approximate coherence of anything and everything. Most disciplines in philosophy demand some basic set of a prioris . . . and impose one where it does not exist, effectively if often surreptitiously falsifying themselves from the outset. There are a few lines of argument in philosophy which argue conversely for absolute indeterminism, but that's not much fun as it means you can't really 'prove' anything, and so be smarter than the next simian. One beauty of emergence as a concept is that it bridges that false dichotomy. 'Facts' are order- and organization-specific, not universals from this perspective. At the same time, mutability and inconstancy are inherent to systemic behavior, yet they are seldom truly and pervasively random as opposed to being shot through with organization, thresholds, and boundary constraints (with organizational components which are order-specific stable to a degree). Emergence DEMANDS an entirely new approach to philosophy, but for that matter to science and history as well, not least because of how emergence reformulates what 'causality' can and does mean, in my view.
You touch on interesting points regarding the declining complexity of poetry, Eric, and gatekeeping hostility disguised as indifference cloaked in 'objectivity' to any kind of original thinking at all. Just in reading the application guidelines for 'genius grant' foundations, it has struck me that their process is profoundly antithetical to any truly interesting theory or modeling, much as you found. From that perspective, Jim Simon would never have secured a grant from his own foundation either, other than perhaps by personal charisma in the interview or prior reputation (because that is what drives something like 70% of the yes in granting, as I see it). Perhaps I'll find time to opine on these issues as well.
Great comment. The argument you're making against the Causal Exclusion Argument is definitely a good one. Ned Block also makes it in "Do causal powers drain away?" with the idea that, if it really is "bottomless" in that you can keep dividing things smaller and smaller, all notion of causation vanishes, since it's forever "draining away" at each microscale. I think it's an interesting way out of it, but at the same time requires that assumption that there is no basement scale to physics (many physicists might disagree). Additionally, like any reductio, there's always the option of just biting the bullet. In other words, it does away with the Causal Exclusion Argument, but you're also giving up causation in general. I think maybe there's a way to save that (it sounds like you'd be sympathetic to this) which is to say there are no "privileged scales" in the situation where everything is forever draining away. But this path, at most, nets you a sort of relativism about what causes what, rather than clear-cut cases of emergence vs. reduction.
I'm sure my conception here is not original or unique to me, and it is interesting to hear how others approach Causal Exclusion. I have spent literally years working to build up a frame of reference for causal process in history where emergence is intrinsic. I have so much work to do already, I'm beginning to think I won't live long enough to get to that particular, treasured, book.
I would tend to say that 'causation is local.' There is no need to see causation as draining away: what fails is any _global_ causation. Causation is local to scale of order, and often if to lesser extent specific to system of organization.
I in fact have a toy model of general physics in which order scales in a profound way, with causal relationships and order specific to varying shells-tiers of dimensional organization and complexity. I'd love to see someone competent mathematize it---I can't---because I think it's the way out of the dead end into which physics has particle-ized itself. Yes, most physicists would and will hate it, but it's a way around the sense that there is no causal validity at all, just as its a way out of the trap that there is only a single, universal causal relationship set as opposed to a scaled set of relations only some of which are accessible at a given scale of order.
I'd love to discuss some of this with you, Eric, and other issues as well. I've wanted to send you drafts of my text as I think aspects of it would interest you. If that's to your liking, please let me know a path to contact you outside of this. Should you be inclined to check out my approach first, I comment over at The Jagged Spend.
Interesting piece on an interesting person. In regards to what you share around causal emergence, and the notion of it acting like an error correction at the macro scale. I have a question and a though:
Q: There seems to be some inconsistency with the idea that a fully known micro-state (ie. "If everything is accounted for in the microscale atomic-level descriptions") would not also include, theoretically, the causal effects without uncertainty and probability-based approximation. It seems like it still comes down to whether, at the most fundamental level, there is a deterministic or stochastic system at play in the universe. This conflates the idea that while we may theoretically be able to predict future states in chaotic systems, we do not (currently or possibly ever) have enough precision on the current state at the micro level to do so. I think this lead to a dual interpretation of your work to suggest that there is a stochastic system at the micro level, and we then perceive a deterministic system when we operate at any macro level.
Thought: Your analogy of this concept seems like it would equally suit a description of moving between an analog system (real world) and a discrete one (computers). Of course, that is also based on the underlying aspects of information theory as well, but might be a little more intuitive to people. What see in the real world and what you get in the pixels of a photograph is common interface for ADC.
I really enjoyed your use of error correction analogy and information theory.
Very interesting and well written piece as always. It gave me a lot to think about.
I will say however, that the efficient market hypothesis is one of those outdated bits of junk economics that just never died in the public consciousness. It's analogous to those little experiments where you place a jar on the table and have a room guess how many jelly beans are inside. In this case it often turns out that the average guess is more accurate than most or any individual guess, and if you keep repeating this experiment, no one would ever beat the average over time.
Well, this is exactly how the market works except for one thing... when emotion is involved, it knocks the accuracy completely out of whack. The crowd is quite irrational, as are the individuals that make it up, so a well tempered and logical person has a decent chance of beating it. The market is not driven by a cold assessment of available information, as the EMH assumes, but by fear and greed. Unfortunately, it is extremely difficult to distinguish oneself from the fervour of the masses.
‘Later, I think it was Simon DeDeo who presented a talk to all of us, including Jim, on the fall of complexity in poetry. It turns out that if you measure the entropy of published poems they’ve been declining as far back as back goes…’
Sorry, what?!
Yup, that's what I remember from the talk. I forget what measure they used, it was some sort of bag-of-words calculation. But yeah, poetry is supposedly more predictable than it used to be in the past. Another way of saying it uses a smaller vocabulary, by the way, just to give some intuitions (it's easier to guess the next word if there are fewer words used in general).
Coming from the poetry trenches, I would tend to agree without seeing the analysis. Lots of people that sound like a single fairly anonymous collective poet. I tend to think it's related to the increasing prevalence of MFA programs and the academic professionalization of poets, at least in the US. Or maybe I'm hoping that it's that, and not some basic erosion of culture and creativity in the species.
I would love to read the methodology used by Simon DeDeo, and especially the timeframe of observation. It seems to me I've run across similar conclusions before, and I agree with the overall argument: complexity in poetry has at least a modern vector of continuous decline. As someone who has long written complex poetry, I can attest that a) it is impossible to other than self-publish it, and b) published poetry is vastly less complex now even than when I began.
I think there are three reasons for this trend of result:
1) The Well-trodden Path. When few write or publish poetry, the possibility of divergent results are greater, and hence aggregate complexity is inherently likely to be higher. When many publish, even if they are in principle original, the probability of convergence of expression goes up, and aggregate complexity necessarily declines.
2) Conformity Bias. Originality is always possible, but social groups have strong expectations of conformity, or at least of not standing out. As more publish, the numbers who herd together in similar vocabulary, diction, and construction go up, by psychological osmotic pressure, as it were, even more than by intention. In that context, aggregate complexity is simply bound to decline, even if outliers in expression are as frequent as ever (which they are not).
3) Suffocation Under the Bell Curve. Outliers in capacity are few. The numbers (and hence the money) are in the fat part of the bell curve. Low complexity is simply more probable in the middle of the distribution, so that as the total numbers increase, complexity is inherently driven down, even if everyone is 'original' in their creative result.
I would love it if the mass popular production of 'poetry' we see now yielded the beauty and the complexity of the ages . . . but it does nothing of the sort, and IS nothing of the sort. To say more than that would be to run off on a rant; so I shan't.
I made a note to self to google this. Is it not published? Fascinating. To my pale understanding, information entropy is the distance from uniformity of a probability distribution. Poetry has become closer to Random words? By the way I must tout w3w.com at showing how 3 words are like a personal space upon our vast Earth. My office is at quietly.divide.raindrops and my parents are buried near spare.person.parts. Coincidence? Guaranteed, yet it sticks in the brain.
Oops and uneditable: w3w.co
Cheers from
https://w3w.co/quietly.divide.raindrops
Simon DeDeo does lots of cool work. First learned about him on this podcast: https://complexity.simplecast.com/episodes/72-kkxpnFGT
I’ve never heard of the exclusion argument in the philosophy of mind being used to argue against the significance of macro-scale behavior, that would be bizarre indeed! As far as I know, it’s always used as an argument in support of physicalism. The argument being that since the macroscale behavior of neurons supervenes on the microscale, you can’t change the behavior of neurons through some non-physical effect without contradicting physics. That’s what the wiki article also suggests in the link you provided.
I would be very surprised if any serious physicalist argued that macroscale neuronal behavior is not an appropriate level of scale to study, or that cognitive science is not meaningful since it can’t contribute anything beyond physics. But perhaps you’re familiar with a section of the community that I’m not?
It's an inevitable consequence of the argument that it generalizes. This is the crux of Ned Block's reply to the causal exclusion argument, "Do causal powers drain away?" which is probably the most popular piece of the literature besides the argument itself (if it were only about minds, causation wouldn't "drain away"). There's plenty of other admittances of this in the literature, e.g., like Bontly's "The supervenience argument generalizes."
Hi Erik,
I wasn’t disputing whether causal exclusion generalizes to other cases. I was saying that even in the case of the mental, the notion that causal exclusion entails that there is no mental causation seems to me false. Rather it entails that there is no mental causation iff we adopt a non-physicalist theory of mind. The argument as I understand is just that causation doesn’t go over and above its causal parts. Whether we want to say that this rules out higher-level causation depends on your terminology, and whether you think of higher-level causation as needing to be something non-reductive. In other words, I take Jaegwon Kim’s argument to be saying that we either have to accept epiphenomenalism or reductionism about the mental (mental states just are physical states).
Hence Kim’s argument is compatible with mental causation, provided we accept reductionism. It’s purely an argument against interactionism. If this is true, then it doesn’t matter whether the argument generalizes, since even in the case of the mental it doesn’t say that there is no mental causation.
I read Block’s piece but am unfamiliar with Bontly. Block gives the interesting argument that if there is no fundamental level of physics, then were would be no causation according to the CEP, since every causal action is grounded in its parts. But even if he’s right and the possibility of no fundamental level argues against CEP, I don’t see how that justifies your assertion in this article that proponents of CEP think there is no mental causation, or that thinking about macro-scale causation is a waste of time since everything happens at the micro scale anyway.
“I don’t see how that justifies your assertion in this article that proponents of CEP think there is no mental causation”
To be clear I meant that I don’t see how it justifies your assertion that CEP proponents think there is no higher level causal emergent behavior, which of course would exclude mental causation. If even in the mental case CEP proponents think there can be mental causation on reductionism, then obviously they will think there can be higher level causation in general cases as well.
Unless of course you’ve been talking about strong emergence all along, where the lower level behavior of atoms is not actually exhausted by the laws of physics. But if I understand you correctly, that’s not what you mean.
Just to see if it clears anything up: The issues have nothing to do with mental causation at all, if that helps. That’s where some of the literature originated but I’d set all mental causation stuff aside entirely. And understanding of “every causal action is grounded in its parts” is not how I’d phrase it, exactly. I’d phrase it as about different levels of description for the causation that does occur (this turns out to be different than “it’s in the parts,” e.g., a macroscale can have parts, they’re just macroscale parts).
Thanks. I agree that your levels characterization is better. Do you agree that a micro scale quantum description of some event won’t miss anything that a macro scale description would? In other words, that the behavior entailed by the emergent efficiencies (among other things) which pop up at macro scales would be predicted by the micro scale description? If no, then wouldn’t this just plainly contradict physics? If yes, then I’m not sure what you think CEP proponents are disputing when you say they think the macroscale doesn’t ‘add’ anything.
The notion of emergence manifesting somewhere between the micro- and macroscales is most interesting, particularly if we think about it in the context of artificial intelligence. It strikes me that, intuitively if not formally, a great number of the most temporally optimistic theories of AI progress (i.e. those that believe we’re closer to AGI rather than further) seem to hope-against-hope that scale will deliver the manifestation you speak of, wherein a non-intelligent ‘micro’ function turns by some miracle of aggregation into an intelligent and cohesive ‘macro’ function. Puts a lot of AI scientists in league with acolytes of the Upanishads and adherents of panpsychism, as far as I can see, relying on the aforementioned ‘magic’ to conjure the emergent property. I can imagine, however, the potential of a kind of science of ‘emergence dynamics’ to determine how multifarious interactions between different elements at the microscale not only produce the adjustments you speak of but fundamentally attune such various inputs as combine in extremely complex processes like the generation of consciousness.
Also interesting your experience applying to be a Simons fellow – it seems as per your account that the people in charge of assessing your application were an almost perfect dispositional inverse of the fellowship’s namesake (or perhaps it’s truer to say, if they seemed clocked out during your presentation and Jim was also prone to nod off during an address, they might in fact have emulated him too thoroughly!). Knowing how to build an institution in the image of a unique founding figure really does seem like the most delicate art in the world. Do you know if the foundation still has any comparable/loosely-comparable figures of genius/high-interest doing work there now?
When does a chair begin to become a chair? Emergence and phase transition everywhere.