Thanks so much Alan for reading. I don't think of this as a form of vague or spontaneous emergence - it's quite structured and non-arbitrary. It's nothing like the example you give in your own essay of the idea that past a certain (arbitrary) level of complexity, consciousness spontaneously arises. I have some thoughts on your own essay that I will comment on there - thanks so much!
Erik P Hoel
Thanks so much for your thoughtful comments. Causal emergence was first shown as a separate phenomenon in (Hoel et al. 2013), but you're right that we then later showed that it is at work in IIT: systems can have higher integrated information at larger scales (Hoel et al. 2016 in references).
As to your comments:
"..."quark clouds" do not constitute atomic physics. Quarks constitute the atomic nucleus while chemistry depends on processes involving electrons in the atomic shells."
Took me a second to find what you're talking about (which is the arrow on the second to bottom step in the ladder in Figure 1). Yes, the picture is infinitely more complicated than shown! I certainly can't draw out all the supervenient arrows also because scientists don't spend a lot of time looking for what Ernest Nagel called "bridge laws." In retrospect I could have just put "elementary particles in the standard model" which covers more bases.
"Second, is "causal emergence" really so surprising? To me "causal emergence" seems to appear whenever an experimentalist is performing a series of measurements and statistical uncertainties average out (essentially the "law of large numbers" in stochastics). Probably even more prominently in quantum mechanics the fundamental level appears to be less deetermined than macroscopic classical physics. Can you elaborate a little more about whether the causal emergence you are talking about is different? Do I misunderstand the concept?"
Well, let's differentiate between whether causal emergence is surprising vs whether it's rare. I think it is very surprising given that conceptually it is not at all obvious how higher scales can do extra causal work and generate more information. So it's conceptually surprising. But I don't think that it's rare, in fact, I think that it's pretty much everywhere to greater and lesser degrees. But it's not so common that it's just whenever there's an average in an experiment (and not all higher scales are averages). You can have plenty of averages, or let the law of large numbers work, without causal emergence. For example, rolling a die over and over won't give you any causal emergence. However, regularly in science experimenters treat higher scales as causally-manipulatable variables. This can be reflective of causal emergence. For instance, assessing the causal relationship between two neurons (not their underling elementary particles). Causal emergence explains why that's not just about using a simpler representation. You're possibly gaining something in this scenario: extra information about the causal structure that's not available at the microscale, because at the scale of neurons there is error-correction, whereas at the scale of elementary particle physics there is no error-correction. If you believe the hypothesis I lay out of the ladder of science being reflective of each step causally emerging from the one below it, causal emergence should be incredibly common, especially in the special sciences (like biology).
As for quantum physics being less determined than classic physics - well, that's a good scale to look for causal emergence at.
"Also I'm a little confused about the exact relationship of under- and overfitting (overfitting seems clear to me but can somethink like underfitting be possible without causal breaks?)"
Glad it came across. I don't think underfitting is possible without causal breaks - that's why I describe teleology as "breaks in the causal chain" and say that it leads to underfitting by a causal model. So I think it's an extended aspect of causal emergence that's unique to agents (or things with equivalent apparent teleology).
"One may argue that pure information-based conecpts such as the "bill of rights" or the "contents of the bible" are totally substrate independent, and that agents exist somehow in between such imaterial objects and material physics. Would you agree with this interpretation?"
It's a nice idea but I wouldn't agree. First, I don't think anything is actually fully substrate-independent, more like things are substrate-constrained to greater or less extents. Second, I don't think there's some immaterial world that anything is closer to.
Thanks so much for your great questions and your interesting thoughts. I will definitely check out your essay!
Erik P Hoel
Hi again Eric P. Hoel,
I am getting back having seen your Discovery article followup to Lorber et al's claims on small-brained performance. Two points,
1. That article really doesn't do much to deflate the implications of Lorber's findings.
2. That was simply a warmup example in my essay (which itself is a warmup example) as it nicely followed Sean Carrol's quote. I re-suggest that there are basic challenges - and lessons to be learned - by looking beyond the textbooks.
Good luck with your work,
Ted Christopher
Erik--
Point taken about the funding! I'm going to be looking for a new job in the coming months and perhaps I'm just trying to be overly optimistic about how popular the worldview expressed in your essay really is! Also I managed not to notice that you are at Columbia the first time I looked this over, we collaborate with the Banta group and the Venkataraman group on different projects, and I'm literally a subway stop north of you. Feel free to come see our lab at City anytime. We design proteins and characterize them with spectroscopy; we have lots of cool lasers.
And we should discuss coding in biology while you're there too. I didn't focus on it in my essay because it didn't fit neatly into my narrative, but I completely agree that coding or perhaps more generally "translation" of information is an ubiquitous phenomenon that is not merely compression from coarse-graining but involves emergent dynamics from interactions that do not appear in the local microscopic dynamics. I think the genetic code is actually the perfect example of that. I taught a course at city college on principles of statistical physics in biology and we spent a week or so discussing what makes translation of nucleic acids into proteins a much more interesting and physically difficult to explain phenomenon than transcription of DNA to RNA. The translation step is what makes it a code rather than just chemistry, because there is no direct structural connection between a codon and an amino acid, and the code only appears when you look at the correlations in the network structure. I actually have a ton of thoughts on this issue and I'd love to pick your brain too and since you're just up the street I think we should probably make it happpen!
Joe
Hi Erik,
You're probably not going to like what I say. I am going to bring up two points concerning your article and none of them are addressed in your article.
The first point is simple. You plopped yourself in the middle of religion without even realizing it. It concerns the Romeo and Juliet argument you bring up. You seemed to miss that there was a wall at the end of the story that neither Romeo or Juliet could circumvent that was Death. Since you discuss the internal architecture of Romeo's brain we will go with that. Romeo has received new information from the environment. Juliet is dead. S_d to S_k no longer exist. Or do they? A new S_k could be, Romeo will meet Juliet in the afterlife. That is religion. I'm not going to get into beliefs and the ends people use to extend them.
At the end of the 19th century, a certain scientist named Planck formulated his "Black body radiation" law. This law used the smallest thing currently known to man to describe the radiation of a black body. But it is not limited to black bodies, it includes the human body, our sun, our galaxy, and the CMB. How is any large scale effecting the smallest scale. To believe your model you would think the Planck's constant doesn't exist above atoms.
I See all the superlatives other people have given your article. I can't join them. I was drawn to your article by a tweet from Natalie Wolchover and calls your article "the %#*&". Good luck in the contest.
Jim Akerlund
FQXi logged me out so I am not really "Anonymous".
Jim Akerlund
Interesting article that tries to understand the influence of the dimensional scales in human psychology and consciousness.
But taking into account that the composition of all living beings (known) is formed of very similar cells and molecules, we should deduce that the greatest differences occur in the higher levels and not in the lower levels as I understand you propose.
Please read my article (http://fqxi.org/community/forum/topic/2925) that also deals with the SPACE SCALES applied to Cosmology.
Congratulations on your proposal!
Dear Erik,
You have already received so many comments that I am not sure, if this will add much to that. Yet ...
In many statements in the essay, there is one point reverberates as a non-violative presumption that the nature is deterministic in the absolute sense at the most fundamental scale (level). For example,"In this reductionist view, a biologist studying a cell is really referring to some astronomically complex constellation of quarks." The function at the cellular level may not be entirely determined by the quarks, if there is a certain degree of indeterminism in the processes at any scale.
In statements like, "Recent research has argued exactly this [14, 15] by demonstrating the possibility of causal emergence: when a macroscale contains more information and does more causal work than its underlying microscale.", it is not clear what is referred to as more vs less information. I suppose, you just showed all along that complete description of microstates carry more information than any abstraction of the same at a macrostate. As you also mentioned Shannon's quantification principle, the detail in microstate can distinguish a state among much larger set of states, therefore contains more information. But the quoted statement seems to violate that.
Though, as I have shown that ignoring certain detail gives rise to emergence of abstraction that is not present in any one the components. This is a process of generalization that applies to all objects, say states, relations, functions, and processes. But in Shannon's view, it corresponds to loss of information.
With respect to transition matrix, Smicro and Smacro, I am not claiming that I understand it completely, but this is what appears to me. Imagine the micro states are designated as (a,b,c,q), and after ignoring the distinction between a, b and c, we call the resultant macrostate as p. So, if the elements of the transition matrix are probabilities of transitions, then the following points can be noted, 1) 'a' may transition into a, or b, or c, with equal probability of 1/3. The same is true for b and c. Whereas, with Smacro, p remains p, and q remains q for all times. So, I conclude that Smicro certainly has more information distinguishing among finer transitions than Smacro. Keeping the same logic as proposed by you, if we ignore all internal state details and transitions of elements in universe U, then U remains U in perpetuity, a very definitive statement, no degeneracy, absolutely deterministic, but that hardly tells anything worth distinguishing. 2) If the elements of transition matrix are probabilities, that means it is not a fully described system, as per the presumption of determinism, or the determinism does not hold, or determinism is limited. That is, Smicro is also already an abstraction of even more fundamental description. Talking the language of quantification of information, I can represent the Smacro with mere 4 bits, but Smicro would take many more. But in my view Smacro is an abstraction of Smicro, where certain details are of no importance.
Given your definition of teleology, are we to understand that bacterium also possesses a sense of goal and purpose, acts accordingly to discover resource?
Rajiv
Thanks so much for reading David. I appreciate it.
I just checked out your article - I think figure 2 is a really wonderful representation of the truly diverse range of scales.
Thanks again,
Erik P Hoel
Thanks so much for reading Rajiv, I appreciate it.
I'll try to address some of these objections, which are rooted in misunderstandings or misconceptions.
"The function at the cellular level may not be entirely determined by the quarks, if there is a certain degree of indeterminism in the processes at any scale."
This is a misunderstanding between determinism and supervenience (where the lower levels determine, or fix, the higher levels). A non-deterministic system will still have strictly fixed supervening levels. The state of the cell supervenes (is determined by) the constellation of elementary particles below it.
"In statements like, "Recent research has argued exactly this [14, 15] by demonstrating the possibility of causal emergence: when a macroscale contains more information and does more causal work than its underlying microscale.", it is not clear what is referred to as more vs less information. I suppose, you just showed all along that complete description of microstates carry more information than any abstraction of the same at a macrostate"
This is the opposite of what is shown in my essay (and the associated papers). The definition of more or less information is pretty clear in the essay itself (given in the equation for effective information), and in the associated papers there are numerous examples demonstrating how to calculate this information and showing it done on model systems.
"If the elements of transition matrix are probabilities, that means it is not a fully described system, as per the presumption of determinism, or the determinism does not hold, or determinism is limited."
It's a metaphysical assumption on your part that all systems are deterministic. And it's simple not true - any open system will experience noise (for instance, a cell being bombarded by cosmic rays). Even in the deterministic universe the only way to get rid of any notion of noise is to avoid partitioning the universe into notionally separate systems (like a cell, by itself). If you can't do this, it leads to very serious problems (discussed in Appendix D). And besides, as said in the essay, the point is moot anyways, because causal emergence can occur even in deterministic systems.
"Given your definition of teleology, are we to understand that bacterium also possesses a sense of goal and purpose, acts accordingly to discover resource?"
Great question. It may be possible to design even simple systems with the kind of teleology I propose. I would view it as a matter of degree, not a binary answer, where humans might have orders of magnitude more "teleology" than a bacterium.
Thanks again,
Erik P Hoel
OK Erik, but further this figure, you should see a framework to explain a lot of Physics mysteries (Black matter and Energy, Uncertainly principle,...) and a way to follow up to improve in Cosmology: Universe scales are infinity (with diferent laws, concepts,..), and no TOEs will be possible.
Erik P Hoel,
What it means an information to an atom?
Hi Eric,
I think you were very busy because untill now I didn't see your comment (nor a rating) on my essay "The Purpose of Life" where my thoughts passed the Planck Wall in order to find a reason for our reality.
I hope that you are still eager to read it.
thank you
and best regards
Wilhelmus de Wilde
PS sorry that I am a little pushing but I would appreciate the thoughts of someone with a worldview like you.
and again sorry Erik for the faulty Eric...
Wilhelmus, I already read your essay two weeks ago.
All the best,
Erik P Hoel
Hi Erik,
You do a most wonderful job....(this is from a confirmed "we really do not know person"). And your participation in the Q&A above fully compliments your essay.
What strikes me as most interesting is:
The agents are up in complexity (and therefore information) and therefore the agents not only causally emerge, but significant aspects of their causal structure cannot be captured by any microphysical model. If this is true then causal emergence, whether through irreducible physical properties or because of measurement and observational tools, may explain why science has the hierarchal large structure that it does. New rungs in the ladder of science causally emerge from those below them. Different scientific fields are literally encodings by which we improve our understanding of nature.
I cannot imagine that Darpa and other government agencies would not be throwing money at you.
I personally would like to see a computer simulation of this causally emerging agent...I want to play with it :)
Say hello at my blog. I think you will find some humor in it.
Don Limuti
Dear Erik,
Thanks for taking the discussion forward.
It appears, your claim of misunderstanding is indeed true. For example, you state, "A non-deterministic system will still have strictly fixed supervening levels. The state of the cell supervenes (is determined by) the constellation of elementary particles below it."
I suppose, for a non-deterministic system, a given microstate description may still allow more than one possible description at supervening level. Is not that the very definition of non-determinism? Then, we cannot assert, "A non-deterministic system will still have strictly fixed supervening levels." There must be something more than strictly determined description of the microstate to give rise to a given supervening state description. I am sure, somewhere our definitions do not coincide.
Misunderstanding goes deeper. In a specific context I stated, "If the elements of transition matrix are probabilities, that means it is not a fully described system, as per the presumption of determinism, or the determinism does not hold, or determinism is limited", it was meant to show the Smicro is also a 'multiply realized' description (not an ultimate description of unique reality). Here, the idea was to show that logic that applies to Smicro to Smacro, must be true even for Snano to Smicro, as well as to remind that you are not building a case here from a deterministic micro world to supervening Smacro, but rather a case for one indeterminate system to another.
"It's a metaphysical assumption on your part that all systems are deterministic."
Ah! This response goes deeper than simple misunderstanding, since if I responded to your statement, "In this reductionist view, a biologist studying a cell is really referring to some astronomically complex constellation of quarks", with, "The function at the cellular level may not be entirely determined by the quarks", I am also questioning the very presumption of determinism itself. I am implying here that 'a cell' is not just mere 'constellation of quarks'. So, I am not sure, how you happened to miss this argument to infer I am for determinism? By the way, I assert here, even greater indeterminism by saying, "one does not require a cosmic rays, or anything external to disturb a system exhibiting indeterminism, all systems at all levels inherently possess limited determinism." That is, even at microscopic level, even a quantum system exhibits indeterminism within limits. Therefore, 'determinism' could not be my metaphysical assumption.
I suppose, there is enough misunderstanding already, therefore, I will hold my further queries at the moment.
Rajiv
Thank you Jack, I will check it out.
sorry Erik,
As you did not leave a comment I just didn't know that.
regards
Wilhelmus