Yes, I agree. I am not perfectly happy with the results of the LLM party as Jon calls it. It is an up and down. One moment you think: oh wow, I get real results and complete proofs in a few minutes. And then Claude nitpicky insists on 82 details that are still totally wrong and you are unsure if anything about it makes sense.
One problem I see is if we use non-communicative models such as SO(N) for N > 2 to model path dependence then we can model path dependence a little bit but not divergence and convergence of world-views. We just rotate unit vectors around on an N-dim sphere without changing their length. If we use linear euclidean models (plain vector arithmetic: attraction toward others or repulsion, added noise) and random encounters we can model divergence and convergence well, but not path dependence. Finding the right model and level of abstraction for a hard problem is hard as well. Nevertheless interesting. -J. On October 12, 2026 7:18:22 PM GMT+02:00, glen <[email protected]> wrote: >Interesting! My gestalt take: If you were modeling shared values, conflict >resolution, negotiation, arbitration, or any such thing, then the formalism >looks fantastic. It's very easy to overcome minor nits like "your axes are >never orthogonal", "values/qualities are not differences in degree but kind", >and "15°, 60° are assumed not measured". > >But I thought we were modeling *consciousness*, which is why I wanted numbers >related to consciousness (e.g. fMRI or EEG or somesuch). Do values supervene >on conscious states? I doubt it. > >The above nits remain, though. TLDR warning. > >1. Order/Safety, Care/Harm ∈ ℝ: While I can be generous when talking about >shared values, it's more difficult in terms of conscious states. But let's >stick with shared values. How do we bind -∞, +∞, 0, and 1? These are important >values in ℝ. We can gloss over continuity, density, etc. But those 4 values >need bindings. E.g. What does "infinite harm" or "infinite care" mean >practically? What does "zero order" or "1 safety" mean? > >And more importantly, by what repeatable, "works everywhere", method do we >measure these things? I can imagine we can take methodological examples from >places like UNESCO for "food insecurity" or "genocide" from the UN, etc. We >may also get some relatively stable suggestions from the economics literature, >maybe some theoretical biology for things like autonomy. > >2. Orthogonality & kind not degree: Do this math work in weirder spaces? My >guess is they do, or they will if changed slightly. Rather than binding to >simple spaces like ℝ⁶, the rhetoric would be more convincing if you bound to a >weird space. Hyperbolic spaces? Graphons? IDK. > >3. The transformations (and the H implied by them) of 15° and 60° seem like >unnecessarily separate assumptions. Once you define the methods of measuring >the axes, these should fall out. But maybe the extra assumptiveness lies in >some kind of _sensitivity_. E.g. sure, Hank's *registration* of the events >during Home→Street→Home may be small. But maybe that's because he's a >coarse-grained, brutish person. Walt is a sensitive guy. ... Or maybe we could >think more in terms of cognitive or memory power. Maybe Walt's simply way way >smarter than Hank. Maybe Walt's value space is ℝ⁹ where Hank's is ℝ³. Then we >need a reduction as well as a transformation. > > >On 10/11/26 5:43 AM, [email protected] wrote: >> >> You can find some numbers here in these examples. Claude helped me to >> simulate two basic examples: the "Breaking Bad" story where a good chemistry >> teacher turns into a bad drug maker, and Jonathan Haidt's theory of moral >> foundations which he explains in his book "The righteous mind". The first >> one uses SO(3), i.e. rotations in 3 dimensions, the second one uses 6 >> dimensions (SO(6)). >> https://github.com/JochenFromm/PhilosophyBook/blob/main/Empirical%20Case%20Studies.pdf >> >> The Python script to check the "Breaking Bad" numbers can be found here >> https://github.com/JochenFromm/PhilosophyBook/blob/main/breaking_bad_model_so3.py >> >> How it works is you have a "valuation" vector that represents what the agent >> considers as good or bad. In these toy examples we have 3 and 6 dimensions, >> but it works for larger dimensions too. Depending on the situations which >> the agent encounters the valuation vector is rotated by rotation matrices >> from SO(N). Rotations in SO(N) for N >= 3 are commutative only when they >> happen around the same axis or if they are very small. Very small or >> infinitesimal angles approximately commute, but larger angles do not >> commute. This non-commutative property (AB != BA) can be used to model the >> part of subjective experience that is path-dependent. It sounds paradox, but >> if we model cognitive processes in this abstract way, we can mathematically >> prove theorems about things like subjective experience and moral values >> which are hard to measure in real life. Jonathan Haidt has used a >> questionnaire to get empirical data for moral values. This is one way to get >> data about emotional valuations but in gene >> ral it is hard to get. >> >> The art is to find the right level of abstraction. Simple models are easier >> to understand but often describe reality poorly. If we use the "spherical >> cow" abstraction to model a cow as a sphere we can describe the movement >> precisely, but lose all properties which are characteristic for a real cow. >> Complex models increase the risk that the thing we want to model is true by >> construction. The art is to find a simple well-defined model that has >> complex results. As David Hilbert said "The art of doing mathematics >> consists in finding that special case which contains all the germs of >> generality". By the way the tombstone from Hilbert in Göttingen contains the >> words "We must know. We will know". I like this attitude of boundless >> curiosity. >> https://en.wikipedia.org/wiki/David_Hilbert >> >> -J. >> >> >> On 2026-10-06 16:38, glen wrote: >>> OK. That's somewhat helpful. But what we need are numbers, right? >>> >>> So let's go with H_k(•) being a/the primary binding point. How do we >>> measure it? What is it in the person's physiology that we measure? It seems >>> like the validation has to occur in the distance measure: >>> >>> d(x,y) = min_Q∈G max_k || H_k(y) - QH_k(x)Q^-1 || >>> >>> So we need real world bindings to those symbols. x,y ∈ people, whatever >>> that means. H_k is the change in the state of the person after *returning* >>> to a prior context. Let's just allow the state (U) to be whatever for now, >>> glucose levels, EEG, survey measured Virtue matrix, whatever. We simply >>> assume it's a high dimensional, dense space. Q is just a basis change for >>> that space. Yeah, therein lies the assumption that we can define such a >>> space. But under all the assumptions, it seems like the simplest operator. >>> >>> All those bindings seem imaginable, to me. The problem I have is the loop. >>> To what do we bind these "closed loops"? I don't think we can bind them to >>> things like loop k = 1 = {home, shower, school}, because H_1 may be 0 one >>> day (on "autopilot") but H_1 >> 1 on another day (you were beaten and had >>> your lunch money stolen). >>> >>> We might be able say k = 1 = {healthy social support group, >>> hypercompetitive work colleagues, casually competitive Tennis group}. These >>> supernodes may make sense. One person's U() may not be impacted by the >>> higher order effects of competition: U_0→1 = 0. Another may not need social >>> support: U_2→0 = 0. >>> >>> But if we compare those two people H_1(x) = QH_1(y)Q^-1, where Q swaps one >>> node for the other. But we would hardly consider those two people to have >>> equivalent worldviews. Or, going back to the bullying example, if X is >>> bullied on Tuesday and Y is bullied on Wednesday is H_1(x) = QH_1(y)Q^-1 on >>> Thursday? Or maybe it takes some time? Maybe H_1(x) = QH_1(y)Q^-1 a week or >>> so later after the initial shock wears off? Or maybe if you shift Y's >>> worldview by one autopilot day, *then* it's equivalent to X's? >>> >>> I guess my point is that for us to know whether the structure of the >>> formalism matches, in any way at all, the structure of the system being >>> modeled, we need a whole "modeling relation". Encode the natural system >>> into the formalism. Decode the formalism into the natural system. A "data >>> first" approach defines the [en|de]coding, then searches for inferential >>> structures on either side. I'm OK with your formalism first approach as >>> long as you commit to the followup of searching for [en|de]codings. Don't >>> bail halfway through and call it "done". 8^D >>> >>> >>> On 10/5/26 2:28 PM, Jochen Fromm wrote: >>>> Good question. Yes, without experimental confirmation it is just fancy >>>> mathematics. The model used for the proofs is an abstract model. It >>>> describes how mindsets/worldviews of adaptive agents diverge and converge. >>>> >>>> In a simple toy model it works like this: >>>> >>>> >>>> A context/nodecan be: the places the agent moves through like home, >>>> school, work, street. >>>> >>>> >>>> An edge means stepping from one context to another: home-street, >>>> home-work, .... >>>> >>>> >>>> Moving along the edges changes the internal valuation matrix of the agent. >>>> Every time an agent moves along an edge, their internal valuation frame >>>> gets rotated or updated by an edge matrix. The experience of making that >>>> specific transition reshapes how they evaluate things. >>>> >>>> >>>> If we reach the same context again we get a loop. A holonomy H measures >>>> how a valuation changes after a loop of contexts. It is the total, >>>> combined shift in the agent's worldview after completing that full loop. >>>> >>>> >>>> A "breaking bad" model of two agents would include two agents who start >>>> with slightly similar worldviews. One interacts with a group that thinks >>>> breaking the rules is bad, and the other joins a group that believes >>>> breaking the rules is necessary to survive because the world is cruel or >>>> rules have been imposed by bad people. >>>> >>>> >>>> A more complex model would be cultural dissemination. As you know Robert >>>> Axelrod created an agent-based model of cultural dissemination based on >>>> local convergence global polarization. It would be interesting to combine >>>> his model with the new gauge theory model. >>>> >>>> >>>> Other areas of experimental verification would be >>>> >>>> + Group psychology and organizational psychology (group formation and >>>> fragmentation, group influence in individual) >>>> >>>> + Political science (emergence of new movements, radicalization pathways, >>>> the threshold of schisms, etc) >>>> >>>> + Sociology (what holds societies together, and what is the role of >>>> socialization and institutions in maintaining social cohesion, to what >>>> extent is the individual shaped by society) >>>> >>>> >>>> -J. >>>> >>>> >>>> >>>> -------- Original message -------- >>>> From: glen <[email protected]> >>>> Date: 10/5/26 8:00 PM (GMT+01:00) >>>> To: [email protected] >>>> Subject: Re: [FRIAM] A field theory of understanding >>>> >>>> I haven't finished reading Jochen's theorems yet. But it seems obvious to >>>> me that he's not claiming experiences are private so much as they are >>>> unique. It seems more about irreversibility or unreachability than >>>> closedness. >>>> >>>> And I'm no expert on Wittgenstein, but my own perspective has been that >>>> his dictum was more about coherent form as opposed to ontology. I.e. if >>>> your language/formalism is inadequate for its target, then it's useless >>>> for talking about that target. It's similar to Rosen's argument that >>>> predicativity is inadequate for describing complex systems. Maybe some of >>>> us just toss up our hands and give up when our language seems inadequate. >>>> But others of us develop new language. And still others push the current >>>> language as far as it can possibly be pushed until we're convinced of its >>>> [in]adequacy. >>>> >>>> But back to the first theorem. This seems much closer to a theory (by >>>> which I mean a body of assertions falsifiable through measurement). But >>>> what still seems missing are the bindings. We don't need bindings for >>>> *every* symbol in the math. Some symbols can be internal. (What that >>>> _means_ is debatable.) The "numerical" sections for both give a good >>>> indication of what we have to bind. Theorem 1 uses >>>> {x,y,z,roll,pitch,yaw,t}. Theorem 2 doesn't seem to have a clear binding >>>> point. Maybe I just missed it. So we can stick to T1. >>>> >>>> I find the binding mysterious. I have no idea what the points in the space >>>> might be. It seems to rely on the word "concept", as in "returning to the >>>> same baseline situation or concept". But that seems like assuming the >>>> consequent to assume something like a concept, then prove something about >>>> independent loops returning to it. So that can't be the binding point. >>>> Maybe H_x is the binding point, some kind of Δconcept. That's easier to >>>> bind because we *can* measure the changes in the brain/CNS before an after >>>> an experience (e.g. showing you a picture while you're in the MRI machine). >>>> >>>> So my question is: Which of these symbols are bound to measurements we can >>>> (or may one day) be able to make? Without that binding, all this math is >>>> merely decorative, great for hypnotizing your math-envious friends but >>>> useless for science. >>>> >>>> On 10/4/26 7:40 AM, Eric Charles wrote: >>>> > As some of you know, I recently finished a draft of The Book (the one >>>> I have been writing for around 15 years now, with a foundation in Nick's >>>> writing, but extended beyond that). I have circulated drafts to some >>>> readers... so if any of you are interested.... >>>> > >>>> > This appears in the epilogue and speaks directly to the question here: >>>> > >>>> > Wittgenstein laid down a gauntlet to 21^st century philosophers, >>>> “Wovon man nicht sprechen kann, darüber muss man schweigen,” which is >>>> often translated to, “About what one can not speak, one must remain >>>> silent.” This has been broadly taken to indicate that we should shut the >>>> heck up about a wide range of topics regarding the mind, under the >>>> assumption that /all/ such talk is a prime example of >>>> that-about-which-we-cannot-speak. /If/ mind-body dualism is true, either >>>> in its classical form or its modern neural form, then your “feeling of >>>> anger” or “experience of blue” is something that exists only in your >>>> mental space, that only you could access, and my “feeling of anger” or >>>> “experience of blue” is something that exists only in my mental space, >>>> that only I could access. And /if/ that is how minds work, then we can try >>>> to talk about such things all we want, but ultimately it is /impossible/ >>>> for us to talk about them sensibly. Wittgenstein’s various though >> t experiments >>>> > demonstrated this repeatedly. Understood in this fashion — understood >>>> in combination with a /separate/ commitment to dualism — many people have >>>> understood Wittgenstein to be one more philosopher offering proof that a >>>> science of psychology is impossible. >>>> > >>>> > However, given [William] James’s influence on Wittgenstein, we do not >>>> see his dictum as pointing to a fatal flaw. Instead, we understand his >>>> dictum as pointing out a dilemma faced by the field of psychology: Those >>>> interested in psychology must /decide/ whether the phenomena they are >>>> discussing are, inherently, an example of that-of-which-we-/cannot/-speak, >>>> or if the phenomena they are discussing might very well be the types of >>>> thing we /can/ speak about sensibly. >>>> > >>>> > We assume, if you got this far, that you are the type of person deeply >>>> interested in psychological happenings. If, after all the scrutinizing you >>>> can muster, you believe you are interested in something that cannot be >>>> talked about sensibly, then you should find a way to deal with the >>>> inevitable sadness and disappointment of that situation, and then you >>>> should try to find other things to be interested in. We, however, prefer >>>> the other option. We believe psychological phenomena are the types of >>>> things that /can/ be talked about sensibly, and we have endeavored to show >>>> that such a position can work, in practice. We believe there is nothing >>>> within our attempt to understand psychological terms that is fundamentally >>>> different than Lavoisier’s attempt to understand the elements. Like >>>> Lavoisier, we may be right or wrong about particular details, but only in >>>> the types of ways that can be readily corrected over time… because the >>>> things we are talking about can be investigated readily, >>>> > with no magic or vital essence, or other mystical factors having been >>>> included. >>>> > >>>> > We believe that psychological phenomena are the types of things that >>>> /can/evolve, that /can/develop, and that /can/be studied. They are aspects >>>> of behavioral design. And that design deserves to be recognized and >>>> accounted for, to be both described accurately and explained well. >>>> > >>>> > >>>> > >>>> > Best, >>>> > Eric >>>> > >>>> > >>>> > <mailto:[email protected]> >>>> > >>>> > >>>> > On Sun, Oct 4, 2026 at 4:40 AM <[email protected] >>>> <mailto:[email protected]>> wrote: >>>> > >>>> > If this is a polite way to say shut up then please ignore the >>>> rest. What Wittgenstein wants to say by "whereof one cannot speak, thereof >>>> one must remain silent" is apparently that things like anxiety, pain, >>>> nothingness, religion, the meaning of life, subjective experience or >>>> feelings lie outside the logical world and cannot be meaningfully >>>> described in words. That is one way to see it. I see it differently. The >>>> problem is if we remain silent we will never understand it. I prefer Karl >>>> Popper's optimistic view to the depressive pessimism of Wittgenstein. >>>> Popper argued that substantial, real-world problems exist in philosophy, >>>> not just linguistic puzzles as Wittgenstein claims. >>>> > >>>> > For me the things we cannot describe meaningful (yet) are the >>>> interesting ones. We just need to find the right languages, theories, >>>> perspectives or metaphors. One useful metaphor for pain is as you said "an >>>> emergency organization of behavior in which the agent deploys physical and >>>> social defenses of various sorts". If we describe pain as a mechanism to >>>> protect a person it is personal by definition. And yet we all can >>>> understand what it is like to feel pain in general. If we watch the movie >>>> "Marley & Me" for instance we can understand what it is like to lose a >>>> loved dog because most of us have experienced such a loss of a cat or dog >>>> or hamster or horse ourselves ( by the way we have a new horse now named >>>> Mack which looks like this https://www.youtube.com/shorts/VDyNyTsMccg >>>> <https://www.youtube.com/shorts/VDyNyTsMccg> ). >>>> > >>>> > This is in my opinion the solution to the hard problem: the film >>>> maker turns a 1st-person view in a 3rd-person view, the cinema conveys the >>>> 3rd-person view to the persons in the audience, and they are able to >>>> understand roughly what-it-is-like-to-be the lead actor because they see >>>> and hear what it goes through and they know what it feels like to >>>> experience pain, loss, and suffering. According to Wittgenstein cinemas >>>> should not be allowed to exist ("one must remain silent"). But people like >>>> them because they want to know what-it-is-like-to-be someone else. Cinemas >>>> help to solve the hard problem. >>>> > >>>> > -J. >>>> > >>>> > On 2026-10-04 03:35, Nicholas Thompson wrote: >>>> > > You can say that. But you end by reasserting the privacy of >>>> pain. And >>>> > > … >>>> > > >>>> > > Wovon man nicht sprechen kann, darüber muss man schweigen >>>> > > >>>> > > You just cannot have it both ways. >>>> > > >>>> > > N >>>> > > >>>> > > Sent from my Dumb Phone >>>> > > >>>> > > On Oct 3, 2026, at 2:26 PM, [email protected] >>>> <mailto:[email protected]> wrote: >>>> > > >>>> > > Claude was not happy about the theorem proofs that Gemini >>>> created. >>>> > > And Jon also mentioned that they are not waterproof. So I asked >>>> Claude >>>> > > and Gemini if they could write a new proof. They could. I asked >>>> them >>>> > > if we could use gauge theory to explain subjective experience - >>>> gauge >>>> > > invariance to describe objectivity, and symmetry breaking for >>>> > > subjectivity. Both Claude and Gemini said yes, and Claude was >>>> able to >>>> > > formulate a proof that both could agree on (combining two AIs >>>> which >>>> > > have different personalities often gives better results than >>>> just >>>> > > using one). Gemini is curious and creative, Claude is accurate >>>> and >>>> > > precise. The result was a proof for a theorem of a new "field >>>> theory >>>> > > of understanding". I understand only half of the mathematics, >>>> but the >>>> > > physicist who use gauge theory for quantum field theories also >>>> > > understand only half of it. >>>> > > >>>> > > I have uploaded the new proofs here. The theorems are named the >>>> same, >>>> > > only the proofs are different. If they are more convincing than >>>> the >>>> > > old ones, then the idea is to replace the proofs in the >>>> appendix with >>>> > > this version. In my opinion the solution to the hard problem >>>> looks >>>> > > similar to this - in combination with the other things written >>>> in the >>>> > > rest of the book. >>>> > > https://jochenfromm.github.io/PhilosophyBook/QualiaTheorems.pdf >>>> <https://jochenfromm.github.io/PhilosophyBook/QualiaTheorems.pdf> >>>> > > >>>> > > Paradoxically the qualia part that is most mysterious for some >>>> > > researchers is not a problem for others at all. Nick for >>>> example wrote >>>> > > he does not believe in qualia and sees emotions as regulations. >>>> In a >>>> > > post he argued "To me, pain is an emergency organization of my >>>> > > behavior in which I deploy physical and social defenses of >>>> various >>>> > > sorts.... To me, the question, “What is it like to be X?”, has >>>> > > been fully answered when you have said where X-like people can >>>> be >>>> > > found and what they will be doing there... for me, feeling is a >>>> kind >>>> > > of doing, an exploring of the world". I would agree to most of >>>> it but >>>> > > for me what an animal is doing alone gives us only a hint what >>>> it >>>> > > feels. If a horse for example is injured and feels pain, it >>>> does not >>>> > > scream or cry, and you see no tears. Only if you know the >>>> animal well >>>> > > you realize that the behavior is slightly different, for >>>> example by >>>> > > the things it is not doing. I am certain animals feel pain just >>>> as >>>> > > intense as we do, but it is hard to recognize from th >>>> > > eir behavior alone. If it would be better visible, would we >>>> still eat >>>> > > pigs and cows in large amounts (this reminds me of "Why We Love >>>> Dogs, >>>> > > Eat Pigs, and Wear Cows" from Melanie Joy)? >>>> > > >>>> > > -J. >>>> > > >> > >-- >8647 ⊥ ɐןןǝdoɹ ǝ uǝןƃ >ὅτε oi μὲν ἄλλοι κύνες τοὺς ἐχϑροὺς δάκνουσιν, ἐγὰ δὲ τοὺς φίλους, ἵνα σώσω. > >.- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... >--- -- . / .- .-. . / ..- ... . ..-. ..- .-.. >FRIAM Applied Complexity Group listserv >Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom >https://bit.ly/virtualfriam >to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com >FRIAM-COMIC http://friam-comic.blogspot.com/ >archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ > 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
.- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
