Crisis of Faith case study: beyond reductionism?

post by MalcolmOcean (malcolmocean) · 2023-06-08T06:11:24.246Z · LW · GW · 9 comments

Contents

  📖 Reading notes
  ↳12017-01-17 evening
    First pass: skepticism, committing to the virtues of rationality
      ...just read a couple more pages. Am now on page 7. Still very challenging.
    Second pass: why is reductionism important to me?
    Third pass: discovering curiosity as I flip around the pages
    Fourth pass: this makes sense now. I can start to listen without feeling like it's going to make me crazy
    Reflection, 6+ years later
    Incidentally: the central point (I think?) of the book
    Comments?
None
9 comments

What follows are some notes to self that I took while reading the first 7 pages of Stuart A. Kauffman's book Reinventing the Sacred. This was over 6 years ago, in January 2017.

I read those pages multiple times, pausing whenever I had some objections arise, and journalling about the process. I ended up having a bit of a Crisis of Faith [LW · GW] with it, and something loosened, but I didn't actually return to the book until years later. The objections were largely coming up based on what I'd read in the Sequences and learned elsewhere from rationalists. I remember thinking at the time it would be a good case study for LessWrong, but I didn't get around to sharing it until now.

It's kinda funny/awkward pasting this in—I've changed a lot in the past years, both in terms of how I write and how I related to all of this stuff. But since the whole thing is about my perspective at the time, there's no point in trying to change any of that, so I've basically only edited it for clarity, grammar, etc. I did take out a couple tangents. The section titles I wrote now with the benefit of hindsight.

If you get bored towards the start, you might skim til it picks up; I think it gets more interesting as it goes on, but it seemed important to include my whole process of course.

📖 Reading notes

↳12017-01-17 evening

First pass: skepticism, committing to the virtues of rationality

I got a few pages into this book and it was already quite challenging, in the sense of "challenging to take the author seriously as having something worth saying". He struck me as being the kind of person that caused Eliezer to rant about "emergence".

Main issues so far are:

noticed myself thinking something akin to "he's not exactly doing a good job of convincing me he's not nuts"

also finding myself thinking "is [transcending and including reductionism] oxymoronic?" it seems like it might be. have just asked this on fb. we'll see what people say

...just read a couple more pages. Am now on page 7. Still very challenging.

I'm noting though that I maybe have some stale beliefs here though!

It feels important to separate "what beliefs LW taught me" and "how LW taught me to think". In particular,

So what *do* I believe (about reductionism), that I think Kauffman doesn't?

One thing to remember is that a lot of the reason that rationalists are suspicious of people who use words like emergence and who make arguments for elements of consciousness being ontologically basic is that these people are typically engaging in some form of mind projection fallacy, plus motivated cognition of various kinds.

I think I can say "enough" for this line of thinking. I think my conclusions can be summarized as "My worldview is ontologically reductionistic but absolutely *not* methodologically reductionistic. (In particular, I reject mental agents as ontologically basic). Kauffman appears to confuse these two kinds of reductionism, which is perhaps somewhat annoying but he may still have much to teach me about methodological reductionism."

Second pass: why is reductionism important to me?

I'm going to start at the beginning of the book again, with all of this in mind, and see how I feel reading it :P

...re-reading the first few pages, it's clear that my summary above was very relevant but not enough to get me into smooth sailing through this book! Having personally clarified the distinction between ontological and methodological reductionism, it's very clear that he is aware of both and is rejecting both as limited. On page 3, which I clearly didn't read closely enough the first time:

It turns out that biological evolution by Darwin's heritable variation and natural selection cannot be "reduced" to physics alone. It is emergent in two senses". The first is epistemological, meaning that we cannot from physics deduce upwards to the evolution of the biosphere. The second is ontological, concerning what entities are real in the universe. For the reductionist, only particles in motion are ontologically real entities. Everything else is to be explained by different complexities of particles in motion, hence are not real in their own ontological right.

But okay, let's back up a bit: why is reductionism important to me? How would I fail if I allowed non-reducible realities?

[Note from 2023: That question ↑ seems to me to be central to how I was able to make progress here. I shifted from defending my view to focusing on what criteria I was looking for in whatever view I might want to adopt. I shifted from "reductionism is correct" to "[my] reductionism doesn't make these particular errors".]

When I think about this, it becomes relatively clear that a better description for the whole thing is more like:

Third pass: discovering curiosity as I flip around the pages

(I still don't quite even know what it *means* for "a couple in love on the banks of the Seine" to be more than whatever it is they are made of. but maybe my concept of "is" is still reductionistic on some level? and maybe there's another way to conceive of "is"?)

let's start at the beginning again, having had that insight. even though I didn't make it to p7 again. I think maybe I'll only keep reading when I stop feeling stuck on these pages. It feels cool that I'm relating to that stuckness excitedly, as a sign of learning/understanding available, etc.

Fourth pass: this makes sense now. I can start to listen without feeling like it's going to make me crazy

(aside: it is certainly NOT helping matters, from my perspective, that Kauffman is bringing up the science vs religion debate here. it seems very very very clear to me that religion is an incidental part of human history)

p3: "a central implication of this new worldview is that we are co-creators of a universe, biosphere, and culture of endlessly novel complexity"

p4: "agency has emerged in evolution and cannot be deduced by physics"

[I need to head to bed, even though I still haven't made it past p7]

wanting to note 2 things:

  1. post-reductionism isn't about not being able to decompose complex systems into parts, but rather about not being able to put them back together again (although there may also be systems that are entangled in such a way that part-ness is impossible to reasonably define)
  2. we might want to distinguish between "ontologically basic" and something else. "ontologically non-reducible", maybe
    1. the point being that you still need some sort of model that says *that* the phenomenon is made of other, more basic parts, even if no model can say *how*
    2. this prevents errors like the supernaturalist ones, and maybe also others
    3. like just because "agency" or "life" or something is a thing in one's ontology, doesn't mean you can have it for free without some complex system that is yet made of atoms
    4. something about construction feels relevant here

Reflection, 6+ years later

It's hard to say what impact this experience I documented above had on that process, but it seems to me likely that even though I didn't read further in Reinventing the Sacred at the time, that the shift from guardedness and defensiveness towards curiosity helped me appreciate post-reductionist insights over the following years, from other books and from just looking at the world.

I no longer consider myself a reductionist, and I don't think I'm making any particular anti-reductionistic errors. I definitely don't think of myself as a post-reductionist, although I suppose that label would apply. It seems to me to be inherently counterproductive to consider myself to be anything that ends in "-ist", except for guitarist and pianist.

I would also no longer say that I think that " 'how LW taught me to think' [...] is roughly-speaking lacking in error." 😆 It feel slightly embarrassed to have written that, to be honest. Literally "Less Wrong", not "we have it all figured out". But there has at times been a "we have it all figured out" vibe around here, even though the official policy is more fallible. My guess is that basically I felt that I had to say that, even in my private notes, in order to feel free to question Eliezer on reductionism.

I'm not entirely sure why I waited 6 years to post this—I had a lot on my plate and this did take a couple hours of editing out of workflowy-bullets... but it seems likely that one of the reasons is that at the time I was probably a bit nervous given that reductionism still seemed cool at the time and I didn't want to subject myself to a ton of public criticism. Now it feels like there's a large fraction of LWers and other friends or peers of mine who've gone through a similar shift, whether from reading David Chapman or whatever other life experiences. Not to say you can't still be cool if you consider yourself a reductionist! But I will say that I experience myself substantially saner and more integrated on the other side of that shift, so I do recommend it.

Anyway, the juiciest thing about this case study, from my current perspective, is that it's a brilliant example of something I didn't come to understand until years later, about trust and distrust. Essentially, Kauffman tripped a few warning signs for me pretty early on that put me in a skeptical stance towards him. Both skeptical towards what he had to say and skeptical towards the idea of it being worth continuing to read the book at all. Guarded, rather than listening.

It's common for rationalists (and, I think, STEM-folk more broadly) to flinch-dismiss things that pattern-match to something they think they've rejected already, without recognizing the pre-post fallacy [LW · GW] (the stage beyond yours often at first looks like the stage before yours). I now see this urge to dismiss reading something as a waste of time as... not a very good heuristic. The recognition that it would be a waste of time to debate someone about something? Great. But if reading or watching something intrigues me, I take that as a sign that there's something there that my system wants to understand.

Anyway, in this case, what allowed me to listen was when I owned my distrust as coming from something I cared about (not making certain anti-reductionistic errors) and then sought to read the book while keeping that in mind. I shifted from making a positive/assertive claim "reductionism is correct" (therefore this piece saying "reductionism is wrong" is wrong) to something more like "well I'm not going to trade reductionism for something worse" (and this is currently failing to prove that it's not worse, but if it were better I would want it).

This resonates with something Logan wrote about courage & defensiveness:

"What is defensiveness trying to accomplish?" According to my working model, defensiveness is trying to preserve access to valuable information you can see from your current perspective. Defensiveness knows that sometimes, when you accidentally fall toward another vantage point, you might get stuck there, forgetting what you used to see and know and care about. Defensiveness would like for you to only gain information, and never lose it.

In 2021 I quote-tweeted someone who said "Books should be a canvas on which you paint your reaction" and mentioned my experience of RtS:

In 2017 I spent 2h re-reading and taking notes on the first 7 pages of a book because I found it so disturbing to my ontology and was trying to integrate it.

(Rereading it last year I now think it's way unnecessarily belligerent but basically correct in its content.)

Re-reading it today in 2023 I didn't even find it particularly belligerent, although it still seems to me like not the best gentle on-ramp for reductionists.

Incidentally: the central point (I think?) of the book

In 2020, I picked up the book again, started once more at the beginning, and—seeing an early reference to Chapter 10: Breaking the Galilean Spell—I decided to jump to the middle of the book. There I came to understand his thesis quite directly, which is basically that while the laws of physics absolutely constrain what is possible on higher levels (ie no supernatural stuff) they can't predict those higher levels for reasons relating to combinatorics & evolution—there are too many possibilities, each of which depends on unexpected remixes of other possibilities. He writes in chapter 10:

If a scientific law is (as the physicist Murray Gell-Mann concisely put it) a compact statement, available beforehand, of the regularities of of a process, then the evolution of the biosphere is partially beyond scientific law.

Here's a recent piece of writing from an astrobiologist & theoretical physicist outlining Assembly Theory, which (it seems to me) operationalizes one part of this quite precisely. They can measure in any sample of matter, whether there are molecules present that are above the threshold that can only come into existence via some life-like process. This is a simple way in which all the explanations don't point downward—some point upward. Why is there a ton of this molecule here? Because it was useful for something bigger, whether a nematode or a pharmaceutical company. There is no other way to account for this particular molecule existing in such quantities given, again, combinatorics.

These solid samples of stone, bone, flesh and other forms of matter were dissolved in a solvent and then analysed with a high-resolution mass spectrometer that can identify the structure and properties of molecules. We found that only living systems produce abundant molecules with an assembly index above an experimentally determined value of 15 steps. The cut-off between 13 and 15 is sharp, meaning that molecules made by random processes cannot get beyond 13 steps. We think this is indicative of a phase transition where the physics of evolution and selection must take over from other forms of physics to explain how a molecule was formed.

I've become increasingly fascinated with applications of universal law to biology (what can we say about life that we could confidently know would be true on all planets?) which often comes from physicists who have also studied complexity theory and emergence (another great example being Scale by Geoffrey West, which looks at geometry and scaling laws and finds fascinating invariances). But it's still just constraints; it can't say what will happen, only what won't.

I feel like in the years since 2017, I've had some key insights about the evolutionary qualities of learning (and what you could call belief-updating) and I've come to appreciate what appear to be some of its invariances. One of them is that it's both necessary to surface any defensiveness/distrust AND to find a way to simultaneously be curious, in order to have deep paradigm-shifting learning, as I happened to do here without fully understand that. If you don't surface any existing defensiveness/distrust, then even if you do learn, the new learning will be compartmentalized from the old one and at odds with it (though this can maybe be integrated later). If you surface the distrust but don't find a way to honor it, you won't be able to listen at all. Something about this seems deeply true to me in both a logical sense and in that it matches a lot of my experiences quite precisely, but I'd love to be able to formalize and verify it even further. This is building on, among other things, Unlocking the Emotional Brain [LW · GW]. I've written a bit more about my perspective on it here.

Kauffman says: the reason "evolution" displaced "God" is that it was an idea worthy of displacing God, and the more I've come to understand about evolution and how it applies on pre-biological scales and memetic scales (David Deutsch is a great source there, who makes a solid rationalist case for post-reductionism) the more I've come to appreciate that evolution is worthy of displacing God not just in terms of explanatory power but in terms of potential for sacred devotion as well—without naive or ignorant faith!

Comments?

I'd love to hear what's on your mind. Some prompts you could respond to:

9 comments

Comments sorted by top scores.

comment by Viliam · 2023-06-08T11:49:05.580Z · LW(p) · GW(p)

Reductionist: "Apples are made of atoms."

Post-reductionist: "I agree. However, it often makes sense to think about 'apples' instead of 'huge sets of atoms', when we are saying things such as 'if you plant an apple seed to the ground, an apple tree will grow out of it'. Trying to express this in the terms of individual atoms would require insane computational capacity. The patterns at high level are repeating sufficiently regularly to make 'apple' a useful thing to talk about. Most of the time, it allows us to make predictions about what happens at the high level, without mentioning the atoms."

Reductionist: "Yes, I know. I think this is called 'a map and the territory'; the territory is made of atoms, but apples are a useful concept."

Post-reductionist: "From your perspective, apples are merely a useful fiction. From my perspective, atoms are real, but apples are real too."

Reductionist: "Let's taboo 'real'. I already admitted that apples are a useful abstraction, in everyday situations. There are also situations where the abstraction would break. For example if we asked 'how many DNA bases can we change before the apple stops being an apple?', there is no exact answer, because the boundary of the concept is fuzzy; the objects currently existing in our world may be easy to classify, but it is possible to create new objects that would be controversial. By the way, talking about the 'apple DNA' already brings the atomic level back to the debate.

On the other hand, a hypothetical superintelligent being with an insane computing capacity (probably would need to exist in a different universe) might be able to describe the apple in the terms of atoms, something like 'apple is a set of atoms that has this super-complicated mathematical property'; it might also be able to describe in terms of atoms how the apple trees grow. If it did many computations regarding apples, it might develop an abstraction for 'apple', in a similar sense how humans have an abstraction for 'prime numbers'. However, we are not such beings, and I am definitely not pretending to be one. Just saying that it is possible in principle."

Post-reductionist: ...well, I don't know what to write here to pass the ITT.

Replies from: Valentine, Ape in the coat
↑ comment by Valentine · 2023-06-08T14:25:13.548Z · LW(p) · GW(p)

Post-reductionist: ...well, I don't know what to write here to pass the ITT.

I can't speak for the post-reductionist view in general. But I can name one angle:

Atoms aren't any more real than apples. What you're observing is that in theory the map using atoms can derive apples, but not the other way around. Which is to say, the world you build out of atoms (plus other stuff) is a strictly richer ontology — in theory.

But in a subtle way, even that claim about richer ontology is false.

In an important way, atoms are made of apples (plus other stuff). We use metaphors to extend our intuitions about everyday objects in order to think about… something. It's not that atoms are really there. It's that we use this abstraction of "atom" in order to orient to a quirky thing reality does. And we build that abstraction out of our embodied interactions with things like apples.

Said more simply: infants & toddlers don't encounter atoms. They encounter apples. That's the kind of thing they use to build maps of atoms later on.

This is why quantum mechanics comes across as "weird". Reality doesn't actually behave like objects that are interacting when you look closely at it. That's more like an interface human minds construct. The interface breaks down upon inspection, kind of like how icons on a screen dissolve if you look at them through a microscope.

(This confusion is embedded in our language so it's hard to say this point about mental interfaces accurately. E.g., what are "human minds" if objects aren't actually in the territory? It's a fake question but I've never found a way to dispel it with accurate language. Language assumes thing-ness in order to grammar. It's like we can only ever talk about maps, where "territory" is a map of… something.)

So really, we already have examples of constructing atoms out of apples (and other things). That's how we're able to talk about atoms in the first place! It's actually the inverse that's maybe impossible in practice.

(…as you point out! "…a hypothetical superintelligent being with an insane computing capacity (probably would need to exist in a different universe) might be able to describe the apple in the terms of atoms….")

I think the core issue here is that the standard reductionist view is that maps can be accurate. Whereas my understanding of post-reductionism basically says that "accurate" is a type error arising from a faulty assumption. You can compare two maps and notice if they're consistent. And you can tell whether using a map helps you navigate. But things get very confusing at a basic philosophical level when you start talking about whether a map accurately describes the territory.

Replies from: Viliam
↑ comment by Viliam · 2023-06-09T16:28:06.936Z · LW(p) · GW(p)

What are the new predictions that post-reductionism makes?

Replies from: Valentine
↑ comment by Valentine · 2023-06-10T13:48:31.402Z · LW(p) · GW(p)

I don't know if it does. It's not that kind of shift AFAICT. It strikes me as more like the shift from epicycles to heliocentrism. If I recall right, at the time the point wasn't that heliocentrism made better predictions. I think it might have made exactly the same predictions at first IIRC. The real impact was something more like the mythic reframe on humanity's role in the cosmos. It just turned out to generalize better too.

Post-reductionism (as I understand it) is an invitation to not be locked in the paradigm of reductionism. To view reductionism as a tool instead of as a truth. This invites wider perceptions which might, in turn, result in different predictions. Hard to say. But the mythic impact on humanity is still potentially quite large: the current reductionist model lends itself to nihilism, but reality might turn out to be vastly larger than strict reductionism (or maybe any fixed paradigm) can fully handle.

Replies from: Viliam
↑ comment by Viliam · 2023-06-11T19:49:32.155Z · LW(p) · GW(p)

This sounds to me like: "I believe the same things as you do, but on top of that I also believe that you are wrong (but I am not)."

Which in turn sounds like: "I do not want to be associated with you, regardless of how much our specific beliefs match".

↑ comment by Ape in the coat · 2023-06-08T12:16:23.357Z · LW(p) · GW(p)

Yeah, that's similar to my impression as well. Frankly, I think that "real" is a terribly confusing category which we owe dozens of fruitless philosophical arguments. And as soon as we switch to a much more useful map-territory distinction they mostly dissolve.

comment by Kaj_Sotala · 2023-06-08T08:01:25.213Z · LW(p) · GW(p)

I skimmed this a bit, but found it a little hard to follow along your notes without knowing what those first seven pages of the book were actually saying. I felt that in order to understand your thought process, I would first need to read your notes and reconstruct the argument the book was making from your responses to it, and then read the post a second time once I had some sense of what it was a reaction to. But that seems pretty error-prone; it would help a lot to have a summary of those seven pages before your notes.

comment by Ape in the coat · 2023-06-08T12:00:03.700Z · LW(p) · GW(p)

I'm under the impression that the difference between reductionism and postreductionism or rationalism and postrationalism, for that matter, is not in the realm of ideas but more like political affiliations and aesthetics.

Things you've said about the usefulness of multi-level models isn't something reductionists disagree with. There is no crux here so I don't really see an opportunity for the genuine crisis of faith. Could you explicitly state a couple of object-level beliefs you had previously when you identified as reductionist and the corresponding ones, that you have now as postreductionist, so that I could see the difference?

The procedure where you keep steelmanning and sanewashing someone's argument to the point where they start making sense is a bit questionable. I can see potential use cases, like when you exist in adversarial epistemic conditions with highly polarized social bubbles, and every argument of the other side arrives to you distorted to be weak and crazy. Then doing your best to find a reasonable sounding version of the argument and try to check whether it was its original form from someone who actually holds this view is a good idea. But what you seem to have done - isn't that.

The way I see it, you've distorted the original ontological argument, turning it into a methodological one. The new version of the argument makes sense, it can be a valid critique of some behaviour patterns inside the community. But this isn't what the author of the argument originally meant! You try to find some common ground and when you can't, you keep misrepresenting his views to yourself until you can find the common ground with this misrepresented version. I guess as a result, you are reading a much more interesting and nuanced book then the author wrote. And most credit for that should go to you, not the author.

comment by shminux · 2023-06-08T07:26:28.318Z · LW(p) · GW(p)

I am quite sure I am missing your point, but just in case...

Why is there a ton of this molecule here? Because it was useful for something bigger, whether a nematode or a pharmaceutical company. There is no other way to account for this particular molecule existing in such quantities given, again, combinatorics.

You might be vaguely gesturing at a weak emergence? (Strong emergence is not a thing.) When you are not a Laplace's demon, but a bounded embedded agent trying to make accurate predictions, that might be one useful way to abstract your observations to simplify this job. If so, emergence is fully compatible with reductionism.