Hacker News Reader

Comment by felipeerias | Hacker News Reader

The problem with trying to reason everything from first principles is that most things didn’t actually came about that way.

Both our biology and other complex human affairs like societies and cultures evolved organically over long periods of time, responding to their environments and their competitors, building bit by bit, sometimes with an explicit goal but often without one.

One can learn a lot from unicellular organisms, but won’t probably be able to reason from them all the way to an elephant. At best, if we are lucky, we can reason back from the elephant.

View on HN

>The problem with trying to reason everything from first principles is that most things didn’t actually came about that way.

This is true for science and rationalism itself. Part of the problem is that "being rational" is a social fashion or fad. Science is immensely useful because it produces real results, but we don't really do it for a rational reason - we do it for reasons of cultural and social pressures.

We would get further with rationalism if we remembered or maybe admitted that we do it for reasons that make sense only in a complex social world.

Yes, and if you read Popper that’s exactly how he defined rationality / the scientific method: to solve problems of life.

>if you read Popper

Thanks, I might actually go do this :) I recently got exposed to a very persuasive form of "rationalism is a social construct" by reading "Alchemy" by Rory Sutherland. But a theme in these comments is that a lot of these ideas are just recycled from philosophers and that the philosophers were less likely to try and induct you into a cult.

I read the book too! I really enjoyed it and I think they Popper and Sutherland would go along quite well.

Both advocate the principle of putting out a bold conjecture, test, learn, adapt.

A lot of people really need to be reminded of this.

I originally came to this critique via Heidegger, who argues that enlightenment thinking essentially forgets / obscures Being itself, a specific mode of which you experience at this very moment as you read this comment, which is really the basis of everything that we know, including science, technology, and rationality. It seems important to recover and deepen this understanding if we are to have any hope of managing science and technology in a way that is actually beneficial to humans.

Brilliant way to phrase this idea! I think its incredible that we can even manage an imperfect way to escape our "social" brain. It's clearly very powerful - math, and thus all of science, only exist because of how weird predaliction to thinking abstractly enough to (sorta) break our social bounds. At least when thinking in these terms, I am sure you can poke holes in it.

Reducibility is usually a goal of intellectual pursuits? I don't see that as a fault.

What the person you are replying to is saying that some things are not reducible, i.e. the the vast array of complexity and detail is all relevant.

That's a really hard belief to justify. And what implications would that position have? Should biologists give up?

Biologists don't try to reason everything from first principles.

Actually, neither do Rationalists, but instead they cosplay at being rational.

> Biologists don't try to reason everything from first principles.

What do you mean? The biologists I've had the privilege of working with absolutely do try to. Obviously some work at a higher level of abstraction than others, but I've not met any who apply any magical thinking to the actual biological investigation. In particular (at least in my milieu), I have found that the typical biologist is more likely to consider quantum effects than the typical physicist. On the other hand (again, from my limited experience), biologists do tend to have some magical thinking about how statistics (and particularly hypothesis testing) works, but no one is perfect.

Setting up reasoning from first principles vs magical thinking is a false dichotomy and an implicit swipe.

Ok, mea culpa. So what distinction did you have in mind?

Reasoning is one information-processing process, performed by humans, with bounds on what it can accomplish. It works in a limited context and is inherently incomplete and imperfect. Other non-logical processes, emergent processes, parallel processes such as evolution, process information in ways reasoning cannot. It perhaps should not be surprising that we may have internal systems of understanding that follow these principles, instead of only those of logic or reasoning.

Reasoning from first principles cannot span very far in reality, as for starters the complexity of the argument quickly overwhelms our capacity for it. Its numerous other limits have been well-documented.

Logicomix, Gödel Escher Bach are some common entry points.

>Gödel Escher Bach

I'm kinda new here but am surprised I haven't seen this book mentioned more. Maybe I just haven't seen it or it's old news but it seems right up HNs alley.

It would imply that when dealing with complex systems, models and conceptual frameworks are, at the very best, useful approximations. It would also imply that it is foolhardy to ignore phenomena simply because they are not comprehensible within your preferred framework. It does not imply biologists should give up.

How reducible is the question. If some particular events require a minimum amount of complexity, how to do you reduce it below that?

I think that chemistry, physics, and mathematics, are engaged in a program of understanding their subject in terms of the sort of first principles that Descartes was after. Reduction of the subject to a set of simpler thoughts that are outside of it.

Biologists stand out because they have already given up on that idea. They may still seek to simplify complex things by refining principles of some kind, but it's a "whatever stories work best" approach. More Feyerabend, less Popper. Instead of axioms they have these patterns that one notices after failing to find axioms for a while.

On the other hand, bio is the branch of science with a single accepted "theory of everything": evolution.

Evolution is a theory of the origin of species via natural selection of heritable traits; evolution is not a theory of biogenesis, the origin of life itself.

That's a fine counterexample to "theory of everything", and fertile ground for spirited debate. But I think it's a distinction thats relevant to <1% of the work that biologists do, so like... does it matter?

Yeah, I almost wrote ‘nearly have a theory of everything’ for that reason, but decided it wasn’t worth the extra words. We have a few plausible outlines of how life started, and IMO it doesn’t really matter all that much which one(s) actually happened. When we ourselves are doing biogenesis, there’s no requirement that it has to happen the way it happened before. It would be interesting to know, though, so if in your estimation we don’t have a theory of everything because of that, I’m okay with that.

Several different definitions are being bandied about. If you think of reduction as understanding a material system in terms of its components, biology is now reductionist, having abandoned vitalism.

Concretely we know that there exist irreducible structures, at least in mathematics: https://en.wikipedia.org/wiki/Classification_of_finite_simpl...

The largest of the finite simple groups (themselves objects of study as a means of classifying other, finite but non-simple groups, which can always be broken down into simple groups) is the Monster Group -- it has order 808017424794512875886459904961710757005754368000000000, and cannot be reduced to simpler "factors". It has a whole bunch of very interesting properties which thus can only be understood by analyzing the whole object in itself.

Now whether this applies to biology, I doubt, but it's good to know that limits do exist, even if we don't know exactly where they'll show up in practice.

That's not really true, otherwise every paper about it would be that many words long. The monster group can be "reduced" into its definition and its properties which can only be considered a few at a time. A person has a working memory of three to seven items.

"Reductionist" is usually used as an insult. Many people engaged in intellectual pursuits believe that reductionism is not a useful approach to studying various topics. You may argue otherwise, but then you are on a slippery slope towards politics and culture wars.

I would not be so sure. There are many fields where reductionism was applied in practice and it yielded useful results, thanks to computers.

Examples that come to mind: statistical modelling (reduction to nonparametric models), protein folding (reduction to quantum chemistry), climate/weather prediction (reduction to fluid physics), human language translation (reduction to neural networks).

Reductionism is not that useful as a theory building tool, but reductionist approaches have a lot of practical value.

> protein folding (reduction to quantum chemistry),

I am not sure in what sense folding simulations are reducable to quantum chemistry. There are interesting 'hybrid' approaches where some (limited) quantum calculations are done for a small part of the structure - usually the active site I suppose - and the rest is done using more standard molecular mechanics/molecular dynamics approaches.

Perhaps things have progressed a lot since I worked in protein bioinformatics. As far as I know, even extremely short simulations at the quantum level were not possible for systems with more than a few atoms.

That's a practical limitation. There's no barrier in principle to a complete reduction.

I meant that the word "reductionist" is usually an accusation of ignorance. It's not something people doing reductionist work actually use.

But that common use of the word is ignorant nonsense. So, yes, someone is wrong on the internet. So what?

The context here was a claim that reducibility is usually a goal of intellectual pursuits. Which is empirically false, as there are many academic fields with a negative view of reductionism.

'Reductionist' can be an insult. It can also be an uncontroversial observation, a useful approach, or a legitimate objection to that approach.

If you're looking for insults, and declaring the whole conversation a "culture war" as soon as you think you found one, (a) you'll avoid plenty of assholes, but (b) in the end you will read whatever you want to read, not what the thoughtful people are actually writing.

'Reducibility' is a property if present that makes problems tractable or possibly practical.

What you are mentioning is called western reductionism by some.

In the western world it does map to Plato etc, but it is also a problem if you believe everything is reducible.

Under the assumption that all models are wrong, but some are useful, it helps you find useful models.

If you consider Laplacian determinism as a proxy for reductionism, Cantor diagonalization and the standard model of QM are counterexamples.

Russell's paradox is another lens into the limits of Plato, which the PEM assumption is based on.

Those common a priori assumptions have value, but are assumptions which may not hold for any particular problem.

Ok. A lot of things are very 'reducible' but information is lost. You can't extend back from the reduction to the original domain.

Reduce a computer's behavior to its hardware design, state of RAM, and physical laws. All those voltages make no sense until you come up with the idea of stored instructions, division of the bits into some kind of memory space, etc. You may say, you can predict the future of the RAM. And that's true. But if you can't read the messages the computer prints out, then you're still doing circuits, not software.

Is that reductionist approach providing valuable insight? YES! Is it the whole picture? No.

This warning isn't new, and it's very mainstream. https://www.tkm.kit.edu/downloads/TKM1_2011_more_is_differen...

Reply on news.ycombinator.com