History of post ‘Choosing between Theories’

Revision 2 · Jun 03, 2025 10:04pm · View this (the most recent) version (v3)

@@ -10,7 +10,7 @@ I’ve skimmed the beginning of the paper by Salmon which Kieren linked to, and

> […] Popper’s account of scientific knowledge involves generalisations and their observational tests.

That sounds like a misrepresentation of Popper’s account of scientific -knowledge, which+knowledge. That account is not about generalizations, but *explanations*, which can’t be obtained by generalizing, only through guesses and criticism.

Salmon references a work I do not own and do not wish to purchase at this time, so it’s hard to say whether he’s wrong or I’m wrong. Which brings us, again, to the question of how to choose between conflicting theories (or claims in this case). The problem is that of *breaking symmetry*, which is an idea by [Elliot Temple](/posts/what-you-should-know-about-elliot-temple), see [Curiosity – Symmetry](https://curi.us/1470-symmetry) and [Curiosity – Epistemology](https://curi.us/1978-epistemology).

Revision 1 · Feb 13, 2025 6:51pm · View this version (v2)

@@ -12,7 +12,7 @@ I’ve skimmed the beginning of the paper by Salmon which Kieren linked to, and

That sounds like a misrepresentation of Popper’s account of scientific knowledge, which account is not about generalizations, but *explanations*, which can’t be obtained by generalizing, only through guesses and criticism.

Salmon references a work I do not own and do not wish to purchase at this time, so it’s hard to say whether he’s wrong or I’m wrong. Which brings us, again, to the question of how to choose between conflicting theories (or claims in this case). The problem is that of *breaking symmetry*, which is an idea by -Elliot Temple,+[Elliot Temple](/posts/what-you-should-know-about-elliot-temple), see [Curiosity – Symmetry](https://curi.us/1470-symmetry) and [Curiosity – Epistemology](https://curi.us/1978-epistemology).

Just knowing that two ideas conflict doesn’t tell you which one is wrong (assuming they really do conflict, and assuming only one of them is wrong). As Elliot writes:

Original · Oct 26, 2021 5:17am · View this version (v1)

# Choosing between Theories

[Kieren](https://twitter.com/krazyander) has [asked](https://twitter.com/krazyander/status/1452804799682920450) about how to choose between (conflicting) theories:

> Well, aside from violent shakings :) a path forward for me would be a Popperian solution to the practical problem of induction (choosing between different theories for practical purposes).
>
> http://www.homepages.ucl.ac.uk/~ucessjb/Salmon.pdf

I’ve skimmed the beginning of the paper by Salmon which Kieren linked to, and found this sentence on p. 117 noteworthy:

> […] Popper’s account of scientific knowledge involves generalisations and their observational tests.

That sounds like a misrepresentation of Popper’s account of scientific knowledge, which account is not about generalizations, but *explanations*, which can’t be obtained by generalizing, only through guesses and criticism.

Salmon references a work I do not own and do not wish to purchase at this time, so it’s hard to say whether he’s wrong or I’m wrong. Which brings us, again, to the question of how to choose between conflicting theories (or claims in this case). The problem is that of *breaking symmetry*, which is an idea by Elliot Temple, see [Curiosity – Symmetry](https://curi.us/1470-symmetry) and [Curiosity – Epistemology](https://curi.us/1978-epistemology).

Just knowing that two ideas conflict doesn’t tell you which one is wrong (assuming they really do conflict, and assuming only one of them is wrong). As Elliot writes:

> “X contradicts Y” means that “Y contradicts X”.

So the ideas are symmetric in that way, and to make progress, you need to find a way of “breaking the symmetry”, as Elliot calls it.

Justificationism, for example, serves as a way to break symmetry. You can ask: which idea has received more support/is better justified etc? If one believes that justificationism can do this job, then one won’t want to get rid of it without replacement. Which is fair, and which is why it’s not enough to point out to people that justificationism is false. They still need a way to break symmetries, so an alternative is needed.

In his book *The Beginning of Infinity* (chapter 1), David Deutsch suggests looking at how “hard to vary” an explanation is. As in: can we make arbitrary changes to an explanation without it losing its ability to explain the phenomenon it purports to explain? This is useful, but when comparing two different explanations, I know of no way to methodically compare their ‘hardness to vary’. In some cases it’s more or less apparent – like when comparing, as Deutsch does, a Greek myth that ‘explains’ the seasons by invoking gods to today’s axis-tilt theory: you could replace one Greek god with another and you’d still be able to explain the seasons. The axis-tilt theory, on the other hand, is hard to vary without it breaking apart. It's not easy to replace the earth's axis with something else and not ruin the explanation in the process. But when comparing other theories, breaking the symmetry using ‘hardness to vary’ can be more difficult, particularly when both seem roughly equally hard or easy to vary.

For example, Kieren is looking for a way to break symmetry between the two opposing claims ‘consciousness requires creativity’ and ‘consciousness does not require creativity’. Deutsch has spoken in favor of the former:

> My guess is that every AI is a person: a general-purpose explainer. It is conceivable that there are other levels of universality between AI and ‘universal explainer/constructor’, and perhaps separate levels for those associated attributes like consciousness. But those attributes all seem to have arrived in one jump to universality in humans, and, although we have little explanation of any of them, I know of no plausible argument that they are at different levels or can be achieved independently of each other. So I tentatively assume that they cannot.

<small>— *The Beginning of Infinity*, chapter 7</small>

For clarity, one cannot be a general-purpose explainer without being creative. So Deutsch argues that creativity (at least the universal kind, if there are non-universal kinds) makes a general-purpose explainer, which in turn leads to consciousness.

That means Deutsch breaks the symmetry in two ways:

1. Consciousness (along with other attributes) seems to have arrived in humans together with humans’ ability to explain things. (I have outlined, in detail, [how this may have happened](/posts/the-neo-darwinian-theory-of-the-mind).)
2. He knows of “no plausible argument” that consciousness is at a different level than creativity or can be achieved without it.

Notably, Deutsch does not use his ‘hard to vary’ criterion to break the symmetry here. He instead invokes a historical guess alongside a lack of alternatives.

Something else you could do is find a contradiction within one of the claims. Or you could find that it conflicts with background knowledge which you currently (and tentatively) deem uncontroversial. (Technically, finding a contradiction is a special case of that, since rejecting contradictions in favor of consistency is an approach that is part of our background knowledge. (Maybe all symmetry breaking involves comparisons with background knowledge in some way?))

For example, I have been asked how I decide between two related claims: that consciousness arises from *all* information processing vs. just *some* information processing (namely the creative kind).

I opted to show that the former claim conflicts with background knowledge: if consciousness arises from *all* information processing, even things like calculators must be conscious. But our best explanations of how calculators work, which are very good and part of our background knowledge in this case, don’t invoke consciousness, so we should conclude that calculators are not conscious. Therefore, it cannot be true that all information processing results in consciousness. We can even *build* calculators – and people do so all the time – without understanding how consciousness works. (Whereas, if Deutsch is right that consciousness arises from creativity, then we can’t, say, build artificial general intelligence without understanding how consciousness works.)

For the related claim that consciousness requires creativity, here’s how I break the symmetry: consciousness is a property of information processing. All information processing people have done so far (except in their minds) is execution only, not creative, and, like with calculators, does not lead to consciousness. Then there’s the problem with Lamarckism: that the mere execution of existing knowledge cannot result in new knowledge. So I ask: if ‘execution-only’ information processing cannot be where consciousness lives, the only place we have left to go is *creative* information processing, do we not? I know of no other kind of information processing. (Maybe there’s a ‘destructive’ kind, in the sense of wiping memory on a computer, but destruction can be automated, so it seems to fall under the execution-only kind.)

In other words, we simply run out of alternatives. There seem to be only two: execution-only information processing and creative information processing. Our best explanations of execution-only information processing do not invoke consciousness, so the only place left to go is *creative* information processing. And with the latter, there’s much more room left since we don’t really understand creative information processing at all while we do understand execution-only information processing pretty well.

I’m not sure breaking the symmetry can be boiled down to a recipe. I’m guessing it is itself a creative act and you can always find new ways to do it. In the context of my [neo-Darwinian approach to the mind](/posts/the-neo-darwinian-theory-of-the-mind), the idea that breaks symmetry is the one that has spread through the mind at the expense of its rivals, and whose total number of copies is therefore greater than that of any one of its rivals.