Search Ideas

More on the difference between a criterion of rationality and the demand for infallibility: #5336

They’re confused about what BoI actually says about HTV. It says HTV is central to the enlightenment and all rationality, that people use it systematically, that we should choose between explanations based on how hard to vary they are, and so on. According to BoI, HTV is not just another mode of criticism. It’s the central mode explaining progress.

As I say here, I’m personally on board with adopting a weaker form of HTV (or really ETV) where we only treat it as one mode of criticism among many and reject explanations if they have arbitrary components. But again, that’s not what BoI says. The bounty is about taking BoI seriously and rehabilitating HTV as presented there.

I also reply that “Since a fully programmed replacement for HTV already exists that meets all these conditions [mentioned in #5428], I’m not sure it’d be fair to hold HTV to a lesser standard.”

I reply:

The word “reliably” refers to the claim in BoI that ppl search for HTV explanations “systematically”; that they follow some yet-to-be-specified system to form preferences this way.

The system/program, being man-made, would of course be fallible. Still, since BoI says HTV is at the heart of all rational endeavors, the system must be universal. Hence the requirement for any pair.

I also reply that it needs to rank any pair, not every pair.

Why should HTV need to algorithmically rank every pair of explanations to be meaningful?

Because, per BoI, we can’t claim to have understood a computational task if we can’t program it. The formation of rational preferences is a computational task. I say this in my reply.

Not a mathematician but AFAIK, Hilbert basically asked if we could make a machine that answers every math question correctly. The answer to that is ‘no’. But an algorithm for rational decision-making isn’t about that.

I point out that he and I have already discussed that. He has so far ignored the tentative outcome of that discussion, which leaves my stance with zero pending criticisms and his with five (at the time of writing), see #5340.

Related: #5354 and comments.

I don’t care whether people agree with me about this. I care about what’s true, what the ideas imply, as written. Ironically, chapter 1 also says theories can have implications their authors did not intend or may not like (p. 27):

What if you’d rather not know? You may not like these predictions.
Your friends and colleagues may ridicule them. You may try to modify
the explanation so that it will not make them… You will fail.

Citation needed. You of all people should know to cite since you hate when others cite from memory. Don’t be a hypocrite.

I reply:

BoI is full of remarks about preference formation. Doing word searches for “choose between” and “rival explanation”, we can find, for example:

“… we should choose between [policies or theories] not on the basis of their origin, but according to how good they are as explanations: how hard to vary.” (209)

“… how does the ‘dictator’ itself choose between rival explanations …?” (341)

“[The] main use [of experience] is to choose between theories that have already been guessed.” (4)

“The role of experiment and observation is to choose between existing theories…” (32)

I later found more evidence running a word search for the string ‘prefer’:

“Whenever a wide range of variant theories can account equally well for the phenomenon they are trying to explain, there is no reason to prefer one of them over the others, so advocating a particular one in preference to the others is irrational.” (21, emphasis mine)

And chapter 13 is virtually all about forming rational preferences.

Then there’s TCS common-preference finding…

The wrapper provides no immunity. If X’s flaw defeats the proposed use, that flaw can criticise the new policy too. ‘This bridge design collapses, but keep using it’ is criticised by the danger of collapse. Therefore, not almost any criticised idea becomes adoptable this way.

Ben explored and addressed this criticism in more detail, see #5382.

I reply that “Not all programs prompt users for input at runtime”

This is addressed in section 5 of Dennis' paper: The Structure of Rational Thought.

‘How can we rationally proceed when all known ideas have pending criticisms?’
By treating ideas as immutable and discrete, even if content overlaps. For example, as a description of reality, general relativity has pending criticisms. But there are no serious contenders. Instead, take the idea: ‘Although GR has pending criticisms as a true description of reality, it’s an excellent approximation, makes precise predictions, and powers technology such as GPS. So we keep using it until we find a replacement.’ As written, this idea has no pending criticisms, so its advocacy is rational.

Quote missing italics.

This does not identify an error in #5387. It only says ‘there are no serious contenders’ can be criticised, which is exactly what #5387 is doing.”

The criterion excludes comparing the strength of criticisms, not comparison altogether. ‘Serious’ could be binary: does the contender solve the same problems, reproduce GR’s successful predictions, and lack a pending criticism that prevents replacing GR?

‘X is true’ and ‘use X for purpose Y while acknowledging that X is not true’ are different proposals. A criticism of X as a true description does not necessarily criticise its bounded use.

This assumes ‘approximation’ means measurable closeness to an unknown final theory. In context, it actually means that GR’s predictions closely approximate observed gravitational behaviour in its tested domains. We can compare predictions with observations without knowing the final theory.

I read “two strings” literally. The strings are the only case-specific inputs. A fact read from a database, sensor, or anywhere else is another input, whether or not it appears in the function signature.

No function of those two strings can always provide such a ranking.

The bounty says the code must accept two strings as input. It doesn’t say only or at most or exactly two. But I see now that this may not have been clear – if you think it’ll help, I can edit the bounty to say ‘at least two’.

Also, when I wrote the bounty, I meant input as in ‘part of the function signature’. If the function body later reads from a database or prompts the user for input, that doesn’t in and of itself disqualify the submission.

… we cannot avoid guesswork about which changes preserve meaning …

If a good-faith submission prompted the user for answers about changes in meaning or other creative input (within reason), that also would not automatically disqualify the submission. (The proviso “within reason” matters or else bad actors will simply prompt the user to tell the program which explanation is harder to vary. As with all other bounties, any submission will have to meet not just the bounty terms but survive all criticism it may get.)

Here also, I can edit the bounty to clarify that user input at runtime is permissible.

I hope I’ve addressed all reasons the bounty might not be doable, but let me know if I missed something.

Bulk idea

Idea: showing the quickest way to rehabilitate an idea.

Could be a visual feature showing something like the shortest path down a discussion tree where a criticism needs to be addressed.

Or all such branches.

Such a thing has probably been done in graph theory before, could look for existing algorithms.