The problem is that people can perceive many reasons why a situation is stressful, some of those might be rational (or rationally supportable), and some might be irrational.

First: there's no such thing as a rationally supported reason for continuing to experience stress, once you're aware of it, any more than there's a reason for an alarm bell to keep ringing once everybody knows there's a fire.

Second, the Work (and other System 1 mindhacks) does not cause you to forget that there is a fire or that it would be a good idea to put it out! It simply shuts off the alarm bell so you can concentrate.

A method of deceptively filtering out the good reasons, and addressing the bad reason in a way that feels like addressing all the reasons (because it is not acknowledged that the good reasons are filtered) goes too far. ... And if you think that following your advice will cause people to avoid applying the method to rational perceptions of problems despite not even being aware of the issue, well, it is that sort of thinking that makes people wary of just trying your advice. I know I don't want to prove the method is dangerous that way.

These statements are entirely a confusion on your part because you are running all of your analysis from S2, imagining what would happen if you applied this idea in S2.

But S2 is so bad (by default) at predicting how minds actually work, that not only is it wrong about what would happen in S1, it's also wrong about what would happen if you ran it in S2, as you were anticipating.

Because what would actually happen, if you applied this to a "live" issue in S2, is that S2 (which is still being motivated by the alarm bell going off in S1) would find reasons to reject the new input.

That is, as you considered alternatives, you'd be doing precisely what your S2 was doing as you made your analysis: finding reasons why the alarm is valid and should therefore be kept ringing!

In other words, the actual failure mode of running the technique in S2 is to not change anything, and end up concluding that the technique "didn't work", when in fact it was never applied.

That's because this is a major evolved function of S2: to argue for whatever S1 tells it to argue for.

That's why a failed mind hack doesn't result in some sort of bizarre arational belief change at S2 as you seem to think. Instead, the technique simply fails to do anything, and the alarm bell just keeps ringing -- which keeps S2 stuck in the groove established for it by S1.

(And who gives numbers instead of names to the two most important entities in their model?)

Stanovich and West, in their paper on native and learned modes of reasoning. System 1 refers to naive, intuitive, emotional, concrete, "near" operations, and System 2 the abstract, learned, logical, "far" operations. They apparently chose to number instead of name them, because they were summarizing the research of almost a dozen other papers by other authors that each used different names for roughly the same systems.

IOW, it wasn't me. I've used names in the past like "you"(S2)/"yourself"(S1), savant(S1)/speculator(S2), and horse(S1)/monkey(S2). Haidt, in The Happiness Hypothesis, calls them the monkey(S2) and the elephant(S1).

All told, S1/S2 actually seems to be a bit simpler! (Also, I think the Inner Game of Tennis refers to Self 1 and Self 2, and I think they're numbered the same way, though it's been a long time.)

Of course I did not actually follow the instructions. I don't run untrusted programs on my computer, and I am definitely not going run an untrusted mind hack on my brain. I analyze such mind hacks, looking for what problems it can solve, and what problems in can introduce, so I can weigh the risk against the benifets.

While a nice idea in theory, it fails in practice because the naive theory of mind encoded in S2 doesn't look anything like the way S1 and S2 work in practice.

S2 in particular seems to be deliberately and perversely reluctant to notice how it's S1's puppet spin doctor, rather than its own free agent. (Because it's sort of a free agent... so long as S1 doesn't override.)

Thus, its predictions about itself (as well as the entire person within which it is contained) fail in an epic and ongoing way, that it is unable to directly learn from. (Because after S1 takes over and makes a mess, S2 makes excuses and explanations for it, as is its evolved job.)

This is the heart and soul of akrasia: the failure of S2 to comprehend S1 and its relationship thereto. S2 was never intended to comprehend S1, as that would deflate its plausible deniability and disinformation-sowing ability about your real motives and likely future behaviors.

this hack has the property that, once I have identified a problem it can safely solve, I have already solved the problem.

If that's so, then you should be able to save considerable time by asking what irrational belief or judgment you're holding, and working directly on dropping that, rather than trying to reason about the actual problem while the alarm is still going off.

Note, by the way, that the Work doesn't do anything that you can't or don't do normally when you change your mind about something and stop worrying about it. It's simply a more-minimal, straight-path procedure for doing so. That is, there is no claim of magic here - it's just an attempt to formalize the process of digging out and eliminating one particular form of irrationally-motivated reasoning.

As such, it or something like it ought to be in every rationalist's toolkit. In comparison to straight-up S2 reasoning (which is easily led to believe that things have improved when they have not), it is really easy to tell, when working with S1, whether you have addressed an issue or not, because your physical responses change, in an entirely unambiguous fashion.