Back Up and Ask Whether, Not Why

[Eliezer Yudkowsky]

Followup to: The Bottom Line

A recent conversation reminded me of this simple, important, and difficult method:

When someone asks you “Why are you doing X?”,
And you don’t remember an answer previously in mind,
Do not ask yourself “Why am I doing X?”.

For example, if someone asks you
”Why are you using a QWERTY keyboard?” or “Why haven’t you invested in stocks?”
and you don’t remember already considering this exact question and deciding it,
do not ask yourself “Why am I using a QWERTY keyboard?” or “Why aren’t I invested in stocks?”

Instead, try to blank your mind—maybe not a full-fledged crisis of faith, but at least try to prevent your mind from knowing the answer immediately—and ask yourself:

“Should I do X, or not?”

Should I use a QWERTY keyboard, or not? Should I invest in stocks, or not?

When you finish considering this question, print out a traceback of the arguments that you yourself considered in order to arrive at your decision, whether that decision is to X, or not X. Those are your only real reasons, nor is it possible to arrive at a real reason in any other way.

And this is also writing advice: because I have sometimes been approached by people who say “How do I convince people to wear green shoes? I don’t know how to argue it,” and I reply, “Ask yourself honestly whether you should wear green shoes; then make a list of which thoughts actually move you to decide one way or another; then figure out how to explain or argue them, recursing as necessary.”

[mike_kenny2]

that sounds liek a great method. i shall have to try it out when someone asks me that question, or when i ask myself that question regarding something i do automatically.

on the other hand, i could see someone simply meaning ‘why do you use qwerty’ as ‘what’s the history of your using qwerty’. i wonder if this linguistic issue might cause us to conflate facts of the past with motivations. when i give a reason why i use qwerty historically it might just be ‘well, that was what was available’ whereas an answe to a ‘why’ question aimed at getting justifications might be ‘well, i don’t know if it’s justified.’

perhaps this sheds some light onto the problem of confusing is with ought. we might confuse ‘why is it’ for ‘why should we’.

[TobyBartels]

on the other hand, i could see someone simply meaning ‘why do you use qwerty’ as ‘what’s the history of your using qwerty’. i wonder if this linguistic issue might cause us to conflate facts of the past with motivations.

I tend to say ‘Why do you use QWERTY?’ for the question that Eliezer was writing about, and ‘How come you use QWERTY?’ for the question of history. As far as I know, this isn’t justified by any recognised rule of English grammar, but it feels right to me.

You can also use more precise phrasing: ‘Why do you choose to use QWERTY?’, ‘How did it come about that you use QWERTY?’. But who wants to use more words than necessary?

[Lowly_Undergrad2]

Can I request a little more explanation? Why should I adopt this method? [/​sarcasm]

[Ben_Jones]

/​asks self if he should beat his wife

/​realises self is not married

[Furcas]

Good advice. I have noticed that, when faced with this sort of question, I tend to seek to justify myself rather than find the real answers, but I wasn’t sure how to fight this tendency. This looks like it will work, although it will take a lot of effort on my part.

[Z._M._Davis]

I worry that blanking your mind is a tad underspecified. If you delete all your current justifications and metajustifications, you become a rock, not a perfect ghost of decision-theoretic emptiness. What we want is to delete just our current notion of a good means, while preserving our ultimate ends, but the human brain just might not be typed that strongly. When asking what is truly valuable, it could sometimes seem that the answer really is “whatever you want it to be”—except that we don’t want it to be whatever we want it to be; we want an answer (even knowing that that answer has to be a fact about ourselves, rather than about some universal essence of goodness). Ack!

[JamesAndrix]

“Why are you using a QWERTY keyboard?” or “Why haven’t you invested in stocks?” and you don’t remember already considering this exact question and deciding it,

It think you mean: If I don’t remember already considering the question “should I do X”, Checking for a cached answer to the why question can just yield an answer to the wrong question.

You’ve got to rewrite the question first, then check to see if you know an answer to the right question.

[Neuroskeptic]

Might it also help to ask yourself, “Why am I choosing not to use a Dvorak keyboard?”—i.e. to reframe the question in terms of the alternative you’re currently forgoing?

[apophenia]

Perhaps a counterfactual? “Why am I using a Dvorak keyboard?”, even though you’re not?

[JulianMorrison]

Neuroskeptic: it wouldn’t help. The problem is screwy human brain circuitry that turns an attempt to look for reasons for any already-assumed decision into a self-deluding hunt for excuses. You have to un-assume the decision before you stand any chance of clear thought.

Just as a general thing, ANY target of retroactive justification would be no use. You’d be setting your subconscious mind to the task of finding excuses that convince you—hacking the reason firewalls of your own brain.

[bigjeff5]

Poppycock.

As soon as you questioned your motives you were 9/​10ths the way to finding the true reasons you use the QWERTY keyboard and on the path to avoiding self-delusion. For each answer you come up with for or against using a QWERTY keyboard you simply need to question your motives for that answer.

If at that point you were still somehow deluding yourself, there is absolutely no way you could possibly tell, and the question is entirely moot.

(A note in advance, my result from the example below really, really surprised me!)

For example, my own reasons for using a QWERTY keyboard are simple: I learned to touch-type QWERTY, and I type reasonably fast with a QWERTY keyboard. Also, all keyboards I have ever used or come across have been QWERTY.

My reasons for not using a DVORAK keyboard are equally simple, though I’ve never considered them before now. They are true of any alternative to QWERTY, so it doesn’t really matter. The following are purely assumptions.

-First, I don’t know DVORAK, so I would have to learn a new style of touch-typing all over again. I am reasonably certain that this will take a significant amount of effort, because it took years of practice to become proficient with QWERTY in the first place, and I know from other personal experience how difficult it is for me to re-train the level of muscle memory I have with QWERTY touch-typing. -Second, as far as I know DVORAK is not superior to QWERTY in any significant way -Third, I have never seen a DVORAK keyboard, so they are likely to be difficult to find or uncommon.

These are all lines of inquiry I can verify. Some with hard data right off the bat, some will require further research, and based on my initial findings I can determine if such research is warranted.

The second assumption seems the easiest to falsify, and a quick Google search gives some interesting results which certainly warrant further research: -The first result for the fastest verified typist in the world is Barbara Blackbum, and she types an incredible four times faster than I do. She uses the DVORAK keyboard, apparently at first because QWERTY was too difficult to learn. This certainly raises the question that DVORAK may be superior to QWERTY, but Blackbum may be unique, and if she had stuck with QWERTY she may have ultimately been just as fast. Are there QWERTY typists who type at similar speeds? Unfortunately, a quick Google search gives little information on the layout used by most typist record holders. However, further searching has revealed that her record was broken earlier this year by Michael Shestov, and he certainly appears to be “special” in that it doesn’t matter the keyboard layout, he types about as fast with all of them. Barbara Blackbum is also very fast with QWERTY, but she is significantly faster with Dvorak, lending credibility to the idea that I, too, may be significantly faster with Dvorak than QWERTY. There are studies that suggest Dvorak is faster for average typists as well, but at this point the only way to definitively know which is superior for me would be to try Dvorak and see what happens. This also happens to be very practical, which is a bonus.

During in the process above (which shows my second assumption is clearly not based in reality) I also came across information that calls into question my first assumption, which was that it would require years to become as proficient with Dvorak (or any typing method) as I am with QWERTY. The basis for this assumption was the difficulty of learning QWERTY, however it seems Dvorak was designed from the beginning to be a more natural and easier to learn than QWERTY, and there have been a small handful of studies which seem to bear this out. Anecdotal evidence from Dvorak users also seems to bear this out, though obviously none of this can actually determine that I will find Dvorak easier. However, it does show that my assumption was based on a false premise, and gives me avenues for actually determining if Dvorak is better than QWERTY for myself (namely, try it for a month, and if I can actually type with Dvorak after this short time, try it for a few more months to see how fast I progress and determine if it is worth maintaining).

For the last assumption, just a little thinking on the problem shows that I was looking at the problem incorrectly: It does not matter how popular Dvorak keyboards are or aren’t, a simple setting change in my OS can turn any QWERTY keyboard into a Dvorak keyboard. I can also switch keys around or make little labels for them, but this isn’t really necessary to have a functioning Dvorak keyboard.

The nail in the coffin for your argument actually has nothing to do specifically with what I stated above about the arguments for or against the QWERTY keyboard and the Dvorak keyboard. The nail in the coffin is the huge difference between what I expected to demonstrate and the conclusion I actually came to.

The original purpose of the exercise was to show that there was no point in switching away from the QWERTY keyboard, but this was OK because I wasn’t fooling myself about anything, it was just the most reasonable position for someone who was used to a QWERTY keyboard to take.

My actual conclusion was that it was definitely worth my time to try the Dvorak style, and in fact I will be switching my settings over as soon as I post this (I would have switched already, just out of curiosity, but I have spent about an hour on this post already, and I don’t want to spend any longer than I need to on it).

While I did not find any evidence that definitively said “Yes, Dvorak is clearly superior to QWERTY, stop looking and just use it” to my satisfaction (I doubt I ever would have), I did find that my reasons for not using Dvorak had no substance at all, and since I have a very easy way to falsify the idea that Dvorak is better for me to use than QWERTY (it’s just a simple OS setting after all) that I must go ahead and try. Since I kept finding my self say “Huh. That’s not what I expected” quite often in the half hour or so I was researching the problem, it became very clear that my expectations need to be adjusted. In order to bring my expectations in line with reality, I need to try this for myself and either have concrete reasons for using the QWERTY keyboard, or to switch to Dvorak. I have a feeling I might be switching to Dvorak permanently, given the research that has been done on the subject, but there is no way to know for sure until I try.

As you said, I should have had all kinds of subconscious barriers to changing my assumptions, and in fact I did—I was pretty sure my reasoning against switching was sound before I wrote it down, but writing it forced me to focus on it consciously, and they seemed rather weak, so I set out to disprove them.

Subconscious barriers cannot stand up to my conscious mind once it is (correctly) trained on them. Unfounded assumptions cannot be maintained once the truth is known. Period. There is nothing your subconscious can do to stop that, so as long as you always question your assumptions, no matter where they came from, and always verify your assumptions by seeking to prove them wrong. If you seek to prove them correct you will never look in the right place, and you can easily fall into the trap you describe. Do this an you can pretty much always overcome your subconscious barriers to knowing the difference between the truth and a rationalization.

[falenas108]

I’m curious, did you end up staying with Dvorak?

[Z._M._Davis]

Julian: “You have to un-assume the decision before you stand any chance of clear thought.”

Of course the decision-theoretic logic of this is unassailable, but I continue to worry that the real-world application to humans is nontrivial.

Here, I have a parable. Suppose Jones is a member of a cult, and holds it as a moral principle that it is good and right and virtuous to obey the Great Leader. So she tries to obey, but feels terrible about failing to obey perfectly, and she ends up having a nervous breakdown and removing herself from the cult in shame. Then, afterwards, as a defensive emotional reaction, she adopts an individualist philosophy and comes up with all sorts of clever arguments to the effect that the Great Leader isn’t special at all and really no one has any duty to obey or even listen to her!

So then Jones reads “The Bottom Line,” and realizes her adoption of individualism wasn’t rational, and was simply a reaction to having been hurt so badly. If she had only been better at obeying the Great Leader, then, as a matter of (subjunctive) fact, she never would have come up with all those clever arguments, and wouldn’t have found them convincing if someone else had told them to her. At this point, as rationalists, we advise Jones to clear her mind, unassume her decision to leave the cult, and reëvaluate the matter cleanly. But this advice might be underspecified: if she is supposed to reëvaluate her choice using her current morality, the decision is obvious: stay free. If she is supposed to reëvaluate using her original cult-morality, the decision is obvious: crawl back to the Great Leader, begging for forgiveness.

You can’t reset yourself to a state a perfect emptiness; you have to reset yourself to something. Nor can you reset yourself to a state of perfect emptiness in order to decide what to reset yourself to; that’s just an infinite regress.

I guess the best answer I can give to someone in such a dilemma (the answer I give myself) is to say, “Rational agents act to preserve their current goal system; my past self was confused, I cannot be bound by the terms of her confusion; I can only act from who I am, now, what I want, now.”

But then that sounds like saying that the principle of the bottom line doesn’t apply across morality changes. Which is a little suspicious.

[JulianMorrison]

ZMD: I’m reminded of the darcs revision control system and its patch theory. Given patches A and B applied in order (AB), it’s possible to calculate two commuted patches A’ and B’ that when applied in the opposite order (B’A’) produce the same result. If you did A, did B, and want to undo A, you commute the patches, and then take A’ away from A to walk you back to B’.

That’s really only an illustrative analogy, but it’s a good one. You could see the algorithm here as “commute the decision to the front and delete it”. So taking your example, the original was “decide, and leave, and become an individualist”. The commuted version is “leave, and become an individualist, and decide”. Then delete the decision. You’re re-deciding in the context of the rest of the status quo as a given. “Given I have left and I have become an individualist, would I now decide to leave?”.

Thinking of it that way is a bit brain-twisty, but it makes sense.

[DanielLC]

Are you sure there’s supposed to be a B’? It looks to me like you just need to calculate A’ = BAB^-1. That way, when you take A’ away, you get back to B instead of B’.

[gregorylent]

no way .. just stop at the quiet mind … the answer will simply be revealed by what happens … you are doing way too much

enjoy, gregory

[Abigail]

I might want to make even these decisions emotionally rather than rationally, unconsciously rather than by conscious thought. Conscious thought might not be the most efficient use of my brain.

Also, do I want to devote all the energy I would need to answer these questions, now? I use a QWERTY keyboard, it seems good enough at the moment, Why? Well, why not?

However for practice at being rational with that conscious thought bit of the brain, this is a useful exercise.

[Abigail]

I disagree with ZM Davis. The method described in the post is a useful method of deciding Means, and may even help to decide Ends, if we see what we thought to be Ends as really Means to more fundamental Ends.

Read what the New Revised Standard Version of the Bible calls 4 Maccabees (I am not sure if the text has other names). In that, people decide that eating pork is to be avoided at all costs, and therefore suffer extremities of torture and even death rather than eating pork. The text is full of Stoic philosophy about how one can achieve ones ends if one is sufficiently single-minded.

In ZM Davis’ parable, Jones might ask “Do I want to obey the Great Leader?” and then ask, Will that give me the supportive community which I need as a human being? Might other communities benefit me more than this one? And backing further up, Do I need a supportive community?

Are the Great Leader’s commands the highest Good? (Moving into moral argument).

The question “Why do I do what I do?” may apply to Ends as well as means.

[Douglas_Knight3]

When someone asks you “Why are you doing X?”, And you don’t remember an answer previously in mind, Do not ask yourself “Why am I doing X?”.

How dangerous is that second step? There’s definitely potential for confabulation when you try to remember if you’d ever actively decided on X.

[Caledonian2]

Z.M., I agree with your analysis up to the point where you suggest that rational agents act to preserve their current value system.

I suggest that it may be useful for you to consider what the purpose of a value system is. When trying to decide between two value systems, a rational agent must evaluate them in some way. Is there an impersonal and objective set of criteria for evaluation?

[Caledonian2]

Z.M., I agree with your analysis up to the point where you suggest that rational agents act to preserve their current value system.

It may be useful to consider why we have value systems in the first place. When we know why we do a thing, we can evaluate how well we do it, but not until then.

[Roko]

@Z.M. Davis:

You raise a very good point that I have already encountered twice before. I have written a blog post on the issue

[Richard_Hollerith2]

Earlier on this page Eliezer writes,

I have sometimes been approached by people who say “How do I convince people to wear green shoes? I don’t know how to argue it,” and I reply, “Ask yourself honestly whether you should wear green shoes; then make a list of which thoughts actually move you to decide one way or another; then figure out how to explain or argue them . . .

That piece of advice is also in Eliezer’s “Singularity Writing Advice” where I saw it in 2001. I decided to adhere to it and for what it is worth have never regretted the decision. It works as far as I can tell even for my outre moral beliefs.

[Z._M._Davis]

“I have written a blog post on the issue”

I’d love to read it, but the link here is broken, nor do I see any new posts on the Transhuman Goodness homepage. I hope you saved a local copy!

[Roko]

here it is again. I edited the post and blogger un-published it...

My essential point is that the tension between [on the on hand] “rationality” and [on the other hand] meaning, personality and value in our lives seems to have come up independently a fair few times, so it would seem likely that there is a serious issue here.

[notsonewuser]

From the standpoint of TDT, I see using Dvorak as the obvious choice, and teaching your kids Dvorak rather than QWERTY, etc. Anything but QWERTY.