A sutra for the second one: Simple answers are lighthouses. Use them to navigate, but don't sail directly towards them.
The No do-overs section reminded me of a recent conversation. A friend was giving me a lift home from a rationality meetup, we got off of the highway, and I told him to turn right. We should have turned left. Once we realised my mistake, I apologized. His response was something along the lines of "We've just been talking for the last three hours. Why do you believe I'd be averse to spending another five minutes with you?"
The feeling I had wasn't really that there was any badness to spending more time talking, but I knew that he was meeting someone else after dropping me off, and I didn't want to make him late. I dislike being late. I projected that feeling on to him.
No do-overs is also often felt when you forget people's names.
Also, I want to congratulate you for writing the exam. :)
1. I like the general premise and metaphor a lot but am a little fuzzy on what you're referring to exactly. I endorse a statement of the form "take arbitrary constraints and run with them rather than resisting them, it'll at least be a fun and creative exercise." But there's probably something more to Boxing your Brain?
2. This reminds me of trying to learn life and death problems in Go. Trying to figure out general heuristics other than "read very carefully" seems hopeless. Perhaps AlphaGo would disagree.
3. Good point.
The thing I'd like to point to in 1) is "sometimes, you just run with the meta-knowledge that your brain is corrupted <right now>, and you just default to your Shoulds". And if any part of you tries to object, you silence it with a "Can't listen to you right now! You're compromised!"
Something like that has been useful for me in certain...sensitive, shall we say, situations.
<insert obvious caveats and problems with this approach here.>
[alkjash just finished his superb Hammertime sequence. I'm taking him up on his Hammertime Final Exam to come up with:
with just 5 minutes of brainstorming + 5 minutes of writing for each. The results are below.]
Sometimes, your brain will come up with really great reasons to do really stupid things. Well, maybe they’re not totally stupid, but you don’t have the time to do some sort of fancy in-depth analysis to see what’s up. Or, you know that your System 1 is compromised. In those cases, just Box Your Brain.
Okay, so boxing AIs doesn’t seem to work that well. Unless you precommit, of course, such that you can’t yourself boxed. That’s the same idea here. You know that you know that you know…(and all the way up) that you’re running off faulty hardware.
So just tell yourself, “Nope, boxing my brain right now.” And go off and do the Virtuous Thing that you already know is Good and Correct. It’s your brain, and you can use it how you want. Sometimes the right thing to do is to stop listening and run with your meta-knowledge.
You know how the right thing to do is always listen to your gut? Except for when your gut’s wrong?
You know how the right thing to do is always to write stuff down? Except for when writing stuff down isn’t helpful?
Yeah, so there’s this thing here that’s sort of like the metacontrarian cycle happening in rationality. At least a good third of rationality is trying to draw generalizations around things. We love to work off abstraction. This is going to lead to quite valuable heuristics; and of course there will be lots of edge cases.
Rules aren’t absolute. Sometimes they get broken. So what does that leave us with? Well, one thing is we can get specific down to the situation to make our judgments, much like the way you can still tell that certain moral decisions are repugnant, even if The System declares that said decision is the “right one”.
Here’s a thing I think at least some people are averse to: doing things again. Imagine you’re chatting with a stranger. The conversation finishes, and the two of you part ways. Then, BAM! You remember something cool that’d be great to bring up.
I postulate that the aversion you feel towards running back to the other person and bringing up said piece of coolness grows faster than your ability to run. That is, there will be a point where it will be in theory possible to catch up to the other party—and it’ll also feel Bad.
This feels like the same reason that someone might feel Bad following up again to ask someone for directions, after already having asked them for directions a few minutes prior. There’s something about “going back and trying to redo the same interaction” that I think feels bad.
And I also think it shouldn’t.
Something something about social roles and how interactions are “supposed” to generally follow some assumptions about how often they happen, and two interactions in rapid succession feels weird.
Except that I think most people are fine with multiple interactions despite the weirdness. Or, at least, there are good ways to do this well.
So leave that Badness behind.
Do it again!
(And again! And again!)