MrGus99's Shortform
post by MrGus99 · 2021-05-03T19:17:43.370Z · LW · GW · 9 commentsContents
10 comments
9 comments
Comments sorted by top scores.
comment by MrGus99 · 2021-05-07T15:24:15.544Z · LW(p) · GW(p)
Removed.
Replies from: Dagon, Robbo↑ comment by Dagon · 2021-05-07T19:33:04.925Z · LW(p) · GW(p)
Well, "0" is not a probability, but aside from that, this is far too general. There are certainly things you should set an expectation of arbitrarily low probability. 1/36 is way too high for some things that I don't want/intend to do, and way too low for some things that I kind of want to do, but am conflicted in ways that I can't just commit. And for many many things I don't even intend to expend the effort of rolling dice or otherwise externally making a decision.
Also, for things that part of me wants but I am so averse to that I'll only agree to a 1/36 chance I'll have to do it, I'm likely to cheat anyway, so I should just pre-commit to "no" and not have to worry about it. I kind of suck that way.
[ edit: I don't mean to talk anyone out of trying this - if it works for you, that's awesome! ]
Replies from: MrGus99↑ comment by MrGus99 · 2021-05-07T20:53:14.291Z · LW(p) · GW(p)
Removed.
Replies from: Dagon↑ comment by Dagon · 2021-05-07T23:16:32.804Z · LW(p) · GW(p)
if you are able to follow through on the respective actions you precommitted to
If I had that much willpower, I'd just do (or not do, if that's my true net preference) the thing.
There are no such situations I can think of where an explicit outside randomness helps me. The classic uses of a mixed strategy (where the adversary is strategically optimizing against your intent) don't apply here.
↑ comment by MrGus99 · 2021-05-08T05:23:51.908Z · LW(p) · GW(p)
Removed.
Replies from: Dagon↑ comment by Dagon · 2021-05-08T15:17:35.007Z · LW(p) · GW(p)
Why I think what is? That this intervention doesn't help? Because if I don't want to do something, and I think I can get away with not doing it, I think that will STILL BE THE CASE after I roll a die. For me, at least (I acknowledge that it may work for others, which is great), I care far less about a statement of intent to obey a random event than I care about the actual behavior. Adding the die roll does not add any information or decision weight.
Replies from: MrGus99↑ comment by MrGus99 · 2021-05-08T17:51:52.732Z · LW(p) · GW(p)
Removed.
Replies from: Dagon↑ comment by Dagon · 2021-05-09T02:01:55.313Z · LW(p) · GW(p)
Definitely I'm confused - I don't see how the die roll helps, over just deciding to do or not do the thing. I think you're describing a decision about whether to commit to something, prior to the actual behavior of doing it (which is a decision as well, though I'm not sure whether you agree on that point). Your description is of a decision to assign an external probability source to the commitment portion of the sequence. I don't understand why you wouldn't prefer to just decide.
I think remain most confused by
But it is NOT okay, to decrease that chance to 0
I don't understand why it's OK to commit to a small chance of doing something I don't want to, but why it's not OK to just not do it (colloquial 0%. Bayesian arbitrary small chance, as circumstances can change).
I think an existence proof would help - what decisions or actions has this worked for for you? How did you pick the odds to use? I can't think of any decisions where I expect it to help me in any way (except certain adversarial games where mixed strategies are optimal, but those are incredibly rare in the real world).