Comment Permalink

DataPacRat13y70

Naturally it's not /necessarily/ true - rules-of-thumb aren't always true, they just tend to be handy in a large portion of circumstances. Their main utility comes from having them cached and easily recalled, so that you don't have to waste a lot of time re-deriving them from scratch; this utility comes at the cost of not being applicable in every conceivable situation, but doesn't mean they're totally useless.

5 Is this rule of thumb useful for gauging low probabilities?

by DataPacRat

1st Jun 2012

1 min read

5

Does something like this seem to you to be a reasonable rule of thumb, for helping handle scope insensitivity to low probabilities?

There's a roughly 30 to 35 out of a million chance that you will die on any given day; and so if I'm dealing with a probability of one in a million, then I 'should' spend 30 times as much time preparing for my imminent death within the next 24 hours as I do playing with the one-in-a-million shot. If it's not worth spending 30 seconds preparing for dying within the next day, then I should spend less than one second dealing with that one-in-a-million shot.

Relatedly, can you think of a way to improve it, such as to make it more memorable? Are there any pre-existing references - not just to micromorts, but to comparing them to other probabilities - which I've missed?

Personal Blog

5

Is this rule of thumb useful for gauging low probabilities?

New Comment

15 comments, sorted by

top scoring

Click to highlight new comments since: Today at 11:07 PM

[-]Incorrect13y90

That's not necessarily true.

Perhaps you are preparing for something much worse than your own death but equally improbable. Alternatively, perhaps you already have your affairs in order and there isn't much valuable stuff you can do to prepare for your own death.

[-]DataPacRat13y70

[-]Vladimir_Nesov13y60

"Is this a useful rule of thumb?" is a bad title for a post.

[-]DataPacRat13y00

Freely acknowledged and accepted. It looks like the LW interface will still let me edit that; do you have a suggestion for a better one?

[-]A1987dM13y40

There's a roughly 30 to 35 out of a million chance that you will die on any given day

This is for a randomly-choosen person; if you know that you don't have any serious illness, that you're going to drive less than the average person drives the average day, and stuff like that, the probability that you will die within 24 hours might well be an order of magnitude lower.

[-]DataPacRat13y00

Do you think it would improve the rule-of-thumb if I swapped out the all-too-specific 'you' and replaced it with 'an average person'?

[-]DanArmak13y20

Edit: badly phrased, see here.

[-]DataPacRat13y00

If you believed you had a very high probability of death soon, then that knocks one of the basic assumptions of this potential-rule-of-thumb out from under it, rendering it much less relevant for that particular situation...

... but as one potential answer for your question - how about taking the time to wrap up your conversations with loved ones in a way that you feel would be suitable as your potential last words with them?

[-]DanArmak13y10

If you believed you had a very high probability of death soon, then that knocks one of the basic assumptions of this potential-rule-of-thumb out from under it

You're right, I should rephrase. What are things you actually do or might do, at your current estimated probability of death within 24 hours, to prepare for death, that fit in 30 seconds daily?

For me the answer is: none. If that is your answer is well, then your rule of thumb is telling you to ignore entirely sufficiently improbable things, on the order of 30/1000000 est. probability or less. Was that your entire intention in proposing this rule?

[-]DataPacRat13y00

It's not my /entire/ intention, as it's possible to group enough individual unlikely items together to collect enough probability-mass to pass that threshold. But it's definitely a significant part of it. The classic example seems to be lottery tickets; if the odds of winning a significant prize are roughly 0.5 out of 1 million, then I can gauge that however much time I've spent dealing with my death for the next day, it makes sense to spend 1/60th of that amount of time dealing with that ticket - which could easily mean that I wouldn't even have time to pull out my wallet before my time becomes better spent dealing with near-immediate death instead.

[-]Eneasz13y10

Now that you've mentioned it, I plan to use it. Seems useful.

[-][anonymous]13y00

Not quite.

You should worry about things to the extent you can change your expected utility.

There's maybe a million to one chance of drawing some particular hand at poker night, and there's also a million to one chance that there will be some disaster (earthquake, zombies, flood). One of those doesn't matter very much and you can't do much about it anyway, the other you can do very much to prepare for and actually make a large expected difference.

Your rule will work if you are well-calibrated to one possible cause of death and you are wondering how much time to spend on another, given that you know the probabilities. If the events (and therefore utilities) are not as comparable, it's best to just use decision theory in some form.

[-]DataPacRat13y00

It may be 'best' to use decision theory - but I've found that it can take more time trying to figure out what a decision theory says about an everyday choice than that choice makes a difference of. So I'm hoping that some variation of this rule-of-thumb allows for a reasonable compromise - while it doesn't always apply, the cases where it does allow you to reap most of the benefits that applying a full-fledged decision theory would, while requiring significantly less mental processing time.

Or maybe it's not useful that way at all - in which case, I'd like to find that out here if I can, before I start relying on it too heavily.

[-]Kindly13y00

Are you sure you trust yourself to know whether it's worth spending 30 seconds to prepare for dying within the next day?

[-]DataPacRat13y00

Part of the point is that, on average, it isn't. There are 86,400 seconds in a day; 30/1e6 of that time period is actually closer to 3 seconds than 30.

Perhaps it might be better if I tried applying this rule over a different scale than a day - maybe a week (~ 210/1e6 chance of dying) or an hour (~ 1.3/1e6 chance of dying).

Moderation Log

Curated and popular this week