Have some third party (or several) that LW would trust hold on to it in secrect.
Nitpick: cryptography solves this much more neatly.
Of course, people could accuse you of having an efficient way of factorising numbers, but if you do karma is going to be the least of anyone’s concerns.
Nitpick: cryptography solves this much more neatly.
But somewhat less transparently. The cryptographic solution still requires that an encrypted message is made public prior to the actions being taken and declaring an encrypted prediction has side effects. The neat solution is to still use trusted parties but give the trusted parties only the encrypted strategy (or a hash thereof).
The cryptographic solution still requires that an encrypted message is made public prior to the actions being taken and declaring an encrypted prediction has side effects.
What kind of side effects ? I have no formal training in cryptography, so please forgive me if this is a naive question.
What kind of side effects ? I have no formal training in cryptography, so please forgive me if this is a naive question.
I mean you still have to give the encrypted data to someone. They can’t tell what it is but they can see you are up to something. So you still have to use some additional sort of trust mechanism if you don’t want the act of giving encrypted fore-notice to influence behavior.
Ah ok, that makes sense. In this case, you can employ steganography. For example, you could publish an unrelated article using a pretty image as a header. When the time comes, you reveal the algorithm and password required in order to extract your secret message from the image.
Ah ok, that makes sense. In this case, you can employ steganography. For example, you could publish an unrelated article using a pretty image as a header. When the time comes, you reveal the algorithm and password required in order to extract your secret message from the image.
Better yet… embed five different predictions in that header. When the time comes, reveal just the one that turned out most correct!
But of four people on LW who would be considered trusted parties, what’s the probability that all four would be quiet after the fifth is called upon to post the prediction or prediction hash?
You’re right, of course. I didn’t think that through. There haven’t been any good “gain the habit of really thinking things through” exercises for a Skill-of-the-Week post, have there?
“Recognizing when you’ve actually thought thoroughly” is the specific failure mode I’m thinking of; but that’s probably highly correlated with recognizing when to start thinking thoroughly.
I feel like such a skill may be difficult to consciously train without a tutor:
Rice’s theorem will tell you that you cannot, without already knowing unknown unknowns, determine which knowledge is safe to ignore.
-- @afoolswisdom
Besides, in the prediction-hash case, they may well not post right away.
Yes, the first thing I thought of was Quirrel’s hashed prediction; but it doesn’t seem that everyone’s forgotten yet, as of last month.
Nitpick: cryptography solves this much more neatly.
Of course, people could accuse you of having an efficient way of factorising numbers, but if you do karma is going to be the least of anyone’s concerns.
Factorization doesn’t enter into it—to precommit to a message that you will later reveal publically, publish a hash of the (salted) message.
But somewhat less transparently. The cryptographic solution still requires that an encrypted message is made public prior to the actions being taken and declaring an encrypted prediction has side effects. The neat solution is to still use trusted parties but give the trusted parties only the encrypted strategy (or a hash thereof).
What kind of side effects ? I have no formal training in cryptography, so please forgive me if this is a naive question.
I mean you still have to give the encrypted data to someone. They can’t tell what it is but they can see you are up to something. So you still have to use some additional sort of trust mechanism if you don’t want the act of giving encrypted fore-notice to influence behavior.
Ah ok, that makes sense. In this case, you can employ steganography. For example, you could publish an unrelated article using a pretty image as a header. When the time comes, you reveal the algorithm and password required in order to extract your secret message from the image.
Better yet… embed five different predictions in that header. When the time comes, reveal just the one that turned out most correct!
Hmm yes, there might be a hidden weakness in my master plan as far as accountability is concerned :-)
None that were not extant in the original scheme, assuming there are at least five people on LW who’d be considered trusted parties.
But of four people on LW who would be considered trusted parties, what’s the probability that all four would be quiet after the fifth is called upon to post the prediction or prediction hash?
You’re right, of course. I didn’t think that through. There haven’t been any good “gain the habit of really thinking things through” exercises for a Skill-of-the-Week post, have there?
Bear in mind that it’s often not worth the effort. I think the skill to train would be recognizing when it might be.
Besides, in the prediction-hash case, they may well not post right away.
“Recognizing when you’ve actually thought thoroughly” is the specific failure mode I’m thinking of; but that’s probably highly correlated with recognizing when to start thinking thoroughly.
I feel like such a skill may be difficult to consciously train without a tutor:
-- @afoolswisdom
Yes, the first thing I thought of was Quirrel’s hashed prediction; but it doesn’t seem that everyone’s forgotten yet, as of last month.