I’m pretty sure that’s an incorrect conclusion. The related, correct conclusion is that if I revealed the statement, and it was a correct prediction, you shouldn’t give me credit, because I hadn’t promised to reveal it. Or if I had said that it was a prediction and then refused to reveal it, that would be epistemically bad.
But it wasn’t a prediction, and I never suggested it was. I was just giving myself the opportunity to prove that I wrote something in advance. This turned out not to be useful. I’m pretty sure it’s epistemically fine to do this, or to share hashes of personal policies with no obligation to share them unless you decide they’re relevant.
But I’ll try to make non-predictions more clearly not predictions in the future, thanks.
Suppose Alice thinks there’s a smallish chance that she will say X and then Bob will say no way you thought of X beforehand and then she will want to prove that she thought of X beforehand. What should Alice do?
She could do something like: write X on a google doc in advance and then if she says X later, she also shares the google doc with its version history (so Bob can see when she wrote it). But that’s consistent with her writing many google docs and selecting one to share after the fact, which is problematic in some cases.
Or she could do something like what I did, which is like proving that you only had one google doc. That seems like the optimal low-effort solution. I agree we should clearly distinguish predictions from stuff like this, though.
I intuit that it would be crazy if there was no way for Alice to show that she thought of X in advance without being accused of evidence-filtering when she doesn’t share.
I guess I’m still assuming the only reason to timestamp a statement is for the prediction-y qualities. “I was just giving myself the opportunity to prove that I wrote something in advance.” Why would this matter at all, if not for the prediction-y qualities of what you wrote? Could be a failure of imagination on my part. Can you give me a concrete example of something someone might want to write down, not share, and later prove they thought of in advance, not for the prediction-y qualities? I guess there’s “I was first so I get the patent”, and in a world where the idea doesn’t work but does contain a trade secret, you wouldn’t want to reveal it, to preserve the secret? Too convoluted—sorry, currently, due to my failure of imagination, despite your statements to the contrary, I think it’s very likely that the message was written for its prediction-y qualities.
(Also—is there a reason I should believe that if all did go according to plan, when you revealed your message, you would also have said “if all had not gone according to plan, I would not have revealed this message”? ’cause I currently think there’s a very low chance you would have said that. There would be at least a 1% chance it would have been advantageous to avoid saying that, for sure.)
Can you give me a concrete example of something someone might want to write down, not share, and later prove they thought of in advance, not for the prediction-y qualities
Sure, here are three, wanting to show you’d thought about something in advance for various reasons:
Alice notices an unfixable vulnerability that would cause a website to go down. Drawing attention to it would have no benefit, but she hides an explanation so that if someone takes down the website, (1) she gets credit for noticing it and (2) she better helps others understand what happened (the fact that she wrote in advance about the website going down, and then the website went down, means people should pay more attention to her explanation).
Bob anticipates a situation (that others don’t anticipate that he anticipates), writes a personal policy on what to do in that situation, and publishes the hash (along with many other hashes). If he acts on that policy, he can prove that he wrote the policy in advance. He can’t prove that he didn’t write alternatives, but he can prove he considered the situation in advance, which is sometimes sufficient.
Carol is contemplating taking a unilateral action. She shares a hash. Now if she takes the unilateral action later, she can prove that she carefully thought about it well in advance. And regardless, if the possibility of the action becomes obvious, she can prove she thought of it before it became obvious.
Also—is there a reason I should believe that if all did go according to plan, when you revealed your message, you would also have said “if all had not gone according to plan, I would not have revealed this message”?
Well, I know that the message says things like “if you’re reading this then …” and asserts things about the past that did not come to pass. But I can’t easily prove that to you, no.
Note also that I mentioned that this whole hash-sharing thing wasn’t actually necessary for what I was doing.
Since the other party to the dialog wants to be secretive, i figure giving imagination help decouples the principle matters and that fact in particular matters.
I can think that you could confess to a crime or claim an attack and withhold it for self-incrimination reasons. Then if confessing becomes advantageous validity of it is pretty solid. The confessed crime could be in the past at the moment of writing so it is not predicting anything.
I now believe I should treat any supposed information coming from you as much more likely to be filtered evidence than I would usually suspect. :(
I’m pretty sure that’s an incorrect conclusion. The related, correct conclusion is that if I revealed the statement, and it was a correct prediction, you shouldn’t give me credit, because I hadn’t promised to reveal it. Or if I had said that it was a prediction and then refused to reveal it, that would be epistemically bad.
But it wasn’t a prediction, and I never suggested it was. I was just giving myself the opportunity to prove that I wrote something in advance. This turned out not to be useful. I’m pretty sure it’s epistemically fine to do this, or to share hashes of personal policies with no obligation to share them unless you decide they’re relevant.
But I’ll try to make non-predictions more clearly not predictions in the future, thanks.
Another angle:
Suppose Alice thinks there’s a smallish chance that she will say X and then Bob will say no way you thought of X beforehand and then she will want to prove that she thought of X beforehand. What should Alice do?
She could do something like: write X on a google doc in advance and then if she says X later, she also shares the google doc with its version history (so Bob can see when she wrote it). But that’s consistent with her writing many google docs and selecting one to share after the fact, which is problematic in some cases.
Or she could do something like what I did, which is like proving that you only had one google doc. That seems like the optimal low-effort solution. I agree we should clearly distinguish predictions from stuff like this, though.
I intuit that it would be crazy if there was no way for Alice to show that she thought of X in advance without being accused of evidence-filtering when she doesn’t share.
I guess I’m still assuming the only reason to timestamp a statement is for the prediction-y qualities. “I was just giving myself the opportunity to prove that I wrote something in advance.” Why would this matter at all, if not for the prediction-y qualities of what you wrote? Could be a failure of imagination on my part. Can you give me a concrete example of something someone might want to write down, not share, and later prove they thought of in advance, not for the prediction-y qualities? I guess there’s “I was first so I get the patent”, and in a world where the idea doesn’t work but does contain a trade secret, you wouldn’t want to reveal it, to preserve the secret? Too convoluted—sorry, currently, due to my failure of imagination, despite your statements to the contrary, I think it’s very likely that the message was written for its prediction-y qualities.
(Also—is there a reason I should believe that if all did go according to plan, when you revealed your message, you would also have said “if all had not gone according to plan, I would not have revealed this message”? ’cause I currently think there’s a very low chance you would have said that. There would be at least a 1% chance it would have been advantageous to avoid saying that, for sure.)
Sure, here are three, wanting to show you’d thought about something in advance for various reasons:
Alice notices an unfixable vulnerability that would cause a website to go down. Drawing attention to it would have no benefit, but she hides an explanation so that if someone takes down the website, (1) she gets credit for noticing it and (2) she better helps others understand what happened (the fact that she wrote in advance about the website going down, and then the website went down, means people should pay more attention to her explanation).
Bob anticipates a situation (that others don’t anticipate that he anticipates), writes a personal policy on what to do in that situation, and publishes the hash (along with many other hashes). If he acts on that policy, he can prove that he wrote the policy in advance. He can’t prove that he didn’t write alternatives, but he can prove he considered the situation in advance, which is sometimes sufficient.
Carol is contemplating taking a unilateral action. She shares a hash. Now if she takes the unilateral action later, she can prove that she carefully thought about it well in advance. And regardless, if the possibility of the action becomes obvious, she can prove she thought of it before it became obvious.
Well, I know that the message says things like “if you’re reading this then …” and asserts things about the past that did not come to pass. But I can’t easily prove that to you, no.
Note also that I mentioned that this whole hash-sharing thing wasn’t actually necessary for what I was doing.
Since the other party to the dialog wants to be secretive, i figure giving imagination help decouples the principle matters and that fact in particular matters.
I can think that you could confess to a crime or claim an attack and withhold it for self-incrimination reasons. Then if confessing becomes advantageous validity of it is pretty solid. The confessed crime could be in the past at the moment of writing so it is not predicting anything.