I don’t understand the logic of this. Does seem like game-theoretically the net-payout is really what matters. What would be the argument for something else mattering?
Suppose that Alice blackmails me and I pay her $1,000,000. Alice has spent $1,500,000 on lawyers in the process of extracting this payout from me. The result of this interaction is that I have lost $1,000,000, while Alice has lost $500,000. (Alice’s lawyers have made a lot of money, of course.)
Bob hears about this. He correctly realizes that I am blackmailable. He talks to his lawyer, and they sign a contract whereby the lawyer gets half of any payout that they’re able to extract from me. Bob blackmails me and I pay him $1,000,000. Bob keeps $500,000, and his lawyer gets the other $500,000. Now I have again lost $1,000,000, while Bob has gained $500,000.
(How might this happen? Well, Bob’s lawyer is better than Alice’s lawyers were. Bob’s also more savvy, and knows how to find a good lawyer, how to negotiate a good contract, etc.)
That is: once the fact that you’re blackmailable is known, the net payout (taking into account expenditures needed to extract it from you) is not relevant, because those expenditures cannot be expected to hold constant—because they can be optimized. And the fact that (as is now a known fact) money can be extracted from you by blackmail, is the incentive to optimize them.
Note that a lawyer who participated in that would be committing a crime. In the case of LH, there was (by my unreliable secondhand understanding) an employment-contract dispute and a blackmail scheme happening concurrently. The lawyers would have been involved only in the employment-contract dispute, not in the blackmail, and any settlement reached would have nominally been only for dropping the employment-contract-related claims. An ordinary employment dispute is a common-enough thing that each side’s lawyers would have experience estimating the other side’s costs at each stage of litigation, and using those estimates as part of a settlement negotiation.
(Filing lawsuits without merit is sometimes analogized to blackmail, but US law defines blackmail much more narrowly, in such a way that asking for payment to not allege statutory rape on a website is blackmail, but asking for payment to not allege unfair dismissal in a civil court is not.)
Then it sounds like the blackmailer in question spent $0 on perpetrating the blackmail, which is even more of an incentive for others to blackmail MIRI in the future.
Then it sounds like the blackmailer in question spent $0 on perpetrating the blackmail
No, that’s not what I said (and is false). To estimate the cost you have to compare the outcome of the legal case to the counterfactual baseline in which there was no blackmail happening on the side (that baseline is not zero), and you have to include other costs besides lawyers.
To estimate the cost you have to compare the outcome of the legal case to the counterfactual baseline in which there was no blackmail happening on the side (that baseline is not zero)
Seems wrong to me. Opportunity cost is not the same as expenditures.
and you have to include other costs besides lawyers.
Sure, I agree that this is true. But as long as you run a policy that is sensitive to your counterparty optimizing expenditures, I think this no longer holds?
Like, I think in-general a policy I have for stuff like this is something like “ensure the costs to my counterparty were higher than their gains”, and then take actions appropriate to the circumstances. This seems like it wouldn’t allow for the kind of thing you describe above (and also seems like the most natural strategy for me in blackmail cases like this).
But as long as you run a policy that is sensitive to your counterparty optimizing expenditures, I think this no longer holds?
What would this look like…? It doesn’t seem to me to be the sort of thing which it’s at all feasible to do in practice. Indeed it’s hard to see what this would even mean; if the end result is that you pay out sometimes and refuse other times, all that happens is that external observers conclude “he pays out sometimes”, and keep blackmailing you.
in-general a policy I have for stuff like this is something like “ensure the costs to my counterparty were higher than their gains”, and then take actions appropriate to the circumstances
Actions like what?
Like, let’s say that you’re MIRI and you’re being blackmailed. You don’t know how much your blackmailer is paying his lawyers (why would you, after all?). What do you do?
And for all you know, the contract your blackmailer’s got with his lawyers might be as I described—lawyers get some percent of payout, and nothing if there’s no payout. What costs do you impose on the blackmailer?
In short, I think the policy you describe is usually impossible to implement in practice.
But note that this is all tangential. It’s only relevant to the original question (about MIRI) if you claim that MIRI were attempting to implement a policy such as you describe. Do you claim this? If so, have you any evidence?
I mean, the policy here really doesn’t seem very hard. If you do know how much your opposing party is paying their lawyers, you optimize that hard. If you don’t know, you make some conservative estimate. I’ve run policies like this in lots of different circumstances, and it’s also pretty close to common sense as a response to blackmail and threats.
Do you claim this? If so, have you any evidence?
I’ve asked some MIRI people this exact question and they gave me this answer, with pretty strong confidence and relatively large error margins.
I have to admit that I still haven’t the faintest clue what concrete behavior you’re actually suggesting. I repeat my questions: “What would this look like…?” and “Actions like what?” (Indeed, since—as I understand it—you say you’ve done this sort of thing, can you give concrete examples from those experiences?)
I’ve asked some MIRI people this exact question and they gave me this answer, with pretty strong confidence and relatively large error margins.
Alright, and what has this looked like in practice for MIRI…?
It means you sit down, you make some fermi estimates of how much benefit the counterparty could be deriving from this threat/blackmail, then you figure out what you would need to do to roughly net out to zero, then you do those things. If someone asks you what your policy is, you give this summary.
In every specific instance this looks different. Sometimes this means you reach out to people they know and let them know about the blackmailing in a way that would damage their reputation. Sometimes it means you threaten to escalate to a legal battle where you are willing to burn resources to make the counterparty come out in the red.
In every specific instance this looks different. Sometimes this means you reach out to people they know and let them know about the blackmailing in a way that would damage their reputation. Sometimes it means you threaten to escalate to a legal battle where you are willing to burn resources to make the counterparty come out in the red.
Why would you condition any of this on how much they’re spending?
And how exactly would you calibrate it to impose a specific amount of cost on the blackmailer? (How do you even map some of these things to monetary cost…?)
Suppose that Alice blackmails me and I pay her $1,000,000. Alice has spent $1,500,000 on lawyers in the process of extracting this payout from me. The result of this interaction is that I have lost $1,000,000, while Alice has lost $500,000. (Alice’s lawyers have made a lot of money, of course.)
Bob hears about this. He correctly realizes that I am blackmailable. He talks to his lawyer, and they sign a contract whereby the lawyer gets half of any payout that they’re able to extract from me. Bob blackmails me and I pay him $1,000,000. Bob keeps $500,000, and his lawyer gets the other $500,000. Now I have again lost $1,000,000, while Bob has gained $500,000.
(How might this happen? Well, Bob’s lawyer is better than Alice’s lawyers were. Bob’s also more savvy, and knows how to find a good lawyer, how to negotiate a good contract, etc.)
That is: once the fact that you’re blackmailable is known, the net payout (taking into account expenditures needed to extract it from you) is not relevant, because those expenditures cannot be expected to hold constant—because they can be optimized. And the fact that (as is now a known fact) money can be extracted from you by blackmail, is the incentive to optimize them.
Note that a lawyer who participated in that would be committing a crime. In the case of LH, there was (by my unreliable secondhand understanding) an employment-contract dispute and a blackmail scheme happening concurrently. The lawyers would have been involved only in the employment-contract dispute, not in the blackmail, and any settlement reached would have nominally been only for dropping the employment-contract-related claims. An ordinary employment dispute is a common-enough thing that each side’s lawyers would have experience estimating the other side’s costs at each stage of litigation, and using those estimates as part of a settlement negotiation.
(Filing lawsuits without merit is sometimes analogized to blackmail, but US law defines blackmail much more narrowly, in such a way that asking for payment to not allege statutory rape on a website is blackmail, but asking for payment to not allege unfair dismissal in a civil court is not.)
Then it sounds like the blackmailer in question spent $0 on perpetrating the blackmail, which is even more of an incentive for others to blackmail MIRI in the future.
No, that’s not what I said (and is false). To estimate the cost you have to compare the outcome of the legal case to the counterfactual baseline in which there was no blackmail happening on the side (that baseline is not zero), and you have to include other costs besides lawyers.
Seems wrong to me. Opportunity cost is not the same as expenditures.
Alright, and what costs were there?
Sorry, this isn’t a topic where I want to discuss with someone who’s being thick in the way that you’re being thick right now. Tapping out.
Sure, I agree that this is true. But as long as you run a policy that is sensitive to your counterparty optimizing expenditures, I think this no longer holds?
Like, I think in-general a policy I have for stuff like this is something like “ensure the costs to my counterparty were higher than their gains”, and then take actions appropriate to the circumstances. This seems like it wouldn’t allow for the kind of thing you describe above (and also seems like the most natural strategy for me in blackmail cases like this).
What would this look like…? It doesn’t seem to me to be the sort of thing which it’s at all feasible to do in practice. Indeed it’s hard to see what this would even mean; if the end result is that you pay out sometimes and refuse other times, all that happens is that external observers conclude “he pays out sometimes”, and keep blackmailing you.
Actions like what?
Like, let’s say that you’re MIRI and you’re being blackmailed. You don’t know how much your blackmailer is paying his lawyers (why would you, after all?). What do you do?
And for all you know, the contract your blackmailer’s got with his lawyers might be as I described—lawyers get some percent of payout, and nothing if there’s no payout. What costs do you impose on the blackmailer?
In short, I think the policy you describe is usually impossible to implement in practice.
But note that this is all tangential. It’s only relevant to the original question (about MIRI) if you claim that MIRI were attempting to implement a policy such as you describe. Do you claim this? If so, have you any evidence?
I mean, the policy here really doesn’t seem very hard. If you do know how much your opposing party is paying their lawyers, you optimize that hard. If you don’t know, you make some conservative estimate. I’ve run policies like this in lots of different circumstances, and it’s also pretty close to common sense as a response to blackmail and threats.
I’ve asked some MIRI people this exact question and they gave me this answer, with pretty strong confidence and relatively large error margins.
I have to admit that I still haven’t the faintest clue what concrete behavior you’re actually suggesting. I repeat my questions: “What would this look like…?” and “Actions like what?” (Indeed, since—as I understand it—you say you’ve done this sort of thing, can you give concrete examples from those experiences?)
Alright, and what has this looked like in practice for MIRI…?
It means you sit down, you make some fermi estimates of how much benefit the counterparty could be deriving from this threat/blackmail, then you figure out what you would need to do to roughly net out to zero, then you do those things. If someone asks you what your policy is, you give this summary.
In every specific instance this looks different. Sometimes this means you reach out to people they know and let them know about the blackmailing in a way that would damage their reputation. Sometimes it means you threaten to escalate to a legal battle where you are willing to burn resources to make the counterparty come out in the red.
Why would you condition any of this on how much they’re spending?
And how exactly would you calibrate it to impose a specific amount of cost on the blackmailer? (How do you even map some of these things to monetary cost…?)