Suppose an author I like says she’ll write a new work if she gets enough donations. Under CDT, it’s clear to me that it can’t make sense for me to donate—my donation can’t increase the probability of me reading the book enough to pay for the cost, and there are much more efficient ways for me to give altruistically. What do other decision theories have to say about this?
Suppose an author I like says she’ll write a new work if she gets enough donations. Under CDT, it’s clear to me that it can’t make sense for me to donate—my donation can’t increase the probability of me reading the book enough to pay for the cost, and there are much more efficient ways for me to give altruistically. What do other decision theories have to say about this?
Short answer: CDT doesn’t donate. EDT, TDT and UDT all donate (assuming enough others are mutually known to be like you).
TDT was literally made for this kind of situation (because it’s just a Newcomblike problem). UDT differs from TDT only in areas a bit more obscure than this. EDT is also designed to handle this perfectly too (ie. to get you the book for minimal price). If you donate evidence does suggest that enough people will donate to get you the book but if you don’t donate evidence suggests that you will not.
TDT has to say that if the scenario where everyone donates you win, and you know that everyone else is using TDT or that the distribution of decision algorithms is likely to give sufficient “donate” outputs to make it better expected utility, then you should donate. Of course, if you have reliable data on others’ decision algorithms, I’m pretty sure CDT and EDT and any other decision theory I’ve read about will boil down to an expected utility calculation or something pretty close.
Basically, as Vaniver says, all good DTs pretty much agree on this. TDT, CDT and EDT all agree that if you have common knowledge of a sufficient number of other people using the same decision theory (or, with more complicated calculations, various possible theories including those three) are interested in the book, you should all donate. This common knowledge, however, is usually the extremely costly, high-information-value part—the part about figuring out whether to donate or not seems trivial by comparison.
Basically, as Vaniver says, all good DTs pretty much agree on this. TDT, CDT and EDT all agree that if you have common knowledge of a sufficient number of other people using the same decision theory (or, with more complicated calculations, various possible theories including those three) are interested in the book, you should all donate. This common knowledge, however, is usually the extremely costly, high-information-value part—the part about figuring out whether to donate or not seems trivial by comparison.
I don’t think this is correct. The CDT agents would all agree that they all should donate and would support the implementation of a simple mutual commitment protocol. If they couldn’t arrange a way to compel each other to not defect on the commons problem they would be sad but defect themselves. Fortunately there are already existent online donation systems are sufficient. You just need one of the ones that returns pledged funds if the target goal isn’t met and a carefully calculated target goal.
At the extremes of perfect CDT agents you’d have to fiddle with the details a little more and, for example, make it forbidden for one agent to donate twice in order to allow that any will even donate once. But we can assume either all those details are handled or the CDT agents aren’t quite that ridiculous and consider the precommitment mechanism adequate. Another thing they would do is arrange a taxation system enforced by people with guns with the relevant commons problems to be solved specified by (necessarily compulsory) voting.
Of course, the other thing groups of CDT agents would do is arrange a free market capitalism system wherein products are payed for and people who don’t pay don’t get the stuff. A more efficient system would also allow the author easy access to a loan based on the awareness of the loan giver of the desire for the books. Then she would actually get most of the money from the sales of said books.
A more efficient system would also allow the author easy access to a loan based on the awareness of the loan giver of the desire for the books. Then she would actually get most of the money from the sales of said books.
Right- where again the primary block is the mutual information required.
As far as I can tell, any decision theory that disagrees with CDT in this case is mistaken. The author (or you) need to sweeten the deal; either the benefits need to be better, or the cost needs to be lower. Typical ways to improve the benefit are to attach status or other goods to the donation- whenever I talk about the Kickstarter projects I back, I make sure to mention that, you know, I backed them.
Yeah, but the conversation is about collective patronage in general, not about specific projects, and it seemed like it would detract from my point to also brag with my comment.
Decision theory and selfish donating
Suppose an author I like says she’ll write a new work if she gets enough donations. Under CDT, it’s clear to me that it can’t make sense for me to donate—my donation can’t increase the probability of me reading the book enough to pay for the cost, and there are much more efficient ways for me to give altruistically. What do other decision theories have to say about this?
Short answer: CDT doesn’t donate. EDT, TDT and UDT all donate (assuming enough others are mutually known to be like you).
TDT was literally made for this kind of situation (because it’s just a Newcomblike problem). UDT differs from TDT only in areas a bit more obscure than this. EDT is also designed to handle this perfectly too (ie. to get you the book for minimal price). If you donate evidence does suggest that enough people will donate to get you the book but if you don’t donate evidence suggests that you will not.
where this assumption is so restrictive the real answer is probably “don’t donate.”
Thus we see that assurance contracts can be useful even for a population EDT/TDT/UDT agents.
TDT has to say that if the scenario where everyone donates you win, and you know that everyone else is using TDT or that the distribution of decision algorithms is likely to give sufficient “donate” outputs to make it better expected utility, then you should donate. Of course, if you have reliable data on others’ decision algorithms, I’m pretty sure CDT and EDT and any other decision theory I’ve read about will boil down to an expected utility calculation or something pretty close.
Basically, as Vaniver says, all good DTs pretty much agree on this. TDT, CDT and EDT all agree that if you have common knowledge of a sufficient number of other people using the same decision theory (or, with more complicated calculations, various possible theories including those three) are interested in the book, you should all donate. This common knowledge, however, is usually the extremely costly, high-information-value part—the part about figuring out whether to donate or not seems trivial by comparison.
I don’t think this is correct. The CDT agents would all agree that they all should donate and would support the implementation of a simple mutual commitment protocol. If they couldn’t arrange a way to compel each other to not defect on the commons problem they would be sad but defect themselves. Fortunately there are already existent online donation systems are sufficient. You just need one of the ones that returns pledged funds if the target goal isn’t met and a carefully calculated target goal.
At the extremes of perfect CDT agents you’d have to fiddle with the details a little more and, for example, make it forbidden for one agent to donate twice in order to allow that any will even donate once. But we can assume either all those details are handled or the CDT agents aren’t quite that ridiculous and consider the precommitment mechanism adequate. Another thing they would do is arrange a taxation system enforced by people with guns with the relevant commons problems to be solved specified by (necessarily compulsory) voting.
Of course, the other thing groups of CDT agents would do is arrange a free market capitalism system wherein products are payed for and people who don’t pay don’t get the stuff. A more efficient system would also allow the author easy access to a loan based on the awareness of the loan giver of the desire for the books. Then she would actually get most of the money from the sales of said books.
Right- where again the primary block is the mutual information required.
Apologies if sidetracking a hypothetical into the real world: kickstarter attempts to solve this problem.
Fuzzies?
As far as I can tell, any decision theory that disagrees with CDT in this case is mistaken. The author (or you) need to sweeten the deal; either the benefits need to be better, or the cost needs to be lower. Typical ways to improve the benefit are to attach status or other goods to the donation- whenever I talk about the Kickstarter projects I back, I make sure to mention that, you know, I backed them.
You missed an opportunity here. ;)
Yeah, but the conversation is about collective patronage in general, not about specific projects, and it seemed like it would detract from my point to also brag with my comment.