Exploring an idea that I’m tentatively calling “adversarial philosophical attacks”—there seem to be a subset of philosophical problems that come up (only?) under conditions where you are dealing with an adversary who knows your internal system of ethics. An example might be Pascal’s mugger—the mugging can only work (assuming it works at all) if the mugger is able to give probabilities which break your discounting function. This requires either getting lucky with a large enough stated number, or having some idea of the victim’s internal model. I feel like there should be better (or more “pure”) examples, but I’m having trouble bringing any to mind right now. If you can think of any, please share in the comments!
The starting ground for this line of thought was that intuitively, it seems to me that for all existing formal moral frameworks, there are specific “manufactured” cases you can give where they break down. Alternatively, since no moral framework is yet fully proven (to the same degree that, say, propositional calculus has been proven consistent), it seems reasonable that an adversary could use known points where the philosophy relies on axioms or unfinished holes/quandaries to “debunk” the framework in a seemingly convincing manner.
I’m not sure I’m thinking about this clearly enough, and I’m definitely not fully communicating what I’m intending to, but I think there is some important truth in this general area....
I think you’ll find some resistance in the philosophical community about acknowledging the adversarial nature of such situations. The given thought experiments are NOT predicated on the “other” agent wanting to harm or trick the agent in question (in fact, motives of Omega or the mugger are conspicuously absent), but the experiments themselves are chosen to find the limits of a decision theory. The creator of the thought experiment is adversarial, not the hypothetical participants.
That said, I fully agree that there’s a blind spot in many of these discussions, related to which agents have what power that gives them orders of magnitude more control over the situation than the agent the problem states is making the decisions. An Omega who cheats to fuck with you makes for uninteresting decision theory question, but IMO is far FAR more likely to actually be encountered by the majority of human-level beings.
Exploring an idea that I’m tentatively calling “adversarial philosophical attacks”—there seem to be a subset of philosophical problems that come up (only?) under conditions where you are dealing with an adversary who knows your internal system of ethics. An example might be Pascal’s mugger—the mugging can only work (assuming it works at all) if the mugger is able to give probabilities which break your discounting function. This requires either getting lucky with a large enough stated number, or having some idea of the victim’s internal model. I feel like there should be better (or more “pure”) examples, but I’m having trouble bringing any to mind right now. If you can think of any, please share in the comments!
The starting ground for this line of thought was that intuitively, it seems to me that for all existing formal moral frameworks, there are specific “manufactured” cases you can give where they break down. Alternatively, since no moral framework is yet fully proven (to the same degree that, say, propositional calculus has been proven consistent), it seems reasonable that an adversary could use known points where the philosophy relies on axioms or unfinished holes/quandaries to “debunk” the framework in a seemingly convincing manner.
I’m not sure I’m thinking about this clearly enough, and I’m definitely not fully communicating what I’m intending to, but I think there is some important truth in this general area....
I think you’ll find some resistance in the philosophical community about acknowledging the adversarial nature of such situations. The given thought experiments are NOT predicated on the “other” agent wanting to harm or trick the agent in question (in fact, motives of Omega or the mugger are conspicuously absent), but the experiments themselves are chosen to find the limits of a decision theory. The creator of the thought experiment is adversarial, not the hypothetical participants.
That said, I fully agree that there’s a blind spot in many of these discussions, related to which agents have what power that gives them orders of magnitude more control over the situation than the agent the problem states is making the decisions. An Omega who cheats to fuck with you makes for uninteresting decision theory question, but IMO is far FAR more likely to actually be encountered by the majority of human-level beings.