TLDR: Everyone is clueless, even experts, and p1 relies on the counterfactual that we would chose actions / general EV in an ideally sorted causal chain.
Would like some clarification, since for me preference in one case before making a decision does not move the consequences or expected value of the decision. If you were to use expert advice to c-prefer A over B, both agents are still trapped in an infinitely causal universe. C-prefer = sorting out causality, which is impossible if we admit p3 (“cosmos-wide consequences”). The premises rely on each other to grant the conclusion.
“Evidence need not contain the reasoning that produced it,” or “testimony can transmit information without transmitting the underlying reasoning.” You would need to believe a doctor who is telling you treatment A is better than treatment B based on evidence. Your framing is that, if the doctor read 1000 pages on each treatment, the deciding agent would inherit the doctor’s experience as evidence.
I am wondering what work this is doing because I cannot quite formulate why I think this is more important than meets the eye. I think this example is unique. Perhaps the whole thing is that we are expounding moral cluelessness, and for this case, existing (being healthy, using doctor’s advice) is not in itself good (asymmetry argument). I say this just to propose assumptions that may be hidden.
Trusting a financial advisor to do A vs. B given you do not know the market is not moral cluelessness, and even if trusting the advisor led to a successful returns, we cannot turn back time to see what would have happened if you went with B (given things can only happen once, turning back time maybe may generate a new reality / causal path).
I guess, if a (more) idealized agent (e.g., an expert) says A is better than B and you have reasons to believe this then we are closing in perhaps on better EV decisions. But I think EV in terms of cluelessness directly equals a fully correct model where one would know exactly what would happen from an action until the ‘end of time.’
What’s most interesting is that waiting for more information generates better decisions. Again, what I am most concerned about here is that no decision would be ideal in a messy future. So trusting testimony or waiting are not converging on a “better” finite ideal, but rather remain blanket unawareness. So normatively we would prefer a decision process where this is not the case (p1).
TLDR: Everyone is clueless, even experts, and p1 relies on the counterfactual that we would chose actions / general EV in an ideally sorted causal chain.
Would like some clarification, since for me preference in one case before making a decision does not move the consequences or expected value of the decision. If you were to use expert advice to c-prefer A over B, both agents are still trapped in an infinitely causal universe. C-prefer = sorting out causality, which is impossible if we admit p3 (“cosmos-wide consequences”). The premises rely on each other to grant the conclusion.
“Evidence need not contain the reasoning that produced it,” or “testimony can transmit information without transmitting the underlying reasoning.” You would need to believe a doctor who is telling you treatment A is better than treatment B based on evidence. Your framing is that, if the doctor read 1000 pages on each treatment, the deciding agent would inherit the doctor’s experience as evidence.
I am wondering what work this is doing because I cannot quite formulate why I think this is more important than meets the eye. I think this example is unique. Perhaps the whole thing is that we are expounding moral cluelessness, and for this case, existing (being healthy, using doctor’s advice) is not in itself good (asymmetry argument). I say this just to propose assumptions that may be hidden.
Trusting a financial advisor to do A vs. B given you do not know the market is not moral cluelessness, and even if trusting the advisor led to a successful returns, we cannot turn back time to see what would have happened if you went with B (given things can only happen once, turning back time maybe may generate a new reality / causal path).
I guess, if a (more) idealized agent (e.g., an expert) says A is better than B and you have reasons to believe this then we are closing in perhaps on better EV decisions. But I think EV in terms of cluelessness directly equals a fully correct model where one would know exactly what would happen from an action until the ‘end of time.’
What’s most interesting is that waiting for more information generates better decisions. Again, what I am most concerned about here is that no decision would be ideal in a messy future. So trusting testimony or waiting are not converging on a “better” finite ideal, but rather remain blanket unawareness. So normatively we would prefer a decision process where this is not the case (p1).