Agreed that, when testable propositions are involved, I use this as a mechanism for artificially adjusting my expectations of the results of an as-yet-unperformed test (or an already performed test whose results I don’t know).
Adjusting my existing knowledge of an already performed test would be… trickier. I’m not sure how I would do that, short of extreme measures like self=hypnosis.
Agreed that arbitrarily increasing my confidence level simply because I can is not a good idea, and therefore that, as you say, I increase it if there are other factors in place that lead me to think it’s a good idea.
That said, those “other factors” aren’t necessarily themselves evidence of likely success, which seems to be an implication of what you’re saying.
To pick an extreme example for illustrative purposes: suppose I estimate that if I charge the armed guards wholeheartedly, I will trigger a mass charge of other prisoners doing the same thing, resulting in most of us getting free and some of us getting killed by the guards, and that this the best available result. Suppose I also estimate that, if I charge the guards, I will likely be one of the ones who dies. Suppose, I further estimate that I am not sufficiently disciplined to be able to charge the guards wholeheartedly while believing I will die; if I do that, the result will be a diffident charge that will not trigger a mass charge.
Given all of that, I may choose to increase my confidence level in my own survival despite believing that my current confidence level is accurate, because I conclude that a higher confidence level is useful.
Of course, a perfect rationalist would not need such a mechanism, and if one were available I would happily share my reasoning with them and wait for them to charge the fence instead, but they are in short supply.
Of course, these sorts of scenarios are rare. But it’s actually not uncommon to enter into situations where I’ve never done X before, and I don’t really know how difficult X is, so a prior probability of 50% of success/failure seems reasonable… but I also suspect that entering the situation with a 50% estimate of success will make failure more likely than entering with an 85% estimate of success… so I artificially pick a higher prior, because it’s useful to do so.
So, while I would not say it’s always beneficial to anticipate success, I would say that it’s sometimes beneficial even if one has good reason to suspect failure.
Whether a trip to Pluto could ever be such an example, and how I might go about artificially raising my estimate of success in that case, and what the knock-on effects of doing so might be… I don’t know. I can’t think of a plausible scenario where it would be a good idea.
Well, this has been a lovely discussion. Thanks for the back and forth; I think we’re in agreement, and your last example was particularly helpful. I think we’ve covered that:
we’re not talking about arbitrarily increasing confidence for no reason (just because we can)
we’re also [probably] not talking about trying to increase belief in something contrary to evidence already known (increase belief in ~X when the evidence supports X). (This is actually the category I originally thought you referring to, hence my mention of “tricking” one’s self. But I think this category is now ruled out.)
this technique is primarily useful when emotions/motivations/feelings are not lining up with the expected outcome given available evidence (success is likely based on prior experience, but success doesn’t feel likely and this is actually increasing likelihood of failure)
there are even some situations when an expectation of failure would decrease some kind of utilitarian benefit and thus one needs to act as if success is more probable, even though it’s not (with the caveat that improving rationality would help this not be necessary)
Agreed that, when testable propositions are involved, I use this as a mechanism for artificially adjusting my expectations of the results of an as-yet-unperformed test (or an already performed test whose results I don’t know).
Adjusting my existing knowledge of an already performed test would be… trickier. I’m not sure how I would do that, short of extreme measures like self=hypnosis.
Agreed that arbitrarily increasing my confidence level simply because I can is not a good idea, and therefore that, as you say, I increase it if there are other factors in place that lead me to think it’s a good idea.
That said, those “other factors” aren’t necessarily themselves evidence of likely success, which seems to be an implication of what you’re saying.
To pick an extreme example for illustrative purposes: suppose I estimate that if I charge the armed guards wholeheartedly, I will trigger a mass charge of other prisoners doing the same thing, resulting in most of us getting free and some of us getting killed by the guards, and that this the best available result. Suppose I also estimate that, if I charge the guards, I will likely be one of the ones who dies. Suppose, I further estimate that I am not sufficiently disciplined to be able to charge the guards wholeheartedly while believing I will die; if I do that, the result will be a diffident charge that will not trigger a mass charge.
Given all of that, I may choose to increase my confidence level in my own survival despite believing that my current confidence level is accurate, because I conclude that a higher confidence level is useful.
Of course, a perfect rationalist would not need such a mechanism, and if one were available I would happily share my reasoning with them and wait for them to charge the fence instead, but they are in short supply.
Of course, these sorts of scenarios are rare. But it’s actually not uncommon to enter into situations where I’ve never done X before, and I don’t really know how difficult X is, so a prior probability of 50% of success/failure seems reasonable… but I also suspect that entering the situation with a 50% estimate of success will make failure more likely than entering with an 85% estimate of success… so I artificially pick a higher prior, because it’s useful to do so.
So, while I would not say it’s always beneficial to anticipate success, I would say that it’s sometimes beneficial even if one has good reason to suspect failure.
Whether a trip to Pluto could ever be such an example, and how I might go about artificially raising my estimate of success in that case, and what the knock-on effects of doing so might be… I don’t know. I can’t think of a plausible scenario where it would be a good idea.
Well, this has been a lovely discussion. Thanks for the back and forth; I think we’re in agreement, and your last example was particularly helpful. I think we’ve covered that:
we’re not talking about arbitrarily increasing confidence for no reason (just because we can)
we’re also [probably] not talking about trying to increase belief in something contrary to evidence already known (increase belief in ~X when the evidence supports X). (This is actually the category I originally thought you referring to, hence my mention of “tricking” one’s self. But I think this category is now ruled out.)
this technique is primarily useful when emotions/motivations/feelings are not lining up with the expected outcome given available evidence (success is likely based on prior experience, but success doesn’t feel likely and this is actually increasing likelihood of failure)
there are even some situations when an expectation of failure would decrease some kind of utilitarian benefit and thus one needs to act as if success is more probable, even though it’s not (with the caveat that improving rationality would help this not be necessary)
Does that about sum it up?
Thanks again.
Works for me!