Hey, indeed I have a bit of a unique perspective on this relative to other EAs. I would say it really depends if we are talking about a model of EV (which I think in practice is typically what we have) or the real EV. I would not trade off “true” EV but in practice I am pretty skeptical about models of EV really getting at truth, particularly when they diverge from other forms of evidence. Another way to frame it is I am comfortable with risk but discount EV models based on uncertainty as I think in practice uncertainty tends to lead to less dramatic outcomes (closer to no effect) as we gain more evidence/certainty.
Hey, indeed I have a bit of a unique perspective on this relative to other EAs. I would say it really depends if we are talking about a model of EV (which I think in practice is typically what we have) or the real EV. I would not trade off “true” EV but in practice I am pretty skeptical about models of EV really getting at truth, particularly when they diverge from other forms of evidence. Another way to frame it is I am comfortable with risk but discount EV models based on uncertainty as I think in practice uncertainty tends to lead to less dramatic outcomes (closer to no effect) as we gain more evidence/certainty.