To continue with my theme about the bias towards action, I would note the following. Suppose that one periodically samples a random variable to decide whether the correct action is to leave some situation alone, or to intervene. Assuming that one continues sampling after getting back “do nothing”, but that an “intervene” decision is final, it should be clear that “intervene” will always win eventually, if the random variable has even a tiny probability of coming up “intervene”, even if the vast majority of the probability mass is on “do nothing”.
So in light of that, if one is going to continue to stand around and talk about intervening, one should probably bias further and further away from intervening as time passes, to account for the fact that eventually the coin will come up “intervene” through bad luck no matter what the correct decision is.
To continue with my theme about the bias towards action, I would note the following. Suppose that one periodically samples a random variable to decide whether the correct action is to leave some situation alone, or to intervene. Assuming that one continues sampling after getting back “do nothing”, but that an “intervene” decision is final, it should be clear that “intervene” will always win eventually, if the random variable has even a tiny probability of coming up “intervene”, even if the vast majority of the probability mass is on “do nothing”.
So in light of that, if one is going to continue to stand around and talk about intervening, one should probably bias further and further away from intervening as time passes, to account for the fact that eventually the coin will come up “intervene” through bad luck no matter what the correct decision is.