Of course you shouldn’t plan to reset the zero point after actions! That’s very different.
I use this sparingly, for observing big new facts that I didn’t cause to be true. That doesn’t change the relative expected utilities of various actions, so long as my expected change in utility from future observations is zero.
Of course you shouldn’t plan to reset the zero point after actions! That’s very different.
I use this sparingly, for observing big new facts that I didn’t cause to be true. That doesn’t change the relative expected utilities of various actions, so long as my expected change in utility from future observations is zero.