Update: I tentatively believe I’ve resolved the confusion around action invariance, enabling a reformulation of the long term penalty which seems to converge to the same thing no matter how you structure your actions or partition the penalty interval, possibly hinting at an answer for what we can do when there is no discrete time step ontology. This in turn does away with the long-term approval noise and removes the effect where increasing action granularity could arbitrarily drive up the penalty. This new way of looking at the long-term penalty enables us to understand more precisely when and why the formulation can be gamed, justifying the need for something like IV.
In sum, I expect this fix to make the formulation more satisfying and cleanly representative of this conceptual core of impact. Furthermore, it should also eliminate up to half of the false positives I’m presently aware of, substantially relaxing the measure in an appropriate way—seemingly without loss of desirable properties.
Unfortunately, my hands are still recovering from carpal tunnel (this post didn’t write itself), so it’ll be a bit before I can write up this info.
Update: I tentatively believe I’ve resolved the confusion around action invariance, enabling a reformulation of the long term penalty which seems to converge to the same thing no matter how you structure your actions or partition the penalty interval, possibly hinting at an answer for what we can do when there is no discrete time step ontology. This in turn does away with the long-term approval noise and removes the effect where increasing action granularity could arbitrarily drive up the penalty. This new way of looking at the long-term penalty enables us to understand more precisely when and why the formulation can be gamed, justifying the need for something like IV.
In sum, I expect this fix to make the formulation more satisfying and cleanly representative of this conceptual core of impact. Furthermore, it should also eliminate up to half of the false positives I’m presently aware of, substantially relaxing the measure in an appropriate way—seemingly without loss of desirable properties.
Unfortunately, my hands are still recovering from carpal tunnel (this post didn’t write itself), so it’ll be a bit before I can write up this info.