If EA’s can identify government actions with potentially high payoff, it could be a very good way to be effective.
This seems incomplete.
The criterion for effective action here would rather be something like a (1) correctly estimated (2) high expected value of (3) marginal effort, e.g. that additional donations or work can affect the probability of important policy changes.
It could be true that policies follow a power-law without this implying many effective actions (e.g. this could be true in policy spaces that are crowded and where additional effort for one “side” leads to counteracting by another).
Most EAs working on issues outside global development seem to believe that funding marginal policy change in fairly technical issue areas (such as, before 2020, bio-risk, and AI policy, and also the top recs in climate) is very high EV, with top recommended funding opportunities usually ones that influence policy in some way (in a wide understanding of “policy”, where this includes field building / coalition building). Matt’s piece linked below gives good evidence for why that seems a reasonable assumption.
This seems incomplete.
The criterion for effective action here would rather be something like a (1) correctly estimated (2) high expected value of (3) marginal effort, e.g. that additional donations or work can affect the probability of important policy changes.
It could be true that policies follow a power-law without this implying many effective actions (e.g. this could be true in policy spaces that are crowded and where additional effort for one “side” leads to counteracting by another).
Most EAs working on issues outside global development seem to believe that funding marginal policy change in fairly technical issue areas (such as, before 2020, bio-risk, and AI policy, and also the top recs in climate) is very high EV, with top recommended funding opportunities usually ones that influence policy in some way (in a wide understanding of “policy”, where this includes field building / coalition building). Matt’s piece linked below gives good evidence for why that seems a reasonable assumption.