x
Learned logic of modelling harm — LessWrong