Additive and Multiplicative Subagents — LessWrong