Gerrymandering illustrates how automation in policy can be better than discretion. If you give a legislature the power to chop things up as they will, they will do so in ways that benefit the dominant local party. A better alternative is for an algorithm to draw districts instead, regularly revising so as to keep seats across a territory in line with popular vote (so that, e.g., when 40% of Californians vote GOP, GOP gets 40% of the House seats for California).
Bitcoin obeys a similar principle. Rather than delegating monetary policy to trusted authorities, it automates that policy in highly predictable ways, and regularly revises (i.e., the difficulty adjustment) to keep things in line with what's expected. The outcomes needn't be optimal for this system to be superior, note; for there is no guarantee that those trusted parties will enact optimal policies, and they often fail in this task! The best argument for automated policy, then, is not that it is for the best always and everywhere, but that it is typically for the better.
