Are you always right?

I've been thinking a lot recently about feedback systems, in terms of failures and other signals to focus your attention, and also in terms of frames of reference that you can use to measure your progress (or lack thereof). I think the defining quality of any feedback system is that it gives useful feedback, where I would define useful as effectively partitioning the action-space.

I covered a similar concept, partitioning idea-space, in an earlier post on strong and meaningful statements. Much like a space of ideas, you can imagine there being a space of all actions, and you want to reduce that space to a manageable set. Your feedback system should act like a knife cutting into the cake of all possible actions, separating the good actions from the bad ones, and ideally separating the most possible actions with each cut.

My recent situation of being in a semi-permanent state of behind with my posts is an example of feedback that isn't useful because it doesn't partition the action space. All that signal can tell me is "you're still failing". It can't effectively partition my actions into ones that have led to failure and ones that have led to success because they all keep leading to failure. Conversely, though, a feedback system that is always nice to you is just as useless. Instead of saying you're failing all the time, it says you're succeeding all the time. Either way it can't separate your actions from one another.

Always being right is more tempting than always being wrong, for mostly egotistical reasons. There's also an implies both ways problem: improving means you get to be right, but being right doesn't mean you're improving. It's easy to accidentally put yourself into a situation where all your signals are necessarily green. Everything's going great! Your actions are definitely still the correct ones!

But it's important to bring it back to this idea of useful feedback, of meaningfully partitioning some actions from others. If your feedback is no longer doing that, it's stopped being useful feedback. If you want to make better decisions, you'll need to find better feedback, to expose yourself to harsher standards or more difficult problems. It's possible to always be right, but only at the cost of being unable to improve.

I think it's better to be wrong some of the time.