...and this leads to another bit of progress, on the structure of mathematical intuition. The key exercise is to try to explain explicit optimization of strategies, as described in your post, in terms of local ambient control that determines only the action. The strategies then become the assumptions of the proof search that tries to guess what will be the global outcome. The solution comes from two agents being equivalent without the observation, thus “input” is automatically extracted from the agents’ code (if we assume the input to be just part of the code by the time decision problem gets stated). I’ll describe it in more detail later, if this pans out.
...and this leads to another bit of progress, on the structure of mathematical intuition. The key exercise is to try to explain explicit optimization of strategies, as described in your post, in terms of local ambient control that determines only the action. The strategies then become the assumptions of the proof search that tries to guess what will be the global outcome. The solution comes from two agents being equivalent without the observation, thus “input” is automatically extracted from the agents’ code (if we assume the input to be just part of the code by the time decision problem gets stated). I’ll describe it in more detail later, if this pans out.