Formal statements tend to be unclear when they’re really proofs that one’s attempts to place line breaks in somehow failed.
The problem is that proving that the agent performs the action proves that the agent does not perform some other action. From this, the agent can prove very high utility given that action. An agent which proves very high utility given an action does not take any other action.
So: An agent that proves that it takes an action, as long as it has enough time left to extend the proof a bit further, does not take that action.
That’s basically the reason consequences appear consistent.
If the agent can change its mind, then naturally it can’t prove that it won’t. I was assuming a situation where the agent is reasoning further after it has already decided on the action, so that further moral arguments are not.
Formal statements tend to be unclear when they’re really proofs that one’s attempts to place line breaks in somehow failed.
The problem is that proving that the agent performs the action proves that the agent does not perform some other action. From this, the agent can prove very high utility given that action. An agent which proves very high utility given an action does not take any other action.
So: An agent that proves that it takes an action, as long as it has enough time left to extend the proof a bit further, does not take that action.
To create explicit line breaks, add two spaces to the end of each line.
That’s basically the reason consequences appear consistent.
If the agent can change its mind, then naturally it can’t prove that it won’t. I was assuming a situation where the agent is reasoning further after it has already decided on the action, so that further moral arguments are not.