It is useful to have a black list of problem indicators, if you can train yourself so these indicators jump out at you. But also approach this problem from the other side, building a white list of good cognitive states. When should you not scream “Error!”?
But also approach this problem from the other side, building a white list of good cognitive states. When should you not scream “Error!”?
Some noteworthy cases come to mind:
When concluding that a decision you made more than a minute ago was a mistake (because the error signal will be misattributed to the investigation, rather than the mistake).
Within the first minute of starting a task (because it will attach to the act of starting and make you a procrastinator)
When evaluating statements that are especially emotionally salient (because the error signal is more likely to correspond to unpleasant conclusions than to falsehoods)
When noticing that your brain has screamed Error! at you (because you need to clear that signal before evaluating the original error)
You seem to have interpreted my question to mean “When do you scream ‘Error!’ when you shouldn’t?”. While your response contains valuable answers to that question, what I had in mind was building a (nearly) comprehensive list of indicators that you are currently being an effective rationalist, so that when you don’t notice any good indicator, you can scream “Error!”.
I treat not screaming “Error!” as the default state, and I think that you should at least pay attention to problem indicators even if you otherwise feel like you’re grokking well, as doing so might help to prevent you from going down the wrong mental path.
Do you have examples of white list states, where you should be more prone to screaming “Error!” if you don’t have many/any of them?
I treat not screaming “Error!” as the default state, and I think that you should at least pay attention to problem indicators even if you otherwise feel like you’re grokking well, as doing so might help to prevent you from going down the wrong mental path.
I am not dismissing your approach. I am saying do both.
Do you have examples of white list states, where you should be more prone to screaming “Error!” if you don’t have many/any of them?
My first example is that you are pursuing a worthwhile goal, which is an attempt at refining Anna’s “use curiosity” heuristic. (Notice how even though Anna presented it as lack of curiosity is a negative indicator, I see it as curiosity is a positive indicator.)
Okay, that makes sense to me. However, I think that our two lists are working in two different domains. I think that mine is best at determining whether you could be more effective in achieving a goal while yours is better at determining whether your current actions line up with more terminal goals.
It is useful to have a black list of problem indicators, if you can train yourself so these indicators jump out at you. But also approach this problem from the other side, building a white list of good cognitive states. When should you not scream “Error!”?
Some noteworthy cases come to mind:
When concluding that a decision you made more than a minute ago was a mistake (because the error signal will be misattributed to the investigation, rather than the mistake).
Within the first minute of starting a task (because it will attach to the act of starting and make you a procrastinator)
When evaluating statements that are especially emotionally salient (because the error signal is more likely to correspond to unpleasant conclusions than to falsehoods)
When noticing that your brain has screamed Error! at you (because you need to clear that signal before evaluating the original error)
You seem to have interpreted my question to mean “When do you scream ‘Error!’ when you shouldn’t?”. While your response contains valuable answers to that question, what I had in mind was building a (nearly) comprehensive list of indicators that you are currently being an effective rationalist, so that when you don’t notice any good indicator, you can scream “Error!”.
I treat not screaming “Error!” as the default state, and I think that you should at least pay attention to problem indicators even if you otherwise feel like you’re grokking well, as doing so might help to prevent you from going down the wrong mental path.
Do you have examples of white list states, where you should be more prone to screaming “Error!” if you don’t have many/any of them?
I am not dismissing your approach. I am saying do both.
My first example is that you are pursuing a worthwhile goal, which is an attempt at refining Anna’s “use curiosity” heuristic. (Notice how even though Anna presented it as lack of curiosity is a negative indicator, I see it as curiosity is a positive indicator.)
Okay, that makes sense to me. However, I think that our two lists are working in two different domains. I think that mine is best at determining whether you could be more effective in achieving a goal while yours is better at determining whether your current actions line up with more terminal goals.