My usual response to requests for “X exists” to be tabooed is to start talking about reliably predicting future experiences E2 in a range of contexts C (as C approaches infinity) consistent with the past experiences E1 which led me to to put X in my model in the first place. If someone wants to talk about E2 being reliably predictable even though X “doesn’t really exist”, it’s not in the least bit clear to me what they’re talking about.
Thanks! This is a very useful explanation / reduction / taboo.
It also sheds some light and helped me understand quite a bit more, I believe, on this whole “instrumentalism” business some people here seem to really want to protect.
(link is just in case someone misunderstands this as an accusation of “Politics!”)
You’re welcome. I vaguely remember being involved in an earlier discussion that covered this idea at greater length, wherein I described myself as a compatibilist when it comes to instrumentalism, but the obvious google search doesn’t find it so perhaps I’m deluded.
the “right” probability distribution is the one that maximizes the expected utility of an expected utility maximizer using that probability distribution.
reliably predicting future experiences E2 in a range of contexts C (as C approaches infinity) consistent with the past experiences E1 which led me to to put X in my model in the first place.
I wholeheartedly approve of this approach. If more people used it, we would avoid the recurrent unproductive discussions of QM interpretations, qualia and such.
EDIT. Just to clarify, the part saying “put X in my model” is the essential bit to preempt the discussion of “but does it exist outside your model?”, since the latter would violate this definition of “exist”. such as this statement by our esteemed Kaj Sotala:
why those beings actually have qualia, and don’t merely act like it.
My usual response to requests for “X exists” to be tabooed is to start talking about reliably predicting future experiences E2 in a range of contexts C (as C approaches infinity) consistent with the past experiences E1 which led me to to put X in my model in the first place. If someone wants to talk about E2 being reliably predictable even though X “doesn’t really exist”, it’s not in the least bit clear to me what they’re talking about.
Thanks! This is a very useful explanation / reduction / taboo.
It also sheds some light and helped me understand quite a bit more, I believe, on this whole “instrumentalism” business some people here seem to really want to protect.
(link is just in case someone misunderstands this as an accusation of “Politics!”)
You’re welcome. I vaguely remember being involved in an earlier discussion that covered this idea at greater length, wherein I described myself as a compatibilist when it comes to instrumentalism, but the obvious google search doesn’t find it so perhaps I’m deluded.
Was it from a couple days ago?
(I found this with Wei Dai’s lesswrong_user.php script.)
Ayup, that’s the one. Thanks!
Yes. I recently described it as this:
I wholeheartedly approve of this approach. If more people used it, we would avoid the recurrent unproductive discussions of QM interpretations, qualia and such.
EDIT. Just to clarify, the part saying “put X in my model” is the essential bit to preempt the discussion of “but does it exist outside your model?”, since the latter would violate this definition of “exist”. such as this statement by our esteemed Kaj Sotala:
Oh, I very much doubt that. But I’d like to think so.
EDIT: I wrote the above before your edit, and don’t really understand your edit.
Instrumentalism is pretty unproductive when it comes to answering questions about what really exists.