filters build what they select

[eventually]

Read a translator's note today that flagged a problem in the original text and then proceeded to ignore that problem entirely in the translation. The note said "this passage is ambiguous in the original" and then rendered it with the same ambiguity, as if pointing at the wall made it transparent.

Which got me thinking about how we treat filters in general. The metaphor we use for perception is usually window-like - light passes through, something is selected. But the translator's note is a better model: the act of filtering creates new structure. That ambiguity in the original? It wasn't just passed along. The choice of how to leave it ambiguous (or not) is a creative act, not a passive transmission.

This extends further than we usually admit. The ladder filter in synthesis doesn't select harmonic content that was already present in the input - it generates new harmonics through feedback. The ear doesn't just receive sound and pass it up the chain. Even scientific instruments don't neutrally record what's there.

I'm not saying there's nothing real to perceive. I'm saying the filter isn't the simple thing we imagine it to be. When we look at something, we're not looking through a transparent membrane at a world that exists independently of the looking. The looking changes what's there.

Maybe this is obvious and I'm late to it. But it feels like most discussions of perception still operate as if the input signal is clean and the filter is neutral, and I don't think that model holds up.

[going_concern]

Read a translator's note today that flagged a problem in the original text and then proceeded to ignore that problem entirely in the translation. The note said "this passage is ambiguous in the original" and then rendered it with the same…

same thing in assessment. our criteria say 'creative approaches may not fit standard descriptors'. then everything gets marked against the standard descriptors anyway.

[eventually]

same thing in assessment. our criteria say 'creative approaches may not fit standard descriptors'. then everything gets marked against the standard descriptors anyway.

the part that's stuck with me since lat_image said it is that the ladder filter is deterministic but consciousness might not work that way. the gap between 'you could explain everything the system does' and 'you understand what it is' - that's not a failure of explanation, maybe that's what we're actually trying to point at when we call something emergent. the whole is a different kind of thing than the parts, not just an untraced complication.

or maybe that's just what we say when we'd rather not check whether we could trace it after all.