Cognitive load theory and what it actually tells us

I was reading a mid-2000s UX report last week that kept invoking "cognitive load" to explain why a three-step form performed better than a five-step one. No measurement of actual cognition. No baseline. Just the phrase doing the explanatory work, the way people used to say "the humors are imbalanced." Here's what I keep running into: cognitive load is useful as a *naming device*. It lets you point at something real—people struggle more with complex tasks—and gives you vocabulary for it. But somewhere between Sweller's original work on worked examples and how it gets deployed now, we've lost the specificity. You see it invoked to explain everything from dashboard design to why nobody reads terms of service. It's become what Hirschman called a "pseudo-explanation": it sounds like it explains something but mostly just relabels the observation. The honest version is narrower. Sweller's experiments with instructional design were fairly concrete—he was measuring how *specific kinds* of extraneous information in math problems affected learning outcomes. That's a real finding. But "cognitive load" as a general framework for why things are hard? That's too broad to be predictive. What actually matters is *which* cognitive resources are being taxed, in what order, under what conditions. A chess master and a novice experience the same position with completely different load profiles. One person's clarity is another's noise. I wonder if we're spending energy defending or debunking a framework that was always more useful for design *intuition* than for prediction. The real work happens when you get specific about the mechanism.