Cognitive load

The useful split, borrowed from cognitive load theory, is between the load a task carries inherently and the load the design added on top. Choosing a delivery date is intrinsic: the decision exists whether or not anyone designs a screen for it. Nine ungrouped options with names taken from a file format specification, no answer filled in, and a paragraph of qualifying copy above them are extraneous, and extraneous load is the only kind a redesign can actually remove. That distinction is what keeps the term from collapsing into “this feels busy”, because it names which of the busyness was avoidable.

Working memory is the resource being spent, and it is small and short. The classic figure of seven items has been revised downward for years, and the practical reading is that anything a reader has to carry from one screen to the next is expensive: a code from an email, a total from a previous step, the meaning of an abbreviation defined earlier. The cheapest fixes are the boring ones. Group the choices so the eye reads three clusters instead of nine items. Write the labels in the words the reader already has. Answer what you can already answer, since a smart default turns a decision into a confirmation. Hide the rare controls behind progressive disclosure rather than ranking everything at once. Keep the earlier answers on screen, which is the whole argument for a check answers page.

This matters most for the people the accessibility literature calls cognitively disabled, which is a very wide group: readers with a learning disability, an attention or memory condition, aphasia, a brain injury, or simply the ordinary decline of processing speed with age. It also covers a much larger temporary population, since anyone doing this on a phone, on a bus, tired, in a second language, or in the middle of something else is running with less working memory than they would in a usability lab. Cognitive accessibility guidance largely reads as good design advice for everybody, and that is not a coincidence.

Be honest about measurement. Load is not read off a screen: it is inferred from proxies such as task completion time, error rate, backtracking, pupil dilation, or a self-report scale like NASA-TLX, all of which measure something adjacent to the thing. A design review can still be rigorous without a number. Count the decisions the screen asks for, count the ones it could have answered itself, count the words the reader would have to look up, and count what they are being asked to remember from a previous screen. Those four counts are arguable in a way that “it feels cluttered” is not.