Option count beyond comparison capacity is wasted

What it is

A cover tool lays out twelve thumbnails. The user parks on the two at top left; the other ten never even dwell; the one adopted is the one glanced at. Extra options that never enter comparison might as well not have been generated. Comparison capacity is how many options a person can hold in a contrast set at once. Count past that window carries zero decision weight.

This is not regret after a pick, and it is not options looking alike. It is that an unread candidate did not take part in the choice.

Why it happens

Simultaneous comparison leans on visual working memory and pairwise contrast. Most people can stably contrast about two to four items side by side; beyond that they scan-and-satisfice, and items outside the window never join the set. Generation is paid per item; decision information is counted per processed item — the two split once count overflows.

Sequential browse (swipe one by one) and juxtaposition are different capacities. Sequence trades space for time, but earlier items are pushed out of memory by later ones, so what is actually compared is still a short window. “Generate many” therefore does not become “choose better”; it stockpiles unread items.

Studying it

Fix the prompt; set option count to 2, 3, 5, 8, 12. Use eye tracking or scroll dwell to mark whether each option was processed (fixation or dwell above a threshold). Independent variables: simultaneous versus sequential presentation, thumbnail versus full text. Dependent variables: share of options processed, whether the final pick landed on an unprocessed item, rate of going back for a second look.

Split “picked without having looked” from “looked and suffered.” The first is capacity overflow; the second is choice load. Total time alone mixes the two failures.

Where it stops holding

When the user arrives with an explicit retrieval criterion (“landscape, dark, no faces”), scan capacity exceeds preference-comparison capacity and extras become a filter. Experts with stable evaluation dimensions for a genre can scan more. On a very narrow screen even three options become a sequence, and capacity decays by sequential rules. When generation is cheap and the user is explicitly browsing, unread items are inventory rather than failure — provided the interface does not pretend all twelve were compared.

Applying it

• On the first response, keep the juxtaposed count inside the comparison window, usually two to four complete options; put the rest behind “another batch,” not a single dump.
• In juxtaposition, each option should be fully visible on the first screen. Anything that shows as a half thumbnail never entered the window.
• If a larger set is required, switch to filtered browsing and show seen / unseen. Do not use a large grid to imply that comparison is happening.
• Check: from a recording, count how many options were visible for more than two seconds. If the final pick often lands on an item visible under two seconds, count has already exceeded comparison capacity.

Related

• Same group: L3.01.1 Multiple options turn generation into a choice · L3.01.3 Options need discriminable differences
• Nearby: L3.07 Multi-option Generation and Side-by-side Comparison · L1.01 Mismatch Between Stochastic Output and Deterministic UI
• Search terms: comparison capacity · option count · satisficing