Choice architecture research has spent two decades establishing that defaults shape what people pick, but far less is known about what a default does to how people feel about the picking. The question this article takes up is narrow and testable: when a flavour menu is trimmed and a default is installed, does the shift show up only in selection shares, or does it also move the confidence with which a first pick is held? A 14-percentage-point movement in first-pick confidence, observed when a six-flavour set collapsed to two with one marked as default, is the thread worth pulling.
What a Default Actually Does to the Picker
The standard account of defaults runs through effort and endorsement. People accept the pre-selected option because switching costs something — a click, a deliberation, a moment of doubt — and because the presence of a default is read as a weak signal that someone competent already did the thinking. That account is well supported and, for retail design, usually sufficient.
But confidence is a separate dependent variable, and it behaves differently from choice share. Confidence is a metacognitive judgment: it reflects how easy the decision felt to make, how clearly the options could be compared, and how much the chooser believes the outcome was under their own control. Those inputs are sensitive to menu size in ways that selection shares are not.
The two-flavour result fits this. A binary menu with one default produces fast, low-friction comparison. The chooser evaluates one alternative against an anchor, reaches a judgment quickly, and — critically — experiences the fluency of that judgment as evidence of its quality. The six-flavour menu does the opposite. More options mean more pairwise comparisons, more attribute trade-offs, and more opportunity for the chooser to notice that they cannot rank the set cleanly. The decision still gets made, but it is made against a background hum of unresolved comparison.
That hum is what the confidence measure is picking up. The 14-point shift is not evidence that the two-flavour chooser is better informed. It is evidence that they are less aware of what they did not consider.
Fluency, Not Accuracy, Drives the Number
The mechanism is best understood through processing fluency and its documented misattribution. People routinely read the ease of a mental operation as a signal about the content of that operation. A statement that is easy to read is judged more likely to be true; a name that is easy to pronounce is judged more familiar; a decision that resolves quickly is judged more sound. This is not a failure of intelligence so much as a feature of how metacognitive cues are used — the system samples its own processing as data.
Applied to flavour menus, the implication is uncomfortable. Shrinking six options to two does not add information. It removes the visible evidence of trade-offs the chooser never resolved. The two-flavour picker holds their choice with more conviction partly because the menu gave them less to be uncertain about, not because the menu gave them a better basis for certainty.
There is a second contributor worth naming: the default itself functions as a social cue. When one of two options is pre-marked, the chooser is not comparing two peers — they are comparing a recommendation against an alternative. That asymmetry simplifies the evaluation into a yes/no question ("do I want the recommended one?") rather than an open ranking. Yes/no questions resolve faster and, in self-report, resolve more confidently. The confidence gain is real as an experience; it is simply not tracking decision quality.
What the Confidence Shift Predicts Downstream
Confidence is not a passive readout. It feeds forward into how people behave after the choice, and this is where the practical stakes sit.
Higher first-pick confidence is associated with lower rates of post-choice switching, less information-seeking before repurchase, and stronger resistance to contradictory feedback. In a flavour context, that translates into a customer who reorders without browsing, does not sample the alternatives they were never shown, and discounts a bad experience as an anomaly rather than a signal to explore. For a shop, that is retention. For the customer, it is a narrower palate and a smaller evidence base for future preferences.
There is a parallel here to the broader finding that confidence and accuracy are only loosely correlated across many judgment domains. People can be highly confident and wrong, and the gap widens when the environment has been simplified in ways that hide the error. A two-flavour menu is exactly such an environment: it is not deceptive, but it is quiet about what it omitted. The chooser's confidence is calibrated to the menu they saw, not to the space of options that existed.
This matters for how the 14-point figure should be read. It is not a claim that narrow menus produce worse choices. It is a claim that narrow menus produce more confidently held choices, and that the confidence is partly borrowed from the menu design rather than earned from the comparison. Whether that is good or bad depends entirely on what the chooser is trying to do — settle quickly, or build a durable map of what they like.
Designing Menus That Keep Confidence Honest
The forward question for anyone building a flavour selection interface is not whether to use defaults — the evidence for defaults is strong and the friction argument is sound — but how to keep the confidence signal from outrunning the information behind it. Three directions are worth testing.
Show the omitted set without restoring the burden. A collapsed "also available" line preserves the fact that six flavours exist without forcing six pairwise comparisons. The chooser keeps the fluency of a binary decision and gains an accurate sense of the menu's breadth. If confidence is being inflated by hidden alternatives, making them visible at low cost should moderate the effect without eroding the speed advantage.
Separate recommendation from default. Marking one option as "recommended" while leaving the selection blank forces an active choice and keeps the social cue without the pre-selection. Comparing confidence under this arrangement against the true-default arrangement would isolate how much of the 14 points comes from the social signal versus the friction reduction — a distinction the current result cannot make.
Measure confidence at the point of choice, not after consumption. First-pick confidence collected immediately is a fluency readout; the same question asked after tasting is contaminated by the experience itself. Keeping the two measures separate is what allows a shop to see whether a design change moved the chooser's felt certainty or the product's actual performance. Conflating them hides the mechanism.
The most useful next study would hold menu size constant and vary only the default, then hold the default constant and vary menu size, with confidence as the primary outcome in both arms. That 2×2 would tell us whether the 14-point shift is a menu-size effect, a default effect, or — as the fluency account predicts — an interaction that only appears when both are present. Until then, the honest reading is that defaults do more than steer the pick. They change how settled the picker feels, and that feeling travels.