Too many options is a design problem

Twenty years ago the hard part of most decisions was finding options. Jobs meant whatever was within commuting distance. Software meant the three vendors your industry used. Now the job market is remote-global, every category of tool has forty credible alternatives, and AI will generate twelve more variants of anything while you're still reading the first one. The scarce resource stopped being options a while ago. The scarce resource is a decided mind.

This shows up sharpest where AI adoption is furthest along. A June 2026 synthesis of practitioner reports on AI-assisted software engineering (Garousi, arXiv) describes engineers facing "a flood of completions, refactorings, and alternatives" — the effort has moved from writing code to evaluating and pruning plausible candidates. That's a grey-literature synthesis, not a controlled study, so hold it loosely. But it names the shape of the problem: generation is cheap, and the bill got forwarded to selection.

So what does the science of "too many options" actually say, now that its most famous results have been through the replication wringer?

What's left of the jam study

The origin story: in 2000, Iyengar and Lepper set up a supermarket tasting booth and found that a display of 24 jams attracted more browsers than a display of 6, but converted them to buyers at a tenth of the rate. "Less is more" became a keynote staple, and Barry Schwartz's The Paradox of Choice (2004) made it a worldview.

The audit was less kind. A 2010 meta-analysis of 50 experiments by Scheibehenne, Greifeneder, and Todd found a mean effect of assortment size on choice that was essentially zero — for every study where big menus hurt, another showed they helped or did nothing. The current consensus comes from Chernev, Böckenholt, and Goodman's 2015 meta-analysis in the Journal of Consumer Psychology (99 observations, N = 7,202): choice overload is real but conditional. It reliably appears when four moderators line up — the choice set is complex, the task is difficult, the chooser's preferences are uncertain, and the goal is to minimize effort. Account for those, and the overall effect turns significant again.

Read that moderator list twice, because it's the whole article. Overload is not a property of the menu. It's a property of the match between the menu and the chooser. The single moderator you control on a Tuesday afternoon is preference uncertainty: whether you know, before looking, what you're optimizing for. Twenty-four jams overwhelm someone who hasn't decided whether they care about sugar, price, or novelty. They barely slow down someone who has. That's the fact; the interpretation — that articulated preferences are the cheap lever — is mine, but it falls straight out of the data.

Decision fatigue, audited

The companion legend fared worse. The famous "hungry judges" paper (Danziger et al., 2011) reported Israeli parole approvals falling from about 65% to near zero before food breaks. It's still quoted as if settled. It isn't. Glöckner's November 2016 simulation study showed that because favorable rulings take longer (~7.4 vs ~5.2 minutes), judges rationally avoiding long cases right before a break produce a similar plunge with no depletion at all — an artifact, compounded by non-random case ordering that other critics flagged. The underlying "ego depletion" theory took its own hit when a 2016 multi-lab replication (23 labs, over 2,000 participants) failed to find the canonical effect.

The newest field evidence is blunter still. A registered report in Communications Psychology (Andersson et al., February 2025) analyzed 231,076 triage decisions by 174 Swedish nurses and found the data favored the null — Bayes factors above 22 against any decision-fatigue effect on these professional judgments. Meanwhile a January 2026 integrative review in *Frontiers in Cognition* surveys what survives: across 23 field studies, long unbroken sessions, heavy decision loads, and high complexity do correlate with more conservative, lower-quality choices — but the mechanism now looks like shifting motivation and attention, not a willpower tank running dry.

Honest summary: your brain is not a phone battery, and skipping lunch will not turn you into judge zero-percent. But hundreds of unstructured, effortful judgments in a row genuinely degrade output. Count the decisions you take without a structure — that's the number that predicts trouble.

The practitioner workaround

While the lab sorted itself out, operators converged on triage. The canonical version is Bezos's 2016 shareholder letter: most decisions are reversible "two-way doors" deserving a lightweight process and roughly 70% of the information you wish you had; the rare one-way doors get the slow, deliberate treatment. A decade later "decision velocity" is standard startup vocabulary, and the essays keep multiplying — most of them restating the same move: classify by reversibility first, then spend attention where reversal is expensive.

Herbert Simon got there in 1956 with satisficing: set an acceptability threshold, take the first option that clears it, stop looking. Note what both frameworks quietly require. You cannot classify a door or set a threshold without criteria decided in advance. Every working answer to abundance smuggles in the same step — a small act of legislation before the browsing starts.

Weights before scores

Which is the actual cure. Shrinking the menu doesn't scale (the menus are infinite now), and "trust your gut" just runs the overload at higher speed. Decide what matters before you look at what's available.

Recommendation, in procedure form — criteria-first shopping:

1. Before opening a single listing, write your criteria. Seven maximum; if you have twelve, some are decorations.
2. Weight them 1–10, and commit. This is the legislation step. Done before exposure, weights encode your priorities. Done after, they encode your crush.
3. Cap intake. Take the first 5–8 options that pass one hard filter. Chernev's moderators say complexity fuels overload; you control the complexity.
4. Score every option against every criterion. Cells, not vibes. Pros-and-cons lists lie by omission; a filled grid can't.
5. If you itch to change a weight after seeing the scores, stop. Write down what the itch is telling you first. Sometimes it's a real criterion you forgot. Usually it's a favorite lobbying for a recount — and tweaking weights until your favorite wins is astrology with spreadsheets.

Then stress-test: find the smallest weight change that flips the winner. If "price" moving from 6 to 7 crowns a different option, your verdict is a coin toss wearing a lab coat — go gather better information on exactly that criterion and nothing else. If no plausible shift flips it, you're done; stop researching. That flip-point check is the core move in Flipweight, but a spreadsheet and ten minutes will do.

The decision rule to keep: reversible and cheap to re-decide — satisfice against a pre-written threshold and move. Expensive or one-way — weights first, then scores, then find the flip point. Either way, the menu opens last.