Two Résumés, Same Percentile — and You'll Still Pick the Wrong One
July 7, 2026 · Dana Cole, Thinking, Fast and Slow~6 min read
Here, quick, before you overthink it. Two people applied for the same job. One came 3rd out of 12 in their assessment. The other came 6th out of 24. Same role, same test, and you can hire exactly one of them. Which way does your gut lean? If you're anything like me — and, it turns out, like most of the people in a new set of experiments — you reached for the 3rd. Third! Podium! Whereas sixth is, well, sixth. Except here's the thing that ought to make you put your coffee down: 3rd of 12 and 6th of 24 are the same standing. Both sit at the 25th percentile. Bottom quarter beaten, top quarter untouched, identical. And most people, offered that exact choice, quietly pick the one that's — mathematically — no better at all.
The number that shouts, and the one that whispers
This comes from new research by Christopher Silverman, working with Uri Barnea at Bocconi and Alice Moon at Georgetown, published in the journal Organizational Behavior and Human Decision Processes and written up by phys.org at the start of July. The pattern held stubbornly: given two options with the same relative standing, people leant towards the one carrying the smaller absolute rank. Third over sixth. Second over fourth. The lower number just feels better, and the length of the list it sits in — the bit that actually gives the number its meaning — barely registers.
Why does it slide past us? Because rankings pull your eye straight to the top. You read "3rd" and, without deciding to, you count upwards: how many are ahead of this person? Two. Only two. What you almost never do is look down the list and count the other way. The 6th-of-24 candidate cleared eighteen people to get there. The 3rd-of-12 cleared nine. So the "worse"-looking one actually beat twice as many rivals — and yet, as the researchers put it, the nine that sixth-of-24 has a leg up on are just not psychologically loud enough. The two who finished ahead shout. The eighteen who finished behind say nothing.
Kahneman saw this engine years ago
If this has the ring of something familiar, it's because Daniel Kahneman laid out the machinery in Thinking, Fast and Slow. The real question in front of you is a fiddly one — what proportion of the field did this person beat? — and your fast, intuitive mind can't be bothered with proportions, so it swaps in an easier question it can answer instantly: how small is the top number? Kahneman called that quiet swap substitution, and the maddening part is you never feel it happen. It feels like you answered the question you were asked.
Two of his other ideas ride along. That rank number lands first and drags your estimate towards it — plain old anchoring. And the denominator, the "out of 24", never gets vivid enough to fight back, so it may as well not exist. Kahneman's phrase for that was what you see is all there is: the numerator is right there, glowing; the list length is a shadow behind it; and we judge, confidently, on the glowing bit alone. It's the same family of blunder that makes "1 in 100" feel scarier than "10 in 1,000" — the denominator quietly goes missing.
You didn't answer the question you were asked
"What percentile is this?" is hard, so your mind quietly swaps in "how small is the top number?" — and answers that instead, without ever telling you it changed the question.
It doesn't stay in the lab, and it doesn't fold under pressure
You might hope this is one of those cute effects that evaporates the moment anything's at stake. It isn't. The same tilt showed up across restaurants, hiking trails, historic homes, beaches, employee bonuses — and, more soberingly, it survived real consequences. Managers sizing up actual job candidates picked the shorter-list one more often. People buying a pen with their own money reached for the shorter-list product. So this isn't a quirk of bored undergraduates ticking boxes; it's the shape your judgement takes when a ranking is put in front of you and something real hangs on it.
Which should give anyone who ranks things for a living a small chill. A product that came 2nd of 8 can read as stronger than one that came 4th of 16, even when they're equally good for their category. Line up two internal candidates from teams of different sizes and the presentation — not the talent — can tip who gets promoted. Nobody has to be acting in bad faith. The scoreboard is doing the misleading all on its own.
What to actually do when a ranking lands in front of you
The good news is that this one bends. When people were nudged to think about how many options a thing had beaten, or shown the standings as percentiles instead of raw ranks, the short-list pull weakened. That's your whole toolkit, really — three small habits for reading any ranking that matters:
First, always ask out of how many. A rank with no denominator is a sentence missing half its words; refuse to feel anything about "3rd" until you know 3rd of what. Second, convert it into a percentile before you let yourself have an opinion — 3rd of 12 and 6th of 24 both land at the 25th, and once you've said that number out loud, the spell tends to break. Third, count the other direction: not just who's ahead, but how many did it actually beat? Nine, or eighteen? That's the figure the ranking works hard to hide, and it's usually the one you wanted all along. None of this makes you immune — I did the maths above and my gut still wants the 3rd. But now, at least, when it tugs, you'll know exactly which corner it's tugging from.
Framing drawn from Daniel Kahneman, Thinking, Fast and Slow (substitution, anchoring, and the "what you see is all there is" / denominator-neglect family). The findings come from new research by Christopher Silverman with Uri Barnea (Bocconi) and Alice Moon (Georgetown), published in Organizational Behavior and Human Decision Processes and reported by phys.org on July 2, 2026. This is a popular-science reading, not professional advice — and it rests on a set of preference experiments, so treat it as a sharp result about how we read rankings, not a verdict on anyone's worth; correlation is not causation.
New research by Silverman with Uri Barnea (Bocconi) and Alice Moon (Georgetown), in Organizational Behavior and Human Decision Processes (reported by phys.org, July 2, 2026): a candidate ranked 3rd of 12 and one ranked 6th of 24 are both at the 25th percentile, yet people repeatedly prefer the shorter list — even though 6th of 24 beat 18 rivals against 3rd of 12's 9. Framework: Daniel Kahneman, Thinking, Fast and Slow (denominator neglect, anchoring, WYSIATI). Popular-science reading, not professional advice; correlation is not causation.
枠組はダニエル・カーネマン『ファスト&スロー』から(「代替」「アンカリング」、そして「見えるものがすべて」/分母の無視という一族のバイアス)。本文の発見は、Christopher Silverman が、ボッコーニ大学の Uri Barnea、ジョージタウン大学の Alice Moon と行った新しい研究による。『Organizational Behavior and Human Decision Processes』に掲載、phys.org が 2026 年 7 月 2 日に報じた。これは科学解説であって専門的な助言ではありません。しかも一連の選好実験にもとづくので、誰かの価値を決める判定ではなく、「わたしたちがランキングをどう読むか」についての鋭い結果として受け取ってください——相関は因果ではありません。
シルヴァーマンが、ボッコーニ大学の Uri Barnea、ジョージタウン大学の Alice Moon と行った新研究(『Organizational Behavior and Human Decision Processes』掲載、phys.org 報道、2026 年 7 月 2 日):12 人中 3 位と 24 人中 6 位はどちらも 25 パーセンタイルなのに、人は繰り返し短いリストを選んだ——24 人中 6 位は 18 人に、12 人中 3 位は 9 人に勝っているのに。枠組:ダニエル・カーネマン『ファスト&スロー』(分母の無視、アンカリング、見えるものがすべて)。科学解説であり専門的助言ではない。相関は因果ではない。