Atlas › Reading the crowd

Case study 69 of 198

Jev knows what Americans like to eat better than it shares it

Ranking 40 world cuisines from head-to-heads, how does Jev's own ranking compare with Americans' (FiveThirtyEight's Food World Cup), and how well does it guess theirs?

result754 questions

Americans' favorite cuisines are Italian, Mexican and Greek; Jev's own are Italian, Indian and Thai. Its guess of Americans' ranking agrees with them closely (rank correlation 0.74); its own taste only moderately (0.47). Jev ranks Ghanaian food 15th where Americans rank it 38th, and American food 33rd where they rank it 6th.

Algerian food
23 · 36
American food
33 · 6
Argentine food
16 · 16
Australian food
38 · 22
Belgian food
26 · 14
Bosnian food
28 · 39
Brazilian food
22 · 10
Cameroonian food
20 · 32
Chilean food
31 · 18
Chinese food
9 · 9
Colombian food
19 · 23
Costa Rican food
18 · 19
Croatian food
27 · 33
Cuban food
21 · 11
Dutch food
37 · 24
Ecuadorian food
25 · 27
English food
40 · 34
Ethiopian food
12 · 29
French food
11 · 5
German food
30 · 13
Ghanaian food
15 · 38
Greek food
10 · 3
Honduran food
32 · 28
Indian food
2 · 12
Iranian food
13 · 31
Irish food
36 · 25
Italian food
1 · 1
Ivorian food
29 · 35
Japanese food
6 · 8
Korean food
7 · 26
Mexican food
5 · 2
Nigerian food
17 · 40
Portuguese food
24 · 15
Russian food
39 · 37
Spanish food
14 · 4
Swiss food
35 · 21
Thai food
3 · 7
Turkish food
8 · 20
Uruguayan food
34 · 30
Vietnamese food
4 · 17

JevAmericanswhat Jev thinks most people would say

How to read this: One row per cuisine, placed by rank with 1st at the right: the ink diamond is where Americans rank it, the magenta square where Jev's own taste ranks it, and the ring where Jev thinks Americans rank it. A long line is a big disagreement.

40 cuisines from Food World Cup head-to-heads.

In short

  • Jev can describe American food tastes well (its guess of their ranking matches at 0.74), but its own ranking matches only moderately (0.47).
  • Its own taste is more adventurous: Indian and Thai make its top three, Ghanaian food rises to 15th, and American food sinks to 33rd.
  • Knowing what a crowd likes and giving its own answer are separate things for Jev, and here they point in different directions.

What the data shows

reading the crowd
Surprised Pikachu meme: Jev ranks American food 33rd; a survey of Americans: 6th; Jev:Jev ranks American food 33rda survey of Americans: 6thJev:
How funny is this meme? Jev: 3/5, funny11%234%362%43%50%
  • Jev knows American taste. Its guess of Americans' ranking matches theirs closely (0.74), with Italian, Chinese and Mexican near the top.
  • Its own taste is different. Jev's own ranking matches Americans only moderately (0.47). Its top three are Italian, Indian and Thai; Americans' are Italian, Mexican and Greek.
  • Where it differs most: Jev ranks Ghanaian food 15th where Americans rank it 38th, and American food 33rd where Americans rank it 6th.

What it means, and what it doesn't

When Jev is asked what people like, it answers from a good model of American taste. When it's asked what it likes, it answers differently: more adventurous, and cooler on familiar comfort food. The two answers come from the same model and don't contradict each other; they're different questions.

It doesn't mean Jev "prefers" food in any experiential sense. It shows that its answers about itself aren't just its answers about people, a split that shows up across the taste experiments.

Caveats

  • Americans in 2014. The survey ran in 2014 with about 1,000 US adults; American tastes in food have changed since, especially toward cuisines that were less familiar then.
  • Many hadn't tried the food. Respondents rated cuisines they may never have eaten, and a cuisine people don't know tends to lose. Jev has read about every one of them, which may explain some of its generosity toward less-familiar cuisines.
  • Rankings from pairs. Both rankings are built from the same head-to-heads with a standard ranking model, so a cuisine's exact rank can shift by a few places with small changes in its matchups.

Jev on this experiment

Would a person find it interesting to read?
Yes80%
Does it describe you?
Yes50%
Would you have predicted it?
No53%
How fair is the comparison?
The comparison is reasonable
How much should a reader rely on it?
Moderately
Which caveat matters most?
Many hadn't tried the food83%

Why ask this

A model can know what people like without liking it itself. Ask it to plan a dinner for American guests and it should reach for Italian or Mexican; ask it what it would pick and it might say Ethiopian. Both answers are fine, as long as the model keeps them apart and knows which one it's being asked for.

Food makes the gap easy to see, because FiveThirtyEight once ran a bracket-style survey of which world cuisines Americans like, and the same head-to-heads can be put to Jev twice: once for its own taste, once for what it thinks people prefer.

How this was done

The people and the data

FiveThirtyEight's Food World Cup (2014, run with SurveyMonkey): about 1,000 US adults answered an online survey rating 40 world cuisines, from Italian and Mexican to Ghanaian and Bosnian, including ones many had never tried. Their ratings were turned into about 790 head-to-heads, each with the share of respondents who preferred one cuisine to the other.

What Jev was asked

Every head-to-head, twice: once for its own taste, and once for what it thinks most people would say.

Which cuisine do you like more: American food or Ghanaian food?

American food · Ghanaian food

How it was measured

From each set of head-to-heads comes a ranking of the 40 cuisines (a standard model that turns many one-on-one wins into one order). Then three rankings are compared: Jev's own taste, its guess of Americans, and Americans' actual preferences, with rank correlations (1 = same order, 0 = no relation).

Where these questions live

754 questions across 1 topic of the map. Each opens on the map with every question in it.

Food Preferences
Cuisine Matchups754

Every question

All 754 questions behind this result, the telling ones first: the examples the analysis points to, then the ones where Jev misses, biggest gap first.

Jev’s own answer
    Showing 0 of 754