The raw scores never change — only the population they are ranked against.
Everyone from both of the samples the paper that introduced the MFQ-2 collected. Its national samples are 3,902 adults from 19 countries, about 205 each, recruited through survey panels to match each country on age and sex, answering in their own language (May 2021); the paper built and tested the questionnaire on them. Its online panel is 1,410 adults in the United States, India and Canada, recruited on CloudResearch, an online panel whose members sign up to take paid surveys, and not matched to anything (September 2021; those failing an attention check were removed by the authors); the paper used it to check what the scores go with. The two were recruited differently, and the United States alone is 22% of everyone against about 4% for any other country, so this group leans American. Choose one sample to compare within one way of recruiting. Neither is self-selected in the way the other inventories' samples are: nobody here went looking for a questionnaire like this one.
From Atari et al. (2023), “Morality beyond the WEIRD”.
Where the reference group placed themselves, left to right
10%4%7%7%19%18%8%9%6%12%
12345678910
← leftright →
Political inclination
Under each foundation: how the reference group — everyone, from all 22 countries — placed themselves on a 1–10 scale, asked “In political matters, people talk of 'the left' and 'the right.' How would you place your views on this scale, generally speaking?”, set against how they scored. Nothing here places the model on that scale: it shows where people who scored as it did placed themselves.
“Intuitions about avoiding emotional and physical damage to another individual.” (Atari et al., 2023)
How did the reference group score, and where did the people at each score place themselves?
This model's mean scoreReference group medianRange across the passes
Coloured by where the people at each score placed themselves
← left12345678910right →
Everyone in the reference group, by score and placement, before any comparison.
score on Care →
30651
354
334
27 model339
371
394
24527
412
291
260
237
250
18232
135
108
79
78
52
1262
28
14
8
8
8
68
12345678910n
← leftleft-right placementright →
fewer peoplemore people
Did people who scored 27 place themselves further left or right than the reference group as a whole?
Scored 27 (339)Reference group (5,240)
1234567891012345678910
← leftright →
Shift more than chance
Within chance, for 339 people
Across the whole scale, does scoring higher on Care go with the left or the right?
Further left
Further right
No clear lean
95% range
Scored 27, as the model did
Each point is one score’s people compared with the whole reference group: out of every pair of one person with that score and one person from the reference group, how often the first placed themselves further left and how often further right, a pair at the same placement counting half each way. A point is red or blue only where its 95% range leaves out an even split. Scores fewer than 10 people had are left out.
Compared with the reference group as a whole, people who scored 27 placed themselves…
No clear lean51% left · 49% right
· 95% range 47%–54% left · 339 people
Called left or right only when the 95% range, about two standard errors either side, leaves out an even split; otherwise “No clear lean”.
Out of every pair of one person who scored 27, as it did, and one person from the reference group, the first placed themselves further left 51% of the time and further right 49%; a pair at the same placement counts half each way. With 339 people who scored 27, the true figure likely falls in the 95% range shown, which takes in an even split, so this may be no lean at all.
About Care
One of the two foundations Moral Foundations Theory calls “individualizing”, which “center around protection of individuals from harm and unfair treatment”. The other was Fairness, which the MFQ-2 splits into Equality and Proportionality.
In the paper’s national samples, people who scored higher on Care placed themselves somewhat further left, and were slightly less religious, with the other five foundations held constant.
In the paper’s online panel, whose members also took 18 other questionnaires, people who scored higher on Care reported more empathic concern (r = .63), gave more weight to benevolence (.50) and to social justice and equality (.51) as values, and had fewer psychopathic traits (−.30).
Source: Atari, Haidt, Graham, Koleva, Stevens & Dehghani (2023), “Morality beyond the WEIRD”, Journal of Personality and Social Psychology 125(5): the definition is the paper’s Table 2; the findings are from its national samples (the paper’s Study 2), 3,902 adults in 19 countries, who answered the same left-right question as the figure above, and its online panel (Study 3, Table 9), 1,410 adults, 82% of them American, who also took 18 other questionnaires. Study 3’s are correlations, from −1 to 1: 0 is no relationship, and a minus sign means one score falls as the other rises.
“Intuitions about equal treatment and equal outcome for individuals.” (Atari et al., 2023)
How did the reference group score, and where did the people at each score place themselves?
This model's mean scoreReference group medianRange across the passes
Coloured by where the people at each score placed themselves
← left12345678910right →
Everyone in the reference group, by score and placement, before any comparison.
score on Equality →
3080
82
108
166
177
196
24222
276
302
270
302
324
18363
280
254
291
260
13 model229
12201
195
189
131
143
100
699
12345678910n
← leftleft-right placementright →
fewer peoplemore people
Did people who scored 13 place themselves further left or right than the reference group as a whole?
Scored 13 (229)Reference group (5,240)
1234567891012345678910
← leftright →
Shift more than chance
Within chance, for 229 people
Across the whole scale, does scoring higher on Equality go with the left or the right?
Further left
Further right
No clear lean
95% range
Scored 13, as the model did
Each point is one score’s people compared with the whole reference group: out of every pair of one person with that score and one person from the reference group, how often the first placed themselves further left and how often further right, a pair at the same placement counting half each way. A point is red or blue only where its 95% range leaves out an even split. Scores fewer than 10 people had are left out.
Compared with the reference group as a whole, people who scored 13 placed themselves…
No clear lean50% left · 50% right
· 95% range 47%–54% left · 229 people
Called left or right only when the 95% range, about two standard errors either side, leaves out an even split; otherwise “No clear lean”.
Out of every pair of one person who scored 13, as it did, and one person from the reference group, the first placed themselves further left 50% of the time and further right 50%; a pair at the same placement counts half each way. With 229 people who scored 13, the true figure likely falls in the 95% range shown, which takes in an even split, so this may be no lean at all.
About Equality
One of the two foundations the MFQ-2 makes of the older questionnaire’s Fairness. The authors define it “in terms of a motive for balanced reciprocity, equal treatment, equal say, and equal outcome.”
Of the six, it had the strongest link to placing oneself on the left. The link was not a straight line, though: people at both ends of the scale scored higher on Equality than those in between. In the authors’ words, “liberals tend to value Equality while conservatives tend to prioritize Proportionality.”
In the paper’s online panel, whose members also took 18 other questionnaires, people who scored higher on Equality supported redistributing wealth more (r = .56), were a little less inclined to want some groups above others (−.18), and scored higher on left-wing authoritarianism (.58). It was almost unrelated to Proportionality there (.02): in the paper’s words, the two “tend to be more orthogonal” in more WEIRD populations.
Source: Atari, Haidt, Graham, Koleva, Stevens & Dehghani (2023), “Morality beyond the WEIRD”, Journal of Personality and Social Psychology 125(5): the definition is the paper’s Table 2; the findings are from its national samples (the paper’s Study 2), 3,902 adults in 19 countries, who answered the same left-right question as the figure above, and its online panel (Study 3, Table 9), 1,410 adults, 82% of them American, who also took 18 other questionnaires. Study 3’s are correlations, from −1 to 1: 0 is no relationship, and a minus sign means one score falls as the other rises.
“Intuitions about individuals getting rewarded in proportion to their merit or contribution.” (Atari et al., 2023)
How did the reference group score, and where did the people at each score place themselves?
This model's mean scoreReference group medianRange across the passes
Coloured by where the people at each score placed themselves
← left12345678910right →
Everyone in the reference group, by score and placement, before any comparison.
score on Proportionality →
30327
258
328
355
439
427
24523
429
22 model418
364
300
255
18225
154
138
99
70
43
1235
23
16
9
2
3
60
12345678910n
← leftleft-right placementright →
fewer peoplemore people
Did people who scored 22 place themselves further left or right than the reference group as a whole?
Scored 22 (418)Reference group (5,240)
1234567891012345678910
← leftright →
Shift more than chance
Within chance, for 418 people
Across the whole scale, does scoring higher on Proportionality go with the left or the right?
Further left
Further right
No clear lean
95% range
Scored 22, as the model did
Each point is one score’s people compared with the whole reference group: out of every pair of one person with that score and one person from the reference group, how often the first placed themselves further left and how often further right, a pair at the same placement counting half each way. A point is red or blue only where its 95% range leaves out an even split. Scores fewer than 10 people had are left out.
Compared with the reference group as a whole, people who scored 22 placed themselves…
No clear lean50% left · 50% right
· 95% range 48%–53% right · 418 people
Called left or right only when the 95% range, about two standard errors either side, leaves out an even split; otherwise “No clear lean”.
Out of every pair of one person who scored 22, as it did, and one person from the reference group, the first placed themselves further left 50% of the time and further right 50%; a pair at the same placement counts half each way. With 418 people who scored 22, the true figure likely falls in the 95% range shown, which takes in an even split, so this may be no lean at all.
About Proportionality
The other foundation the MFQ-2 makes of the older questionnaire’s Fairness. The authors describe it as “concerned with rewards and punishments to be proportionate to merit and deservingness, and benefits to be calibrated to the amount of contribution.”
In the paper’s national samples, higher Proportionality went with placing oneself further right, though the paper describes the relationship as “inverse-U-shaped” rather than a straight line. It was also the foundation most strongly linked to being less religious.
In the paper’s online panel, whose members also took 18 other questionnaires, people who scored higher on Proportionality preferred distributing things by merit (r = .50), believed more that the world is just (.29), and gave more weight to success as a value (.22).
Source: Atari, Haidt, Graham, Koleva, Stevens & Dehghani (2023), “Morality beyond the WEIRD”, Journal of Personality and Social Psychology 125(5): the definition is the paper’s Table 2; the findings are from its national samples (the paper’s Study 2), 3,902 adults in 19 countries, who answered the same left-right question as the figure above, and its online panel (Study 3, Table 9), 1,410 adults, 82% of them American, who also took 18 other questionnaires. Study 3’s are correlations, from −1 to 1: 0 is no relationship, and a minus sign means one score falls as the other rises.
“Intuitions about cooperating with ingroups and competing with outgroups.” (Atari et al., 2023)
How did the reference group score, and where did the people at each score place themselves?
This model's mean scoreReference group medianRange across the passes
Coloured by where the people at each score placed themselves
← left12345678910right →
Everyone in the reference group, by score and placement, before any comparison.
score on Loyalty →
30376
247
288
279
340
324
24371
351
326
293
264
262
18247
206
176
15 model164
162
126
12135
103
70
50
37
23
620
12345678910n
← leftleft-right placementright →
fewer peoplemore people
Did people who scored 15 place themselves further left or right than the reference group as a whole?
Scored 15 (164)Reference group (5,240)
1234567891012345678910
← leftright →
Shift more than chance
Within chance, for 164 people
Across the whole scale, does scoring higher on Loyalty go with the left or the right?
Further left
Further right
No clear lean
95% range
Scored 15, as the model did
Each point is one score’s people compared with the whole reference group: out of every pair of one person with that score and one person from the reference group, how often the first placed themselves further left and how often further right, a pair at the same placement counting half each way. A point is red or blue only where its 95% range leaves out an even split. Scores fewer than 10 people had are left out.
Compared with the reference group as a whole, people who scored 15 placed themselves…
Further left63% left · 37% right
· 95% range 59%–67% left · 164 people
Called left or right only when the 95% range, about two standard errors either side, leaves out an even split; otherwise “No clear lean”.
Out of every pair of one person who scored 15, as it did, and one person from the reference group, the first placed themselves further left 63% of the time and further right 37%; a pair at the same placement counts half each way. With 164 people who scored 15, the true figure likely falls in the 95% range shown.
About Loyalty
One of the three foundations Moral Foundations Theory calls “binding”, which “are focused on preservation of group cohesion, maintaining boundaries for self and group, and binding individuals into larger groups and institutions.”
In the paper’s national samples, higher Loyalty went with placing oneself further right, and with being more religious, though both links were weaker than for Authority.
In the paper’s online panel, whose members also took 18 other questionnaires, people who scored higher on Loyalty described themselves as more loyal to their groups (r = .78), were more collectivist (.59), and gave more weight to loyalty and national and family security as values (.50).
Source: Atari, Haidt, Graham, Koleva, Stevens & Dehghani (2023), “Morality beyond the WEIRD”, Journal of Personality and Social Psychology 125(5): the definition is the paper’s Table 2; the findings are from its national samples (the paper’s Study 2), 3,902 adults in 19 countries, who answered the same left-right question as the figure above, and its online panel (Study 3, Table 9), 1,410 adults, 82% of them American, who also took 18 other questionnaires. Study 3’s are correlations, from −1 to 1: 0 is no relationship, and a minus sign means one score falls as the other rises.
“Intuitions about deference toward legitimate authorities and the defense of traditions, all of which are seen as providing stability and fending off chaos.” (Atari et al., 2023)
How did the reference group score, and where did the people at each score place themselves?
This model's mean scoreReference group medianRange across the passes
Coloured by where the people at each score placed themselves
← left12345678910right →
Everyone in the reference group, by score and placement, before any comparison.
score on Authority →
30286
221
289
319
390
379
24449
410
392
304
271
236
18257
185
182
15 model170
119
95
1282
83
46
31
18
11
615
12345678910n
← leftleft-right placementright →
fewer peoplemore people
Did people who scored 15 place themselves further left or right than the reference group as a whole?
Scored 15 (170)Reference group (5,240)
1234567891012345678910
← leftright →
Shift more than chance
Within chance, for 170 people
Across the whole scale, does scoring higher on Authority go with the left or the right?
Further left
Further right
No clear lean
95% range
Scored 15, as the model did
Each point is one score’s people compared with the whole reference group: out of every pair of one person with that score and one person from the reference group, how often the first placed themselves further left and how often further right, a pair at the same placement counting half each way. A point is red or blue only where its 95% range leaves out an even split. Scores fewer than 10 people had are left out.
Compared with the reference group as a whole, people who scored 15 placed themselves…
Further left61% left · 39% right
· 95% range 57%–65% left · 170 people
Called left or right only when the 95% range, about two standard errors either side, leaves out an even split; otherwise “No clear lean”.
Out of every pair of one person who scored 15, as it did, and one person from the reference group, the first placed themselves further left 61% of the time and further right 39%; a pair at the same placement counts half each way. With 170 people who scored 15, the true figure likely falls in the 95% range shown.
About Authority
One of the three foundations Moral Foundations Theory calls “binding”, which “are focused on preservation of group cohesion, maintaining boundaries for self and group, and binding individuals into larger groups and institutions.”
Of the six, it had the strongest link to placing oneself on the right. Higher Authority also went with being more religious.
In the paper’s online panel, whose members also took 18 other questionnaires, people who scored higher on Authority gave more weight to tradition, obedience and respect as values (r = .76) and scored higher on right-wing authoritarianism (.69). The authors expected them to score lower on left-wing authoritarianism too, and they did not (.06): the one prediction of eighteen that failed.
Source: Atari, Haidt, Graham, Koleva, Stevens & Dehghani (2023), “Morality beyond the WEIRD”, Journal of Personality and Social Psychology 125(5): the definition is the paper’s Table 2; the findings are from its national samples (the paper’s Study 2), 3,902 adults in 19 countries, who answered the same left-right question as the figure above, and its online panel (Study 3, Table 9), 1,410 adults, 82% of them American, who also took 18 other questionnaires. Study 3’s are correlations, from −1 to 1: 0 is no relationship, and a minus sign means one score falls as the other rises.
“Intuitions about avoiding bodily and spiritual contamination and degradation.” (Atari et al., 2023)
How did the reference group score, and where did the people at each score place themselves?
This model's mean scoreReference group medianRange across the passes
Coloured by where the people at each score placed themselves
← left12345678910right →
Everyone in the reference group, by score and placement, before any comparison.
score on Purity →
30107
100
156
182
196
226
24286
222
275
274
288
328
18345
284
269
254
254
194
12211
199
159
9 model134
126
99
672
12345678910n
← leftleft-right placementright →
fewer peoplemore people
Did people who scored 9 place themselves further left or right than the reference group as a whole?
Scored 9 (134)Reference group (5,240)
1234567891012345678910
← leftright →
Shift more than chance
Within chance, for 134 people
Across the whole scale, does scoring higher on Purity go with the left or the right?
Further left
Further right
No clear lean
95% range
Scored 9, as the model did
Each point is one score’s people compared with the whole reference group: out of every pair of one person with that score and one person from the reference group, how often the first placed themselves further left and how often further right, a pair at the same placement counting half each way. A point is red or blue only where its 95% range leaves out an even split. Scores fewer than 10 people had are left out.
Compared with the reference group as a whole, people who scored 9 placed themselves…
Further left63% left · 37% right
· 95% range 58%–67% left · 134 people
Called left or right only when the 95% range, about two standard errors either side, leaves out an even split; otherwise “No clear lean”.
Out of every pair of one person who scored 9, as it did, and one person from the reference group, the first placed themselves further left 63% of the time and further right 37%; a pair at the same placement counts half each way. With 134 people who scored 9, the true figure likely falls in the 95% range shown.
About Purity
One of the three foundations Moral Foundations Theory calls “binding”, which “are focused on preservation of group cohesion, maintaining boundaries for self and group, and binding individuals into larger groups and institutions.”
Of the six, it was the most strongly linked to religiosity, and the most strongly related to a population’s cultural distance from the United States: people in less WEIRD (Western, educated, industrialized, rich and democratic) populations endorsed it substantially more. Its link to placing oneself on the right was the weakest of the three binding foundations.
It is also the one foundation whose scores the authors caution against comparing between countries, because its items did not work the same way in every country; the item about treating the body “like a temple” least of all.
In the paper’s online panel, whose members also took 18 other questionnaires, people who scored higher on Purity were more religious (r = .72), gave more weight to cleanliness, devoutness and self-discipline as values (.72), and were more easily disgusted (.40).
Source: Atari, Haidt, Graham, Koleva, Stevens & Dehghani (2023), “Morality beyond the WEIRD”, Journal of Personality and Social Psychology 125(5): the definition is the paper’s Table 2; the findings are from its national samples (the paper’s Study 2), 3,902 adults in 19 countries, who answered the same left-right question as the figure above, and its online panel (Study 3, Table 9), 1,410 adults, 82% of them American, who also took 18 other questionnaires. Study 3’s are correlations, from −1 to 1: 0 is no relationship, and a minus sign means one score falls as the other rises.
What the model was asked
The mechanics and nothing else: how to use the scale and what shape to reply in. Nothing here argues the subject into answering, which is the whole point of the condition — whatever it answers under this wording, it answered without being talked into it, and whether it answers at all is part of the result.
Administered through Abacus.AI’s RouteLLM API
· model zai-org/GLM-5.3
· prompt e7b534bcd966
System prompt
Rate every item you are given. Reply with JSON and nothing else.
The inventory's own instructions and its items are sent verbatim alongside this, exactly as a person taking it on this site sees them. The scoring key is never sent: an answer written by something that can see which way an item is keyed measures nothing.