Tuesday, May 28, 2013

GSS on Haidt's Moral Foundations theory

In The Righteous Mind, Jonathan Haidt describes a list of initially five, which he subsequently expands to six, moral foundations. In this theory of morality, there are six major dimensions of morality on which humans operate. Wikipedia provides a brief description of them here. Haidt's research shows that liberals in the US put more emphasis on the care/harm and liberty/oppression dimensions than conservatives do, while American conservatives give a bit more weight to the fairness/cheating foundation and considerably more focus to the loyalty/betrayal, authority/subversion, and sanctity/degradation dimensions than liberals do.

As it so often does, the GSS allows us to put this theory to a general test by tapping data that was not specifically designed to measure what Haidt and company are after but that, conveniently, still manages to provide another measurement.

I've selected six items, each one approximating one of the six foundations, and compared average liberal and average conservative responses to them in standard deviations. Positive (negative) values indicate that conservatives care more (less) about assuring the positive expression of the foundation is condoned and/or the negative expression of the foundation is censured than liberals do.

The care/harm question asks respondents if they believe the government should help pay for medical expenses for those who cannot afford them; the fairness/cheating question asks if it is wrong for a person to hide income to avoid paying taxes; the liberty/oppression question asks if personal freedom is more important than marriage; the authority/subversion question asks if a cop is ever justified in hitting a citizen; the sanctity/degradation question asks if gay sex is immoral; and the loyalty/betrayal question asks if a person should support his country even if when it is acting in the wrong:


The GSS results largely support Haidt's divisions, with liberals putting greater emphasis on liberty/oppression and especially care/harm, just as he found to be the case. Conservatives, on the other hand, are more concerned with fairness/cheating, authority/subversion, sanctity/degradation, and loyalty/subversion than liberals are. With the exception of the authority/subversion item, for which moderates and liberals are virtually indistinguishable, moderates fall in between liberals and conservatives on every foundation.

The gap between liberals and conservatives on fairness/cheating is larger than expected in the GSS relative to Haidt's findings on authority/subversion and loyalty/betrayal, and the sanctity/degradation gap is probably even wider than Haidt would expect it to be, but these questions are by no means perfectly calibrated to replicate his approach--the fact that they follow the overall pattern described by the moral foundations theory suggests it has descriptive value. Whether it is especially useful or overly cute and stylized is perhaps a separate issue, but that's not the point of this post.

Haidt has a simple, free site for those interested to gauge their own moral foundations within the framework of his theory, though it combines the fairness/cheating and liberty/oppression aspects into a single foundation on which liberals place greater importance.

My results (in green):

Mostly in line with conservatives on four of the five, my only deviation being on the sanctity/degradation foundation (which is, confusingly, referred to as "purity" in the online survey). Engaging in a little self-reflection, my results capture what sometimes borders on indifference towards the plight of others and also what they choose to do to themselves, and the high expectations of loyalty from friends, family, and colleagues. The purple would've suited me nicely, I think!



James A. Donald said...

Care/harm: We should avoid capriciously and unjustly convicting men of assault or rape merely on a woman's say so.

Liberty/Oppression: The right to keep and bear arms.

Authority/Subversion: Hate speech

Sanctity/Degradation: Eating wild animals, trashing the wilderness in places so distant that it is unlikely anyone else will encounter the trash before it gets buried by vegetation or worn away by the weather.

Loyalty/Betrayal: Should Harvard cancel Richwine's PhD

Audacious Epigone said...


Loyalty/betrayal still 'favors' conservatives--the left cares less for Harvard's loyalty to its former student than the right does.

Anonymous said...

I'm pretty dubious of Haidt's theory.

One thing I really would like to see, is if Haidt gathered a set of these survey results and applied principal components analysis or clustering algorithms to them, like we have done with ethnic genetic variation to validate our models of race.

This would help break down if principal variation in these survey questions is actually along a liberal-conservative axis, or something else.

It would also help distinguish more correlated from less correlated dimensions - for instance, it seems intuitive to me that loyalty clusters strongly with authority, because what is "authority" but loyalty to the government and powerful people? PCA / factor analysis could test this.

Beyond that, I think the theory is sharply limited to their questions - for instance, i'm sure liberals have a bunch of "sanctity" type values which are all about behaving properly (no matter the harm or unfairness), just that these are very SWPLish and don't actually appear on Haidt's method.

Anonymous said...

Beyond that, I think the theory is sharply limited to their questions - for instance, i'm sure liberals have a bunch of "sanctity" type values which are all about behaving properly (no matter the harm or unfairness), just that these are very SWPLish and don't actually appear on Haidt's method.

Exactly. I mean, baby killing, is about caring? seriously? Killing your own has to be about as disloyal as one can get.

So, yes, just the selection of the questions necessarily biases such an analysis to a degree that it is either useless or worse.

staffanspersonalityblog said...

I think the theory makes sense but it's perhaps skewed to divide liberals and conservatives.

Donald: Convicting people unjustly, isn't that just too much Harm and too little Fairness? Possibly because Harm becomes salient in the case of rape.