Very counterintuitively, extreme political actions seem to be associated with people who start out ambivalent about issues in the context of high levels of polarization.
Political extremism varies across people and contexts, but which beliefs will a person support through extreme actions?
We propose that ambivalent attitudes, despite reducing normative political actions like voting, increase support for extreme political actions.
We demonstrate this hypothesized reversal using dozens of measures across six studies (N = 13,055). The effect was robust to relevant covariates and numerous methodological variations and was magnified when people’s attitudinal or ideological positions were more polarized.
It appears to occur because being conflicted about political issues can feel psychologically uncomfortable, making extreme actions more appealing. Notably, this emerged when people thought ambivalence was justified, whereas leading them to consider ambivalence unjustified suppressed the effect, suggesting that ambivalent people are coping with but not necessarily trying to reduce their ambivalence. These results highlight the interplay of affective and cognitive influences in extreme behavior, showing that beliefs people feel justifiably conflicted about can promote extremism.
Joseph J. Siev and Richard E. Petty, "Ambivalent attitudes promote support for extreme political actions" 10(24) Science Advances (June 12, 2024) (open access). Hat tip to Guy in the comments.
6 comments:
Damn it! I wanted you trash the paper. Now I'll have to read it completely and ponder it's meaning.
It isn't obviously wrong. It even makes a certain amount of twisted sense. And, it has lots of data to back it up.
If there is a problem with it, it is a lot less obvious than most flawed behavioral/political science papers.
It isn't quite as intuitive, once you wrap your head around it, as the notion that most terrorists have similar psychological profiles to war heroes. But, the most common sense way to think about it is as a variant of the idea of the fanaticism of recent converts, which is a fairly familiar idea.
Another adjacent phenomena is that Islamic extremism in my lifetime was to a great extent a product of people becoming literate enough to read the Quran directly without the intervention of Imams rather than having it filter through people who were theologically trained in the nuances and doctrines that tamed the raw source text.
Likewise, the Second Great Awakening in the U.S. (which is the root of most Evangelical Christianity in the U.S.) grew extreme for basically the same reason. Self-proclaimed preachers started to teach the Bible based upon what they read themselves with no other context and without the nuances and doctrines that tamed it that clergy traditionally received in a formal theological education.
Not the same, but similar in mechanism and character.
And now for something completely different...
A normal distribution that is bounded on the left can be truncated. Or possibly the left and right side have different variances. I've done a superficial search for how this might affect the PDF (from a Fisherian standpoint) and have not found the right taste in approach, which might be at the "is this the right distribution?" level. Do you have any recommendations on an undergrad level review?
A normal distribution (physicists like to call it a Gaussian distribution, but social scientists snub Gauss because he wasn't one of them) is symmetrical by definition. It has a single variance statistic for the entire distribution even if it is truncated.
A Poisson distribution assigns probabilities to discrete values rather than continuous ones (usually integers) and is usually truncated at zero on the left. See https://en.wikipedia.org/wiki/Poisson_distribution But it also has a single set of properties for the entire distribution.
I took statistics as an upper division calculus/analysis based math course as an undergraduate with hard copy books that I still have, and often use my father's statistics book, possibly from his undergraduate or graduate school days, as well (one he used, not one he wrote) because it is better written.
I don't have a good go to reference online that I can point you to, and I am not thinking of a good distribution to use right off. The Student's t-distribution is a generalized normal distribution which is good for data sets for long tails. https://en.wikipedia.org/wiki/Student%27s_t-distribution
Other distributions out there are the F-distribution, https://en.wikipedia.org/wiki/F-distribution, the Chi-squared which is a non-parametric goodness of fit test https://en.wikipedia.org/wiki/Chi-squared_distribution, and the Gamma distribution https://en.wikipedia.org/wiki/Gamma_distribution.
There is a good list of commonly used probability distribution functions (not to be confused with parton distribution functions and portable document format) at https://en.wikipedia.org/wiki/Probability_distribution#Common_probability_distributions_and_their_applications
If I knew what you were trying to model I could give you some less vague suggestions.
I wanted to do an honors project as a math major on the properties of a normal distribution crossed with linear positive quantities and power laws distributions, to answer questions like what level of income inequality would we expect if income is normally distributed. But my proposal was not accepted.
Post a Comment