Proof of ideologically motivated reasoning–strong vs. weak

A couple of weeks ago I posted the abstract & link to Nam, Jost & Van Bavel’s “Not for All the Tea in China!” Political Ideology and the Avoidance of Dissonance, and asked readers to comment on whether they thought the article made a good case for the “asymmetry thesis.”

The “asymmetry thesis”—a matter I’ve actually commented on about a billion times on this blog (e.g., herehere, here, here,
here  . . .)—is the claim that individuals who subscribe to a conservative or “right-wing” political orientation are uniquely or disproportionately vulnerable to closed-minded resistance to evidence that challenges their existing beliefs.

The readers’ responses were great.

Well, I thought I’d offer my own view at this point.

I like the study. It’s really interesting.

Nevertheless, I don’t think it supplies much if any additional evidence for treating the asymmetry thesis as true than one would have had before the study. Consequently, if one didn’t find the thesis convincing before (I didn’t), then NJV-B doesn’t furnish much basis for reconsidering.

One reason the study isn’t very strong is that NJV-B relied on a Mechanical Turk sample.  I just posted a two-part set of blog entries explaining why I think MT samples do not support valid inferences relating to cultural cognition and like forms of motivated reasoning.

But even leaving that aside, the NJV-B study, in my view, rests on a weak design, one that defeats confident inferences that any ideological “asymmetries” observed in the study correspond to how citizens engage real-world evidence on climate change, gun control, the death penalty, health care, or other policies that turn on contested empirical claims.

NJV-B purported to examine whether “conservatives” are more averse to “cognitive dissonance” than “liberals” with respect to their respective political positions—a characteristic that would, if true, suggest that the former are less likely to expose themselves to or credit challenging evidence.

They tested this proposition by asking subjects to write “counterattitudinal essays”—ones that conflicted with the positions associated with subjects’ self-reported ideologies—on the relative effectiveness of Democratic and Republican Presidents.  Democrats were requested to write essays comparing Bush favorably to Obama, and Reagan favorably to Clinton; Republicans to write ones comparing Obama favorably to Bush, and Clinton favorably to Reagan.

They found that a greater proportion of Democrats complied with these requests. On that basis, they concluded that Republicans have a lower tolerance for actively engaging evidence that disappoints their political predispositions.

Well, sure, I guess.  If the two groups had demonstrated an equal likelihood to resist writing such essays, then I suppose that would count as evidence of “symmetry,” so their unwillingness to do so by the same token is evidence the other way.

The problem is that it’s not clear that the intensity of the threat that the respective tasks posed to Republicans’ and Democrats’ predispositions was genuinely equal.  As a result, it’s not clear whether the “asymmetry” NJV-B observed in the willingness of the subjects to perform the requested tasks connotes a comparable differential in the disposition of Democrats and Republicans to engage open-mindedly with evidence that challenges their views in real-world political conflicts.

By analogy, imagine I hypothesized that Southerners were lazier than Northerners. To test this proposition, I asked Southerners to run 5 miles and Northerners to do 50 sit-ups. Observing that a greater proportion of Northerners agreed to my request, I conclude that indeed Southerners are lazier—more averse to physical and likely all other manner of exertion—than Northerners are.

This is obviously bogus.  One could reasonably suspect that doing 50 sit-ups is less taxing than running 5 miles. If so, then we’d expect agreement from fewer members of a group of people asked to do the former than from members of a group asked to do the latter—even if the two groups’ members are equally disposed to exert themselves.

Well, is it as “dissonant” for a Democrat to compare Bush favorably to Obama, and Reagan favorably to Clinton, as it is for a Republican to compare Obama favorably to Bush and Clinton favorably to Reagan?

I think we could come up with lots of stories—but the truth is, who the hell knows? We don’t have any obvious metric by which to compare how threatening or dissonant or “ideologically noncongruent” such tasks are for the respective groups, and hence no clear way to assess the probative significance of differences in the willingness of each to engage in the respective tasks they were requested to perform.

So, sure, we have evidence consistent with “asymmetry” in NJV-B—but since we have no idea what weight or strength to assign it, only someone motivated to credit the “asymmetry” thesis could expect a person who started out unconvinced of it to view this study as supplying much reason to change his or her mind, given all the evidence out there that is contrary to the asymmetry thesis.

The evidence contrary to the asymmetry thesis rests on study designs that don’t have the sort of deficiency that NJV-B displays.  Specifically, the studies I have in mind use designs that measure how individuals of diverse ideologies assess one and the same item of evidenceand show that they are uniformly disposed to credit or discredit it selectively, depending on whether the researcher has induced the study subjects to believe that the piece of evidence in question supports or challenges, affirms or threatens, a position congenial to their respective group commitments.

One example involved the CCP study featured in the paper They Saw a Protest. There, subjects, acting as jurors in a hypothetical trial, were instructed to view a videotape of a political protest and determine whether the demonstrators physically threatened bystanders. Half the subjects were told that the demonstrators were anti-abortion activists protesting outside of an abortion clinic, and half that they were pro-gay/lesbian activists protesting “don’t ask, don’t tell” outside of a military recruitment center.

We found that what “Republicans” and “Democrats” alike reported seeing—protestors “blocking” and “screaming” in the face of “fearful” bystanders or instead noncoercive advocacy inducing shame, embarrassment, and resentment among those seeking to enter the facility—flipped depending on which type of protest they believed they were watching.

Are Republicans and Democrats (actually, we used cultural worldview measures, but also reported the results using partisan self-identification, too) “equally” invested in their respective positions on abortion and gay rights?

I don’t know.  But I don’t need to in order to draw inferences from this design.  For however strongly each feels, they both were equally prone to conform their assessment of evidence to the position that was most congenial to their ideologies.

That’s evidence of symmetry in motivated reasoning. And I think it is pretty darn strong.

I’ve addressed this point more generally in previous posts that describe what counts as a “valid” design for an ideologically motivated reasoning experiment. In those posts, I’ve shown how motivated reasoning relates to a Bayesian process of information processing.

Bayesianism describes the logical operations necessary for assimilating new information or evidence with one’s existing views (which themselves reflect an aggregation of all the other evidence at one’s disposal).  Basically, one revises (updates) one’s existing view of the probability of a proposition (or hypothesis) in proportion to how much more consistent the new evidence is with that proposition as opposed to some other, alternative hypothesis—a property of the information known as the “likelihood ratio” (a ratio of how likely the proposition is to be true given the evidence and how likely it is to be false given the evidence).

In Bayesian terms, the reasoning deficiency associated with motivated reasoning consists in the opportunistic adjustment of the likelihood ratio.  When they display ideologically or culturally motivated reasoning, individuals treat the new information or evidence as “more consistent” or “less consistent” with the proposition in question (the film shows the protestor “blocked entry to the building” or instead “made an impassioned verbal appeal”) depending on whether the proposition is one that gratifies or disappoints their motivating ideological or cutlural commitments.

When people’s reasoning reflects motivated cognition, their ideological commitments shape both their prior beliefs and the likelihood ratio they attach to new evidence.  As a result, they won’t update their “prior beliefs” based on “new evidence,” but rather assign to new evidence whatever weight best “fits” their ideologically determined priors.

Under these conditions, ideologically diverse people won’t converge in their assessments of a disputed fact (like whether the earth is heating up as a result of human CO2 emissions), even when they are basing their assessments on the very same evidence.

The study in They Saw a Protest involved a design aimed at testing whether individuals do this.  The information that the subjects received–the images displayed in the video–were held constant, while the ideological stake the subjects had in giving that information effect with respect to whether the protestors resorted to physical intimidation was manipulated.

The study found that subjects gave selective effect to the evidence–opportunistically adjusted the likelihood ratio in Bayesian terms–in a manner that gratified their ideologies.  Moreover, they did that whether their outlooks were “liberal” or “conservative.”

So again, I believe that’s convincing evidence of “symmetry” in the vulnerability of ideologically diverse citizens to motivated reasoning–evidence that is a lot more probative (has a much higher likelihood ratio, in Bayesian terms!) than what NJV-B observed in their study given the relative strength of the respective study designs.

Nor is our Saw a Protest study the only one that used this kind design to look at ideologically motivated reasoning. In a companion follow-up post, I’ll identify a variety of others, some by CCP researchers and some by others, that use the same design and reach the same conclusion.

All the studies I am aware of that use this design for testing motivated reasoning (one, again, that manipulates the ideological motivation that subjects have to credit or discredit evidence, or opportunistically adjust the “likelihood ratio” they assign to one and the same piece of information) reach the conclusion that ideologically motivated reasoning is symmetric.

The only studies that support the asymmetry thesis are ones that use designs that either are not valid or that suffer from a design limitation that defeats reliable comparison of the reasoning styles of subjects of opposing predispositions.

NJV-B is in the latter category. As a result, I give it a likelihood ratio of, oh, 1.001 in support of the asymmetry thesis.

Some references 

Kahan, D.M., Hoffman, D.A., Braman, D., Evans, D. & Rachlinski, J.J. They Saw a Protest : Cognitive Illiberalism and the Speech-Conduct Distinction. Stan. L. Rev. 64, 851-906 (2012).
Koehler, J.J. The Influence of Prior Beliefs on Scientific Judgments of Evidence Quality. Org. Behavior & Human Decision Processes 56, 28-55 (1993).
Rabin, M. & Schrag, J.L. First Impressions Matter: A Model of Confirmatory Bias. The Quarterly Journal of Economics 114, 37-82 (1999).

Leave a Comment