These were questions that I posed in a workshop I gave last Thurs. at Duke University in the political science department. I’ll give my (provisional, as always!) answers after “briefly” outlining the presentation (as I remember it at least). Slides here.
1. What is ideologically motivated reasoning?
It’s useful to start with a simple Bayesian model of information processing—not b/c it is necessarily either descriptively accurate (I’m sure it isn’t!) or normatively desirable (actually, I don’t get why it wouldn’t be, but seriously, I don’t want to get into that!) but b/c it supplies a heuristic benchmark in relation to which we can identify what is distinctive about any asserted cognitive dynamic.
Consider “confirmation bias” (CB). In a simple Bayesian model, when an individual is exposed to new information or evidence relating to some factual proposition (say, that global warming is occurring; or that allowing concealed click me!possession of firearms decreases violent crime), she revises (“updates”) her prior estimation of the probability of that proposition in proportion to how much more consistent the new information is with that proposition being true than with it being false (“the likelihood ratio” of the new evidence). Her reasoning displays CB when instead of revising her prior estimate based on the weight of the evidence so understood, she selectively searches out and assigns weight to the evidence based on its consistency with her prior estimation. (In that case, the “likelihood ratio” is endogenous to her “priors.”) If she does this, she’ll get stuck on an inaccurate estimation of the probability of the proposition despite being exposed to evidence that the estimate is wrong.
Motivated reasoning (MR) (at least as I prefer to think of it) refers to a tendency to engage information in a manner that promotes some goal or interest extrinsic to forming accurate beliefs. Thus, one searches out and selectively credits evidence based on the congeniality of it to that extrinsic goal or interest. Relative to the Bayesian model, then, we can see that goal or interest—rather than criteria related to accuracy of belief—as determining the “weight” (or likelihood ratio) to be assigned to new evidence related to some proposition.
MR might often look like CB. Individuals displaying MR will tend to form beliefs congenial to the extrinsic or motivating goal in question, and thereafter selectively seek out and credit information consistent with that goal. Because the motivating goal is determining both their priors and their information processing, it will appear as if they are assigning weight to information based on its consistency with their priors. But the relationship is in fact spurious (priors and likelihood ratio are not genuinely endogenous to one another).
“Ideologically motivated reasoning” (IMR), then, is simply MR in which some ideological disposition (say, “conservativism” or “liberalism”) supplies the motivating goal or interest extrinsic to formation of accurate beliefs. Relative to a Bayesian model, then, individuals will search out information and selectively credit it conditional on its congeniality to their ideological dispositions. They will appear to be engaged in “confirmation bias” in favor of their ideological commitments. They will be divided on various factual propositions—because their motivating dispositions, their ideologies, are heterogeneous. And they will resist updating beliefs despite the availability of accurate information that ought to result in the convergence of their respective beliefs.
In other words, they will be persistently polarized on the status of policy relevant facts.
2. What is the cultural cognition of risk?
The cultural cognition of risk (CCR) is a form of motivated reasoning. It posits that individuals hold diverse predispositions with respect to risks and like facts. Those predispositions—which can be characterized with reference to Mary Douglas’s “group grid” framework—motivate them to seek out and selectively credit information consistently with those predispositions. Thus, despite the availability of compelling scientific information, they end up in a state of persistent cultural polarization with respect to those facts.
The study of CCR is dedicated primarily to identifying the discrete psychological mechanisms through which this form of MR operates. These include “culturally biased information search and assimilation”; “the cultural credibility heuristic”; “cultural identity affirmation”; and the “cultural availability heuristic.”
These mechanisms do not result in confirmation bias per se. CCR, as a species of MR, describes the influences that connect information processing to an extrinsic motivating goal or interest. Often—maybe usually even—those influences will conform information processing to inferences consistent with a person’s priors, which will also reflect his or her motivating cultural predisposition. But CCR makes it possible to understand how individuals might be motivated to assess information about risk in a directionally biased fashion even when they have no meaningful priors (b/c, say, the risk in question is a novel one, like nanotechnology) or in a manner contrary to their priors (b/c, say, the information, while contrary to an existing risk perception, is presented in an identity-affirming manner).
Recent research has focused on whether CCR is a form of heuristic-driven or “system 1” reasoning. The CCP Nature Climate Science study suggests that the answer is no. The measures of science comprehension in that study are associated with use of systematic or analytic “system 2” information processing. And the study found that as science comprehension increases, so does cultural polarization.
This conclusion supports what I call the “expressive rationality thesis.” The expressive rationality thesis holds that it CCR is rational at the individual level.
CCR is not necessarily conducive to formation of accurate beliefs under conditions in which opposing cultural groups are polarized. But the “cost,” in effect, of persisting in a factually inaccurate view is zero; because an ordinary individual’s behavior—as, say, consumer or voter or participant in public debate—is too small to make a difference on climate change policy (let’s say), no action she takes on the basis of a mistaken belief about the facts will increase the risk she or anyone else she cares about faces.
The cost of forming a culturally deviant view on such a matter, however, is likely to be “high.” When positions on risk and like facts become akin to badges of membership in and loyalty to important affinity groups, forming the wrong ones can drive a wedge between individuals and others on whom they depend for support—material, emotional, and otherwise.
It therefore makes sense—is rational—for them to attend to information in issues like that (issues needn’t be that way; shouldn’t be allowed to become that way—but that’s another matter) in a manner that reliably aligns their beliefs with the ones that dominate in their group. One doesn’t have to have a science Ph.D. to do this. But if one does have a higher capacity to make sense of technical information, one can be expected to use that capacity to assure an even tighter fit between beliefs and identity—hence the magnification of cultural polarization as science comprehension grows.
3. Ideology, motivated reasoning & cognitive reflection
The “Ideology, motivated reasoning & cognitive reflection” experiment ) (IMRCR) picks up at this point in the development of the project to understand CCR. The Nature Climate Change study was observational (correlation), and while it identified patterns of risk perception more consistent with CCR than alternative theories (ones focusing on popular deficiencies in system 2 reasoning, in particular), the results were still compatible with dynamics other than “expressive rationality” as I’ve described it. The IMRCR study uses experimental means to corroborate the “expressive rationality” interpretation of the Nature Climate Change study data.
It also does something else. As we have been charting the mechanisms of CCR, other researchers and commentators have advanced an alternative IMR (ideologically motivated reasoning) position, which I’ve labeled the “asymmetry thesis.” The asymmetry thesis attributes polarization over climate change and other risks and facts that admit of scientific investigation to the distinctive vulnerability of conservatives to IMR. Some (like Chris Mooney) believe the CCR results are consistent with IMR; I think they are not but that they really haven’t been aimed at testing the asymmetry thesis.
The IMRCR study was designed to address that issue more directly, too. Indeed, I used ideology and party affiliation—political orientation—rather than cultural predisposition as the hypothesized motivating influence for information processing in the experiment to make the results as commensurable as possible with those featured in studies relied upon by proponents of the asymmetry thesis. In fact, I see political orientation variables as simply an alternative indicators of the same motivating disposition that cultural predispositions measure; I think the latter are better, but for present purposes political was sufficient (I can reproduce the data with cultural outlooks and get stronger results, in fact).
In the study, I find that political orientations exert a symmetrical impact on information processing. That is, “liberals” are as disposed as “conservatives” to assign weight to evidence based on the congeniality of crediting that evidence to their ideological predispositions (in other words, to assign a likelihood ratio to it that fits their goal to “express” their group commitments).
In addition, for both the effect is magnified by higher “cognitive reflection” scores. This result is consistent with—and furnishes experimental corroboration of—the “expressive rationality” interpretation of the Nature Climate Change study.
4. So—“is ideologically motivated reasoning rational? And do only conservatives engage in it?”
The answer to the second question—only conservatives?—is I think “no!”
I didn’t expect a different answer before I did the IMRCR experiment. First, I regarded the designs and measures used in studies that were thought to support the “asymmetry thesis” as ill-suited for testing it. Second, to me the theory for the “asymmetry thesis” didn’t make sense; the motivation that I think it is most plausible to see as generating polarization of the sort measured by CCR is protection of one’s membership and status within an important affinity group—and the sorts of groups to which that dynamic applies are not confined to political ones (people feel them, and react accordingly with respect to, their connection to sports teams and schools). So why expect only conservatives to experience IMR??
But the way to resolve such questions is to design valid studies, make observations, and draw valid inferences. I tried to that with the IMRCR study, and came away believing more strongly that IMR is symmetric across the ideological spectrum and CCR symmetric across cultural spectra. Show me more evidence and (I hope) I will assign it the weigh (likelihood ratio) it is due and revise my position accordingly.
The answer to the second question—is IMR rational—is, “It depends!” The result of the IMRCR study supported the “expressive rationality” hypothesis, which, in my mind, makes even less supportable than it was before the hypothesis that IMR is a consequence of heuristic-driven, bias prone “system 1 reasoning.”
But to say that IMR is “expressively rational” and therefore “rational” tout court is unsatisfying to me. For one thing, as emphasized in the Nature Climate Change paper and the IMRCR paper, even if it is individually rational for individuals to form their perceptions of a disputed risk issue in a way that protects their connection to their cultural or ideological affinity groups, it can be collectively disastrous for them to do that simultaneously, because in that circumstance democratically accountable actors will be less likely to converge on evidence relevant to the common interests of culturally diverse groups. We can say in this regard that what is expressive rational at the individual level is collectively irrational. This makes CCR part of a collective action problem that demands an appropriate collective action solution.
In addition, I don’t think it is possible, in fact, to specify whether any form of cognition is “rational” without an account of whether it conduces or frustrates the ends of the person who displays it. A person might find MR that projects his or her identity as a sports fan, e.g., to be very welcome—and yet regard MR (or even the prospect that it might be influencing her) totally unacceptable if she is to be a referee. I think people would generally be disturbed if they understood that as jurors in a case like the one featured in They Saw a Protest they were perceiving facts relevant to other citizens’ free speech rights in a way that reflected IMR.
Maybe some people would find it unsatisfying to learn that CCR or IMR is influencing how they are forming their perceptions of facts on issues like climate change or gun control, too? I bet they would be very distressed to discover that their assessments of risk were being influenced by CCR if they were parents deciding whether the HPV vaccine is good or bad for the health of their daughter.
Chris Johnston’s book The Ambivalent Partisan is very relevant in this respect. Chris and his co-authors purport to find a class of citizens who don’t display the form of IMR (or CCR, I presume) that I believe I am measuring in the IMRCR paper. They see them as ideally virtuous citizens. It is hard to disagree. And hence it is confusing for me to know what to think about the significance of thing that I think (or thought!) I understood. So I need to think more. Good!