This is last of 3 posts addressing the question “Why cultural worldviews rather than liberal-conservative or Democrat-Republican?” in our studies of risk perception & science communication.
Part 3: The measurement conception of dispositional constructs
This post backs off the “culture dominates ideology” trope — one that could be read into the last two posts but that I actually strongly disavow.
Indeed, my third point—which is actually the most important—is that the question “why cultural worldviews & not left-right?” often is ill-posed. The motivation for it, it often turns out, is if not a mistaken than at least an unappealing (to me) understanding of the point of identifying dispositional sources of conflict over societal risk.
I’ll call the position I have in mind the “metaphysical” conception of cognitive dispositions. I’ll contrast it with another understanding—the one I endorse—that I’ll call the “measurement” conception.
From the point of view of the metaphysical conception, systems of ideas like “liberalism” and “conservativism,” “individualism” and “collectivism,” and even more elaborate constructs are thought to be actual, worldly entities. They are things that are really out there—like trees and lampposts and atoms (in fact, it is a related mistake to think of atoms as worldly phenomena).
Not all of them, but certain of them. Indeed, the primary goal of studying the contribution that these systems make to cognition of politically consequential facts is to identify the “real” one or ones and to expose the nonexistence (or at least inconsequence) of the others. One does this by constructing empirical study designs (or more likely by multivariate statistical tests) that are asserted to “show” that only the “real” one or ones “really” “explain” the relevant state of affairs—or in any case explain “more” of it than does any competing dispositional entity.
The measurement conception sees ideological and cultural constructs as merely tools. Their mission in the scholarly study of perceptions of risk and like facts isn’t to enable demonstration of what “entity” is “really” causing them. Rather, it is to equip us for making sense of what we already know, albeit imprecisely, and even more important for enhancing our ability to manage and control the state of affairs we live in.
We already know the broad outlines of conflict over risk and related facts. It is plain to any socially competent observer that groups whose members display opposing outlooks or styles disagree, often intensely, over diverse packages of risk claims—ones relating to what sorts of behavior or other contingencies threaten society. But we don’t understand this phenomenon well enough to be able to explain, predict, and most importantly of all manage how it effects our collective lives.
The measurement conception says that the key to acquiring that sort of insight isn’t to identify (much less argue about) what “really” causes that sort of conflict but rather to perfect our ability to measure the dispositions associated with the competing sets of risk perceptions with which we are familiar. With reliable and valid measures in hand, we can satisfy (or at least make go about trying to, in the only way that has a chance to succeed) our interests in explanation, prediction, and prescription through appropriately designed scientific tests.
The methods of latent variable modeling are the ones best suited for fashioning such measures. Simply put, these methods aim to enable indirect measurement of some unobservable, or at least unobserved thing on the basis of observable, directly measurable “indicators” or correlates of it. They include the various techniques that psychologists and other social scientists use for measuring diverse sorts of aptitudes and propensities (including attitudes and cognitive styles) that are hypothesized to be the sources of individual differences in one or another behavior, ability, belief, or whathaveyou.
In the study of the cultural cognition of risk (at least as I understand it), the items that make up the “hierarchy-egalitarian” and “individualism-collectivism” scales are nothing more than latent-variable indicators. The responses that study subjects give to them generate patterns, which can then be assessed to confirm that are indeed measuring some unobserved common disposition in those people, and to assess how discerningly they are measuring it.
We hypothesize—and then try corroborate or disprove through empirical studies—that variance in the latent disposition measured in this way generate the distinctive (and very peculiar!) patterns of risk perception that animate debates over issues as seemingly unrelated (at least in any causal sense) as the reality and sources of climate change, the impact of gun control on crime rates, the risks and benefits of the HPV vaccine, etc.
But on this account, our cultural “worldviews” are merely indicators of this latent dispositional propensity. They are not themselves the “thing” that causes conflict over risk or anything else. Nor are they exclusive of other possible measures of the propensities that do.
Indeed, ideologies like “liberalism” and “conservativism” are also indicators of those very propensities. We all know already that they are—we can see that just by looking around us. We can also see that many other characteristics—region of residence and religious affiliation, for example—are also bound up with the outlooks and styles that animate these conflicts.
Indeed, it might well be feasible to combine diverse indicators such as these with each other and with our cultural worldview scales, and thereby generate an even more discerning measure of the latent dispositions or propensities at work in risk conflicts. (It is, in fact, statistically mindless to identify their “independent” influence through multivariate regression, since the covariance that such models “partial out” is exactly what one wants to exploit if one has reason to think they are common indicators of a latent variable.) We have done some work like this, including studies that show how characteristics like gender and race interact with cultural worldviews and others (here & here, e.g.) that try to simulate how collections of attributes treated as cultural profiles or “styles” can influence perceptions.
Indeed, the only justification for preferring a measurement strategy that makes use of fewer rather than more types of indicators is that doing so is, at least for some purpose, more efficient or useful. These are the usual points made in favor of “parsimony,” although stripped of any dogmatic preference for simplicity; the goal is to find the optimal tradeoff between methodological tractability, measurement precision, and ultimately explanatory, descriptive, and prescriptive power.
And it is only on the basis that I would justify the use of our culture measures over of “left-right” ideology ones. That is how my previous two posts should be understood. I emphatically disavow any intention to defend “culture” over “ideology” in the way that is envisioned by the metaphysical conception of cognitive dispositions.
Indeed, the decisive appeal of the measurement conception, for me, is that it avoids all the baggage of a metaphysical style of engagement with social phenomena.