follow CCP

Recent blog entries
popular papers

Science Curiosity and Political Information Processing

What Is the "Science of Science Communication"?

Climate-Science Communication and the Measurement Problem

Ideology, Motivated Cognition, and Cognitive Reflection: An Experimental Study

'Ideology' or 'Situation Sense'? An Experimental Investigation of Motivated Reasoning and Professional Judgment

A Risky Science Communication Environment for Vaccines

Motivated Numeracy and Enlightened Self-Government

Making Climate Science Communication Evidence-based—All the Way Down 

Neutral Principles, Motivated Cognition, and Some Problems for Constitutional Law 

Cultural Cognition of Scientific Consensus

The Tragedy of the Risk-Perception Commons: Science Literacy and Climate Change

"They Saw a Protest": Cognitive Illiberalism and the Speech-Conduct Distinction 

Geoengineering and the Science Communication Environment: a Cross-Cultural Experiment

Fixing the Communications Failure

Why We Are Poles Apart on Climate Change

The Cognitively Illiberal State 

Who Fears the HPV Vaccine, Who Doesn't, and Why? An Experimental Study

Cultural Cognition of the Risks and Benefits of Nanotechnology

Whose Eyes Are You Going to Believe? An Empirical Examination of Scott v. Harris

Cultural Cognition and Public Policy

Culture, Cognition, and Consent: Who Perceives What, and Why, in "Acquaintance Rape" Cases

Culture and Identity-Protective Cognition: Explaining the White Male Effect

Fear of Democracy: A Cultural Evaluation of Sunstein on Risk

Cultural Cognition as a Conception of the Cultural Theory of Risk

« Weekend update: Disentanglement Principle, A Lesson from SE Fla Climate Political Science Lecture (& slides) | Main | Believing as doing . . . evolution & climate change »

Science curiosity and identity-protective cognition ... a glimpse at a possible (negative) relationship 

So here is a curious phenomenon: unlike pretty much every other science-related reasoning disposition, science curiosity seems to avoid magnifying identity-protective cognition!

Let's start with a bunch of culturally contested societal risks, ones on which political polarization can be assessed with the ever-handy Industrial Strength Risk Perception Measure:

Now consider:

click it-- to blow your mind! And for closer look

For each risk, the paired panels chart the risk-perception impact of greater science comprehension and greater science curiosity (in each case “controlling for” the influence of the other), respectively.  They estimate those effects separately, moreover, for a "liberal Democrat" and for a "conservative Republican," designations determined by reference to the study subejects' scores on a composite political ideology and party-identification scale.

As science comprehension (measured with the Ordinary Science Intelligence assessment) increases, so too does the degree of polarization on politically contested risks involving climate change, gun possession, fracking, marijuana legalization, pornography, and the like. 

That’s not a surprise.  The warping effect of identity-protective cognition on cognitive reflection, numeracy, science comprehension and all other manner of critical reasoning proficiency has been exhaustively chronicled, and lamented, in this blog.

But that’s not what happens as science curiosity increases.  On the contrary, in all cases, greater science curiosity has the same general risk-perception impact—in some cases enhancing concern, in some blunting it, and in others having no directional effect—for study respondents of politically diverse outlooks.

Science curiosity is being measured for these purposes with the CCP/Annenberg Public Policy Center “Science Curiosity Scale,” or SCS_1.0.  

SCS_1.0 was developed for use in the CCP/APPC “Evidence-based Science Filmmaking Initiative.” Previous posts have discussed the development and properties of this measure, including its ability to predict engagement with science documentaries and other forms of science information among diverse groups.

So has its relationship to random other non-science related activities, such as taking a peek at what goes on at gun shows and even cracking open a book on religion now & again.

But this feature of SCS_1.0—its apparent ability to defy the gravitational pull of identity-protective cognition on perceptions of disputed risks—is something I didn’t anticipate. . . .

Indeed, I really don’t want to give the impression that I “get” this, it makes “perfect sense,” etc. Or even that there’s necessarily a “there” there.

An observation like this is just corroboration of the fundamental law of the conservation of perplexity, which refers to the inevitable tendency of valid empirical research to generate one new profound mystery (at least one!) for every mystery that it helps to make less perplexing (anyone who thinks “mysteries” are ever solved by empirical inquiry has a boring conception of “mystery” or, more likely, a misconception of how empirical research works).

But here are some thoughts:

1. It does in fact make sense to think of curiosity as the cognitive negation of motivated reasoning.  The latter disposition consists in the unconscious impulse to conform evidence to beliefs that serve some goal (like cultural identity protection) unrelated to figuring out the truth about some uncertain factual matter.  Curiosity, in contrast, is an appetite not only to know the truth but to be surprised by it: it consists in a sense of anticipated pleasure in being shown that the world works in a manner that is astonishingly different from what one had thought, and in being able to marvel at the process that made it possible for one to see that. 

 When one is in that state, the sentries of identity-protect are necessarily standing down.  The path is clear for truth to march in and enlighten . . . .

2. At the same time, these data are pretty baffling to me.  No way did I expect to see this.

The affinity between identity-protective cognition and critical reasoning, I’m convinced, reflects the role the former plays in the successful negotiation of social interactions.  Where positions on disputed issues of risk become entangled in social meanings that transform them into badges of membership in and loyalty to opposing cultural groups, it is perfectly rational, at the individual level, for people to adopt styles of information processing that conduce to formation of beliefs that express their tribal allegiances.

Indeed, not to attend to information in this manner would put normal people—one’s whose personal beliefs about climate change or fracking or gun control don’t have any material impact on the risks they or anyone else face—in serious peril of ostracism and ridicule within their communities.

I’d essentially come to the bleak, depressing, spirit-enervating conclusion, then, that the only reasoning disposition likely to blunt the force of identity-protective cognition was a social disability in the nature of autism.  

But now, for the 14 billionth time, I will have to rethink and reconsider. 

Because clearly the appetite to seek out and consume information about the insights human beings have acquired through the use of science’s signature methods of disciplined observation and inference is no reasoning disability.  And those who in who are most impelled to satisfy this appetite are clearly not using what they learn to forge even stronger links between their perceptions of how the world works and the views that express membership in their identity-defining affinity groups.

Or at least that’s one way to understand evidence like this.  Pending more investigation.

3.  All sorts of qualifications are in order.

a. For one thing, SCS_1.0 is a work in progress.  Additional tests to refine and validate it are in the works.

This is the most informative way I think to examine distribution of reasoning proficiencies across subpopulations. It's also connected to an algorithm I've developed to predict movement in the stock market. email me for details.b. For another, science comprehension and science curiosity are not wholly unrelated! Actually, they aren’t strongly related; in the data set from which these observations come, the correlation is about 0.3. But that's not zero!

I actually tested for “interactions” between science comprehension (as measured by OSI_2.0) and science curiosity (as measured by SCS_1.0), and between the two of them and political outlooks.  The interactions were all pretty close to zero; they wouldn’t affect the  basic picture I’ve shown you above (but I am happy to show more pictures—just tell me what you want to see).

Still I don’t think the effect of science curiosity on identity-protective cognition can be made sense of without closer, more fine-grained examination of how much it alters the trajectory of polarization at different levels of science comprehension.

c.  Also, the impact of science curiosity is interesting only because it doesn’t magnify polarization. It doesn’t make it go away, as far as I can tell.  That’s important—for the reasons stated.  But a reasoning disposition that generated convergence among individuals of diverse cultural outlooks on culturally contested risks (as science comprehension does on culturally uncontested ones) would be much more remarkable—and important. 

We should be looking for that.  I’d say, though, that looking even harder at curiosity might help us detect if there is such a reasoning quality—the ”Ludwick factor” is the technical term   for those who’ve speculated on its possible existence . . .—and how it might be disseminated and stimulated. 

For surely, that is a reasoning disposition the cultivation of which should be cultivated in the citizens of the Liberal Republic of Science.

But in the meantime, this unexpected, intriguing relationship can be contemplated by curious people with excitement and perplexity and with a desire to figure out even more about it.

So what do you think?

PrintView Printer Friendly Version

EmailEmail Article to Friend

References (1)

References allow you to track sources for this article, as well as articles that were written in response to this article.

Reader Comments (30)

I am drawn to the global warming beliefs chart, which seems to show that scientifically curious individuals of either party tend to evaluate global warming as a more significant threat to health and safety. In particular, more curious conservatives are more likely to espouse a belief more commonly associated with liberals. I am interested in whether science curiosity is correlated with political group identification, such that the scientifically-curious conservatives are (for example) more moderate than their conservative peers, and thus more likely to break with their group on one or more issues. If this potential dynamic has already been addressed, could you point me to a source that addresses it? If not, could you comment on whether a potential correlation between scientific curiosity and political affiliation might help make sense of the surprising results that science curiosity seems to indicate willingness to break with one’s group?

February 25, 2016 | Unregistered CommenterAaron Match


You can see the relationship between distribution of science curiosity and political outlooks here.

But as I'm sure you realize, this can't possibly answer your question. The graphics in this post model the impact of science curiosity conditional on a particular value on the political orientation measure (one equivalent to identifying oneself as a "conservative republican" or "liberal democrat". What you want to know is whether the "more curious" *conservative republicans* are in fact more moderate intheir p[olitical outlooks than "less curious" *conservative republicans*--a question that cant' be answered by looking at how curiosity is distributed across individuals whose outlooks are measured w/ that same political orientation scale. I'd need another political outlook scale--a more discerning one-- to see if curiosity is picking up on some sort of liberal political orientation that evades measurement with the left-right measure featured here.

It's possible. I'll think about whether there's a way to figure that out w/ these data ...

Another possibility is that more curious individuals are less intensely partisan; so even at the same score on "left_right," the conservatives are less intense in their views, as would be the liberals (or maybe the partisan intensity effect is asymmetric, who knows).

What would you say is the most plausible serious alternative hypothesis to the one you are thinking of? And then what would you observe to help determine the relative plausibility of the two?

February 25, 2016 | Registered CommenterDan Kahan

I think that the test for what it means to be scientifically curious needs to be further refined.

Going back to this post:, I note that the four choices are different in many aspects that affect audience that have little to do with science. In particular, the science articles are from Science, the publication of the American Association for the Advancement of Science. This journal is not generally available to those who are not themselves members, or who work for an institution which has a subscription. The only saving grace here is that the members of the public being tested would have no idea that these selections were from a restricted source. But given the four things available to pick from, they may have been only trying to impress the test giver with their seemingly higher level tastes. Thus they are exhibiting not curiosity, but a clever form of projecting an artificial identity to fool the test giver into thinking that they are equal colleagues.

If we were trying to test for "What kind of stuff do you like to read?" I think that the materials would need to be grouped by reading level and access points, like which ones are on TV, on the Internet , at supermarket checkout stands and which ones are located further back in the store. Perhaps:

A. People magazine, USA Today, Car and Driver, Popular Science or Discover

B. New York Times, Consumers Reports, National Geographic, Scientific American

But that does not test for curiosity.

People in rural areas might have a greater opportunity for natural observations. Certainly most urban dwellers would make poor astronomers. It could be that one could ask everyday occurrence questions that involve critical thinking and mechanical aptitude. For example:

If your toilet is making funny noises do you:

A. Ignore it

B. Look under the lid. See that the flapper is not sealing correctly. Try to fix it. When the problem continues, go to the hardware store get a kit and go home and replace it.

But that would be confounded by renters. If not paying for the water bill and not current with the rent (and thus not wanting to phone in a complaint to the landlord) A would be the right answer.


If you see a spider do you:

A. Squash it
B. Try to figure out what kind it is.

But that requires books or identification materials, and it would be very useful to have enough prior knowledge to know it is not a black widow before you decide to pick it up.

So those are not a good general tests either. But I don't think that sitting around watching films has necessarily that much relationship to curiosity. Nor is blind acceptance of the latest well publicized "scientific theory".

Somehow there had to be a human who had to start at ground zero, but we are not among them. But how do you effectively measure scientific curiosity when nearly everyone is at a different prior knowledge starting point?

What is scientific curiosity anyway?

February 25, 2016 | Unregistered CommenterGaythia Weis

Thanks for your reply. The crux of what I am wondering is what behaviors are associated with an individual whose beliefs diverge from those of their group, and it seems that curiosity may be one such behavior. However, it occurred to me that if an individual is already more divergent from their group (more moderate), then two measures that are supposed to show independent qualities of that individual (e.g. scientific curiosity and global warming belief) may actually both be measuring the degree of that individual’s moderate-ness. In this particular case, if scientific curiosity among conservatives is related to moderate-ness and if global warming belief is related to moderate-ness, then the positive relationship between science curiosity and perceived global warming risk among conservatives may simply track two different proxies for “moderate-ness”.

The following experiment would resolve my question:

Hypothesis: The tendency of highly scientifically curious conservatives to perceive that global warming poses a more significant risk than is perceived by the average conservative is explained by the fact that scientifically curious conservatives tend to be more moderate conservatives, and thus predisposed to differ from their group on issues such as global warming.

Alternative hypothesis: The tendency of scientifically-curious conservatives to perceive that global warming poses a more significant risk than is perceived by the average conservative is NOT explained by a dynamic in which scientifically-curious conservatives tend to be more moderately conservative.

Data: Survey data of individuals that measures their political views on polarized issues, political group affiliation on a continuum (this is important, and is not obviously a part of the above figure), and scientific curiosity

To start, it would be necessary to test whether there exists a relationship between the moderate-ness of conservative views and the level of scientific curiosity. If there exists no such relationship, then the dynamic I suspect may be at work will be shown to in fact not be. If there does exist a relationship between moderate-ness of conservative views and the level of scientific curiosity, then there could be a way to use such a relationship to explain why scientifically-curious conservatives also believe global warming is of high risk without their scientific curiosity necessarily being the driving factor for such a belief.

February 25, 2016 | Unregistered CommenterAaron Match


I think there’s a highly plausible candidate explanation for this.

The strong polarization that occurs for the increasingly domain knowledgeable in an environment where one or both sides is heavily influenced by culture, is because each faction maintains their own *different* knowledge base. You can see evidence for this in the different gradients in your polarization graphs for climate change and creationism. Creationism stems from very old, ‘received’ culture, i.e. folks are brought up with it. So there’s simply less (fundamentals) to learn from scratch for those who travel to expert status as (mainly) adults; hence lower gradient.

However, the merely science curious haven’t necessarily got as far along with those knowledge bases, or less so on average at least, despite their aspirations. It is not the science aspiration that drives the increasing polarization, it is the level of (different!) domain knowledge that individuals end up soaking themselves in / committing to (their path being vectored by initial cultural bias or innate skepticism of culture). So if you mask out that knowledge factor by whatever means, the enhanced polarization should also go away.

As you note there are confounding factors to a full unravelling though. Not least the overlap you point to suggests that a large chunk of the science curious will presumably also be science comprehending (and this comprehension may also be selective, i.e. it may hit or miss science from particular domains). I think your graphs more or less confirm that topics *directly* related to science, like say climate change or fracking, have more polarization in left-hand charts than topics indirectly related to science, like say p0rn0graphy (this is at least a logical inferance). So there’s more accentuated effects to help unravel stuff with the direct science topics, and therefore these may be the best place to concentrate further efforts on.

So for potential further deconfounding, a good comparison would be to put your existing polarisation graph for increasing OCSI scores, against a graph of increasing science curiosity, but with the latter derived only from a population confined to *low* OCSI scores, say 3 or less. So then you would genuinely (I think and hope) be comparing the effects of ‘increasing aquired knowledge’, against the effects of ‘increasing curiosity but not knowledge’, both within the same directly science related (climate change) domain. I’d expect the right-hand chart to show flat railway tracks running from their starting (default cultural bias) positions.

Note sure why you get blunting or enhancing gradients in existing right-hand charts, but perhaps that’s a part of the above confounding factors. Not thought about it much yet. But if my suggested experiment didn’t produce flat parallel lines (not blunted or enhanced), I can’t currently think why that would be.

I guess another confounding factor is that you are viewing all effects only through the lens of left-right cultural bias. Yet there are other cultural influences in other dimensions, so you will be seeing only the projection of these onto the left-right political axis. May matter not or hardly at all for some domains, maybe much more so for others. For instance religion is another axis. Shouldn’t matter for climate change, is weak (you can see the climate / political gradient is opposite to the religious political gradient, largely cancelling), but may matter for immigration, say. Some minority religious camps poll completely oppositely to the bulk left-right result on religious belief, and there is probably not a fixed left-right proxy for religion on all issue domains.

February 26, 2016 | Unregistered CommenterAndy West


I understand and largely share the judgments that motivate your thoughta about the instrument. A couple of things I would say, though:

1. In the end, what we want is an instrument that predicts as discerningly as possible *behavior* that we would understand to evince "science curiosity." For our purposes, "science curisosity" is a disposition to seek out & consume for personal pleasure information geared to making what science knows comprehensible to non-experts. Science documentaries are the "prototype" of taht sort of science communication for us. If watching science documentaries to satisfy that sort of appetite doesn't count as "science curiosity" that someone wants to measure, then for sure our scale will not be valid in their eyes.

2. For sure, any set of indicators we use to construct the instrument that does measure what *we* have in mind as "science curisosity" will be noisy. Some people will select "science curiosious" responses for reasons that aren't connected to the relevant disposition; others who have the dispositoin will sometimes fail to for reasons that reflect the lack of fit between the indicators and how they would manifest the disposition we are interested in. You minimize noise in part by designing the items as well as you can but even more by using *multiple* items: their covariance will amplify their joint measurement signal, while their noise-- the way in which they vary in manners not related to variance in science curisosity -- will cancel out. If we can't do a good enough job, though, in maximizing signal & minimizing noise in this way, We'll pay for that w/ an instrument that doesn't perform well.

3. I actually like your spider & toilet items. *If* they helped to predict who is going to watch a cool show like your Inner Fish rather than turn the channel to Honey Boo Boo goes to Las Vegas, then I'd treat them as valid indicators of the dispositoin we are interested in! My own bias, though, is to avoid self-reported attitudes-- anything that basically asks people to vocuh for their own curiosity. Self-reproted behavior is better, if it's not obvious to the person that one is in fact trying to assess their possession of socially desirable traits. Best of all are performance-based measures: ones in which people show the disposition we are intersted in by some action that one would associate with it & be less likely to see in a person who doesn't have it.. WE had one of those in SCS_1.0 & will have more in SCS_2.0

February 26, 2016 | Registered CommenterDan Kahan


Your point is a good one: again, as I udnerstand it, is it possible, in effect, that science curiousity correlates with a disposition toward political "moderation" that evades ready detection in our left-right political outlook measure?

One thing I did to explore this was look at whether SCS scores interact with (or moderate) the relationship between political outlooks and cultural worldviews. . If science curious "liberals" or "conservatives" are less likely than noncurious ones to adopt egalitarian & hierarhical views, respectively, or communitarian or individualistic ones, respctively, that would be consistent with the "scs = political moderation" hypothesis. But this wasn't so: political outlooks have the same relation to cultural worldvikews among both high & low SCS study subjects. I'll post something on this.

In addition, the tendency of high SCS isn't unikformly toward moderation across the contested risks featured in this post. HIgh SCS conservatives aren't less extreme on risks that illegal imigrants pose; high SCS liberals aren't too moderate on fracking. HIgh OSI seems to "moderate" conservative views at least smuch as high SCS on pornography & marijuana legalization. What's distinctive, I think, in the pattern of rsponses of high SCS is lack of polariazation on issues on whch high OSI seems to amplify partisan divisions; the "overall views" of high SCS are moderate in some cases,but not others.

But this is just scratching at the issue. If I had "policy items" & political sophistical (a good measure of intensity of partisanship) in this data set, I think I could do a better job seeing if SCS predicts more "modrate" views generally conditional on political outlooks.

I'll keep my eyes peeled. Thanks!

February 26, 2016 | Registered CommenterDan Kahan


I think it won't be possible w/o more data -- collected specifically to test competing conjectures here -- to address all of your counter explanations, or even any of them in a super convincing way. But I might be able to address certain of them to some extent by using the cutlural worldview variables instead of left-right & reporting the data on how OSI & SCS *interact* with each other & w/ worldviews on cuturally polarizing risks-- in other words, do people w/ very high OSI polarize even as their SCS increases, or does higher SCS polarize individuals with low OSI. You proposed looking for such relationships among people w/ differing OCSI scores-- but alas, there is no climate sicence literacy measure in this data set. It is the case, though, that high OSI predicts a high OCSI score, regardless of ideology (even more for OCSI_2.0 than for OCSI_1.0). So I can get reasonably close. Will let you know

February 26, 2016 | Registered CommenterDan Kahan

Is a "curious" person someone who's looking for or just open to alternatives to what they think today? That could be true even if one is not "moderate".

February 26, 2016 | Unregistered CommenterHal Morris


>>'I think it won't be possible w/o more data...'

That the LHS charts polarize because of immersion in 2 different knowledge bases is pretty much a given. The polarization represents increasing cultural bias (from one or both sides, indeed one needs more data than these charts to say who is who). But culture is essentially a defended canon of knowledge, written and thought, and the nearer the centre on that canon individuals are, the more they will diverge from either competing canons or from evidential based knowledge (depending upon the domain / dispute). For issues absolutely core to a culture (creationism for religion, belief in the certainty of climate calamity for climate culture), one can also discern something useful about the gradients too. But for peripheral topics such as immigration or gun control, I think domain knowledge is needed to interpret the gradients (and hence a source of bias may creep into the interpretation).

To address the proposal that the polarization dissapears because most of the highly domain knowledgable are filtered out, I guess requires proving that the higher SCS folks are mostly not high OSI too. But already you do say 'Actually, they aren't strongly related... ...the correlation is about 0.3' (which admittedly surprises me).

Or alternatively per my suggestion, actively removing the high knowledge folks in a particular domain (e.g. in CC domain by picking low OCSI folks only), and showing that the LHS charts retain similar shapes.

I realise both these exercises are easier said than done, but I can't see anything in your current charts that rules my proposal out thus far.

I had an extra thought. While most of the RHS charts do approximate flat railway tracks stretching right, I realised that in cases where there there may be a fairly dominant public story about what the 'right answer' is, then in the absence of detailed domain knowledge (from *either* opposing knowledge base) the science curious may nevertheless become more aware of this dominant 'answer'. This speaks nothing as to whether the 'answer' is actually more related to evidence than culture, it could be either. It just says that one of the 2 sides has (largely, currently), captured authority. This is true of the climate change case, and so may explain why the parallel railway tracks angle upwards rather than go flat right per my original assumption. I'm not so familiar with the other domains, but I presume there's not a strong equivalent of the cc consensus message so heavily pushed by authority.

Re use of cultural worldview variables, this is good to get more leverage. Yet if you mean only your heirarchical / egalitarian variables, these are only slightly wider than the political left-right axis, as each party incorporates h or e as a core value anyhow. The net needs to be much wider, for instance capturing religious cultures and (though you don't believe it exists, so the possibility of), climate culture, neither of which have an h /e main axis.

>>'So I can get reasonably close. Will let you know'

Will still be very interesting, thanks :)

February 27, 2016 | Unregistered CommenterAndy West

'To address the proposal that the polarization dissapears...'

should read

'To address the proposal that the increasing polarization disappears...'

February 27, 2016 | Unregistered CommenterAndy West

Wow. Our speculation from the other thread that "science communication" efforts in a liberal society should be directed towards the cultivation of curiosity seems to be bearing very interesting fruit.

It seems to me that people in the mindset of curiosity may take their priors with them into a situation, but set aside their self-image and self-affiliation and values in determining how they will let their own views be changed. In simpler and more general terms, they do not approach the world from the perspective of who they are.

Dan, is there any difference between general curiosity and science curiosity in terms of their effect on scientifically technical subject matters like fracking and climate change? I notice that those issues are the only ones in which the two sides of the political divide track together with increasing curiosity. In all the rest save gun possession, increasing curiosity acts as a weak attractor of one side to the other side's position, which is held constant.

Maybe gun possession is an example of an issue that really -does- require consensus messaging, the effect of increasing curiosity being -diverging- in opinion.

February 27, 2016 | Unregistered Commenterdypoon


>>'In simpler and more general terms, they do not approach the world from the perspective of who they are.'

Bear in mind that as Dan reminds us, the right-hand charts are still polarized. Just not *increasingly* polarized with higher SCS, as is the case for higher OSI. And so given that folks on the RH side of the RH charts are (largely) not taking a position on the basis of increased knowledge, this position may not in itself be either wiser or necessarily less influenced; it depends why they got there.

I think these interesting charts speak more to knowledge than they do about individuals. I.e. our 'knowledge' on controversial issues is itself culturally biased. Which in turn makes it all the harder to know which side happens to lean more toward the evidential than the cultural.

February 27, 2016 | Unregistered CommenterAndy West

@AndyWest & @AaronMatch:

I have some data ... will post it as a WSMD? JA! (I"ve been tied up in litigation w/ @dypoon over when he is going to collect his MAPKIA prize, so that's slowing me down....)

But meantime: @Andywest suggests that cultural worldviews pretty much reduce to left-right. Hierarchical individualsists & egalitarian communitarians are the groups most polarized on climate change. What does @Andywest estimate the probability to be that a "liberal Democrat" is an Egalitarian Communitarian, and a "conservative Republican" a "hierarchical individualist" (realizing that their population frequency in US is constrained to be 25% insofar as everyone is classified in relation to the population mean on worldview scores)?

Also in meantime, do consider the discussion here of how to think about the distribution of science curiosity (or anything else for that matter) across subpopulations.

I said r = 0.3, and you got that that's "modest." But really I think that way of conveying the information doesn't tell a curious (as it were), reflective person enough to figure out much of consequence.

I think the Likelihood Ratio approach I describe in the linked post is the best heuristic device for conveying that information.

But to get the "consequential" part out, *you* (or whoever is asking the question, "how does science curiosity [or science comrpehension or cognitive reflection or whatever] must specify the relevant "cut off point" along the SCS [or OSI, CRT or whatever] population distribution that *you* think matters.

I showed in the linked post and also in the graphic inset in this post that a person who is "above average" in science comprehension is 1.7x more likely than someone "below average" to attain an SCS score that would put him or her at the 90th percentile in "science curiosity."

Is that "consequential"? You tell me; that is, you tell me if *the 90th* percentile is the "consequential" point for assessing relative probabilities, & if so whether you think the odds of 15:1 (for > avg OSI types) vs. 6:1 (below avg ones) of getting to that level of SCS seem "consequentially" different!

I'm not sure *what* to think about the distribution of SCS across groups. But I think I at least have a pretty decent idea of *how* to characterize the distribution for anyone who is prepared to make use of the information.

February 27, 2016 | Registered CommenterDan Kahan

@HalMorris & @Dypoon:

SCS is more modest than you think, perhaps b/c I failed to stress that as much as I should have in this post.

I/we say something here that makes more clear the current, limited scope of our ambitions. More too in the First ESFI REport.

But some more now, to atone for perhaps not being sufficiently persistent:

Ours is, to begin, a conception of "curiosity" that is very much specific to *scientific knowledge*.

It isn't represented or intended -- or remotely expected, although weird shit happens -- to be a global, object-independent form of "curiosity." Forming a measure of such a disposition is a worthy but super ambitious goal -- no one has come close to pulling it off (and as a result, the dominant view in cognitive psychology is that there isn't any such thing; curiosity is viewed as a transient state... a view we reject at least in the domain of science inormation).

Our understanding of "science curiosity" treats it as a a dispositional trait to seek out & consume for personal satisfaction information on what science knows about the workings of the universe. But as I emphasized to @Gaythia, we see engagement with high-end science documenteries of the sort that are shown on PBS in the US (or BBC in UK or ABC in Australia) as prototypical "science curious" behavior. I think it is very reasonable to aspire to something more here too.

But the very limited achievements of the long project to create an *externally valid* mesure of science interest is something we are very keenly aware of. So we are focusing for now on just getting a good measure that will work to advance our own research in showing how empirical methods can be used to improve the craft of those professionals who specalize in satisfying the appetite of culturally diverse members of a liberal society to know & marvel at the shitload of cool stuff their amazing shared way of life makes possible.

*That* said, I'm still pretty freaked out -- or just curious -- about the information in this post

February 27, 2016 | Registered CommenterDan Kahan


>>'@Andywest suggests that cultural worldviews pretty much reduce to left-right.'

Say what?? This is the exact opposite of what I said, and have always been saying. How could you possibly get this so backwards? Here is me just up-thread:

"The net needs to be much wider, for instance capturing religious cultures and (though you don't believe it exists, so the possibility of), climate culture, neither of which have an h /e main axis."

I.e. cultures (and the worldviews aligned to them) have much wider context than only h/e or indeed only left-right, and have completely different axes to either of these, which tools geared to look only for h/e won't find very well. So there was me urging you to widen your net so you would find more. I see from previous threads you have used your h/e variables as though they reflect a totality of worldviews, when they are just a small subset. However, left-right US political axis is more aligned to h/e than some other cultural axes, hence I noted that although use of these variables would therefore indeed widen your net, they don't widen it by nearly enough. Somehow, you must have completely misread this.

Climate culture exists in it's *own* right, which in the US will project onto left-right (and h/e) by virtue of asymmetrical alliance, but it's maximum strength is seen by measuring the climate culture itself, not any of its projections. Similarly, religious culture projects onto left-right via asymmetrical political alliance, but Christianity is a culture in its own right and indeed long pre-dates even the existence of modern left-right parties.

>>'I said r = 0.3, and you got that that's "modest." But really I think that way of conveying the information doesn't tell a curious (as it were), reflective person enough to figure out much of consequence.'

Maybe. Your original expression of surprise was about why the increasing polarization in the LH charts had disappeared in the RH charts; my proposal is limited only to explaining this. And I agree with your previous that this not so easy to prove without more data. Yet are we still answering this same question? Not sure where you are going with the latter part of your post and 'cut off' points. Are you saying that despite the above 'modest' linkage, high OSI scores still dominate towards the RH side of the RH charts? If this is the case, then my proposal fails, or at the least other effects must be going on too. However, if domain knowledge is seriously diluted on the RH side of the RH charts, then amplification of polarization should be too.

February 27, 2016 | Unregistered CommenterAndy West

correction: However, if domain knowledge is seriously diluted on the RH side of the RH charts (in comparison to the RH side of the LH charts), then amplification of polarization should be too.

February 27, 2016 | Unregistered CommenterAndy West

P.S. So I think what matters is: in the 90th percentile SCS, what's the proportion of 90th percentile OSI scorers in there? In the 80th percentile SCS, what's the proportion of 80th percentile OSI scorers in there? etc. If these proportions are low (and assuming these are a constant proxy for all other OSI scorers in the same percentile bucket) then knowledge increase across the RH chart is much less than across the LH chart, hence amplification of polarization should be much less across the RH chart too, pro rata.

February 27, 2016 | Unregistered CommenterAndy West


Not by any means a complete response to your points, but given what I know of scientists and other science-curious people, I must bet against the correctness of the premise of your domain-knowledge-depletion hypothesis in the case of the risk of marijuana legalization; in my experience, in-domain knowledge of the use of marijuana and other drugs most definitely increases with science curiosity. :-)

February 27, 2016 | Unregistered Commenterdypoon


1. The way I apparently got your point about worldviews & political orientations "backwards" was by interpreting your statement " if you mean only your heirarchical / egalitarian variables, these are only slightly wider than the political left-right axis, as each party incorporates h or e as a core value anyhow." I did indeed mean the cultural cognition worldivew scales; those are the only worldview measures I have. I think I *still* understand you to be saying that you don't think that the worldviews I am referring to -- the ones associated with cultural cognition theory -- add much to left-right political orientation in US.

But what do you think -- how likeliy is a "liberal Democrat" to be an egalitarian communitarian, and how likely a "conservative republican" a hierarchical individualist?

2. On what proportion of the individuals at any percentile of SCS are at any given percentile of OSI, you could actually figure this out w/ the information in the post -- that SCS and OSI, both of which are standardized (i.e., have means of 0 and SDs of 1.0) are correlated point 3... But the math-- which would involve heavy-duty calculus-- would definitely be a major pain the in ass to do. Any sane person will just do a simulation to figure out the answer.

A little poking around convinces me that the answer is that someone at the 90th percentile in OSI is about 1.5x as likely to score at the 90th percentile of SCS than is someone who scores at the 50th percentile of OSI and about 2.5x as likely to do so as someone who scores in the 10th percentile of OSI.

There are, of course, about as many people at 10th as at the 90th percentile in OSI. But there are about 3x as many people around the 50th percentile of OSI than around the 10th or 90th...

You can do the math but basically at around 90th percentile of SCS, there should be slightly more people who are at 50th percentile of OSI than 90th percentile of OSI, and about 5 people at either 50th or 90th percentiles of OSI for every 2 at 10th percentile of OSI.

How does that affect your surmise?

In any case, I can model how much differences in SCS affect polarization among different types of people at any level of OSI pretty easily.

February 27, 2016 | Unregistered CommenterDan Kahan


it occurs to me: you might be getting tripped up a bit by the figures.

They reflect the output of regression models in which the influence of each disposition-- OSI and SCS -- is estimated *controlling for* the influence of the other.

Accordingly, the patterns displayed can't be attributed to the covariance of OSI & SCS, something you are very interested n.

Nevertheless, OSI & SCS can *interact*; that is, the impact of science curiosity might vary depending on how high or low someone is in science comprehension. I understand your hypotheses to be based on that possibility.

We have to ask the data *that* question if we want to know the answer. The model that produced the output (and that modeled the effect that can be observed here in the raw data) wasn't asked.

February 28, 2016 | Registered CommenterDan Kahan

@Dan, I wrote below before seeing your last post, but present it anyhow and will think on your last...

>>' The way I apparently got your point about worldviews & political orientations "backwards"...'

Ah, I see. I probably stated that over-strongly, however...

>>'I think I *still* understand you to be saying that you don't think that the worldviews I am referring to -- the ones associated with cultural cognition theory -- add much to left-right political orientation in US.'

...*relatively* speaking, yes. I'm coming at this from cultures downwards not psychology upwards. In this perspective, all worldviews are ultimately an expression of alignment to cultures. And cultures are themselves collections of co-evolving stories within an umbrella narrative, some of which are more constrained by reality than others (for instance political cultures in a democracy are constrained by the reality of being voted out on occassion). My troubles with e / h are that a) e or h seem like two of the social sub-stories, valid and measurable sub-components of several world-view systems, yet not in themselves neccessarily speaking to the root cultural origin. b) even ignoring a), there is a far wider range of worldviews than expressed by e-h values, so looking only at these will miss the majority (except via overlaps). For instance a worldview dominated by a narrative incorporating the sacrifice of Jesus for our sins plus the End Times, doesn't have a lot of e or h in it, this worldview is on a separate axis. In 1016 say, this narrative was king in a lot of countries, yet it is still a significant player in 2016. Because cultures all live in big long alliance strings, which oppose similar strings in a kind of dynamically wriggling worm (and sometimes even contradictory) sort of way, you can detect the projection of one culture (or even sub-component) on another axis via its (typically asymmetrical) alliances. But that is using one worldview as a proxy to see another, rather than viewing it directly.

>>' likeliy is a "liberal Democrat" to be an egalitarian communitarian, and how likely a "conservative republican" a hierarchical individualist.'

I haven't looked at your e / h value tools in any detail, and not thought about this area. And also in the US there is a huge stripe of Independents that tend to get forgotten in the left-right analyses, yet will pop up when looking at other values, often throwing off expectations. However, to answer your question look at Fig 2 and Fig 6 at this link:

These show the projection of religious and climate culture respectively onto a left-right US political axis. It seems Democrats are about as religious as they are climate calamitous, yet Republicans are significantly higher for the former and dramatically lower for the latter. Yet in both cases the true culture is coming from the top, the green stripe, which consists a culture in its own right with its own associated worldviews, which has asymmetrical alliance with political left and right.

So I would guess that a Lib / Dem is more likely to be e-c than to be a core religious or core climate calamitous believer. I.e. more than 42%. and in fact I would say over 50%, because as noted it seems like an important value in the Lib/Dem narrative to me in recent history, but this is basically a guess 0: I would similarly guess that a Rep / Con is at least as likely to be h-i as to be a core religious believer, i.e. more than 58%. Hence if I also assume symmetry (always a nice start :) then I'd go for ~60% as both likelihoods. They could be higher, but I don't think any single values typically score way up beyond 80% or so, because cultures are *collections* of co-evolving stories. The action of identity protection can give scores that *look* much higher, as you can see for both the cases of religion and climate culture (middle stripes get added on), but these are not true belief. So I'd stay down near to 60%, rather than be tempted higher, even for these relatively aligned values.

Dunno if these guesses are close. But my point is that looking directly at the worldviews generated by independent cultures such as religion and climate culture and others, directly, is better than looking via proxies, be they l/r or e/c, and while two views is better than one, they are still proxies and (if my guesses are correct) proxies with *relatively* aligned values.

If you take out the left-right axis and plot e-c directly against belief in climate change, I assume you'd get a somewhat different shape to the graph, most likely (though the main slope should still be in the same direction if you substitute e/c for L/D and h/i for R/C); likewise if you plot e-c directly against religious belief. If you measure something different, you generally find something different.

>>'How does that affect your surmise?'

Thanks for poking about with the maths, appreciated. My own maths skills (which once included calculus and stats) atrophied away decades ago. It looks like while my surmise still has legs, there is certainly no slam dunk demonstrating it's the only or dominant effect.

>>' around 90th percentile of SCS, there should be slightly more people who are at 50th percentile of OSI than 90th percentile of OSI, and about 5 people at either 50th or 90th percentiles of OSI for every 2 at 10th percentile of OSI.'

Which tantalisingly doesn't provide a definitive yes or no. The integral of that bucket means the overall depth of knowledge will be virtually cut in half. So one would expect a similar reduction in the amplification of polarization. Yet in the absence of other factors this is not enough to create parallel tracks, only tracks that diverge at half the gradient. So either the SCS / OSI correlation is skewed (seems doubtful, I'm somewhat familiar with your OSI and it seems 'good stuff' to me, not looked at the SCS), or there are additional factors in play too, which nevertheless doesn't yet rule rule out the proposal as a significant, or even the biggest single, effect.

February 28, 2016 | Unregistered CommenterAndy West


Take a look at this.

But I'll have some more "tomorrow."

February 28, 2016 | Registered CommenterDan Kahan


>>'Take a look at this.'

Interesting. So the centre of each of e/c and h/i essentially falls within the Independents, which means its less of a core value than I assumed. (and incidentally suggests per other data that left and right are becoming more polarized generally, and in the process leaving a lot of their would-be supporters nearer to a centre that doesn't want to be so partisan). But these e/c and h/i values still appear symmetrical on the left-right axis and with the same balance point, hence are likely not measuring a fundamentally different cultural value on a different axis, which is the case for worldviews generated by (any particular) religion in its own right, or by climate culture in its own right. Incidentally, at the link below I find that higher educated Independents, theoretically the least partisan of all US voters, have amplified polarization on CC just like the Dem / Libs and Rep / Cons. In the case of the skeptical higher educated Independents, this is despite the fact that they don't theoretically have a powerful Rep / Con Identity to defend, which was a main underpinning for your knowing-disbelief. This may tie in to above, have to think about that, yet the simplest explanation is that the cultural bias is not coming intrinsically from left-right anyhow, but from an independent climate culture.

Re covariance, I would want each variable to be controlled for the influence of the other. In a generic domian where we don't know on which side is cultural bias (top or bottom, or both, of chart), the theory is that it is the *domain knowledge itself* which is biasing. It is well known that cultures are supported on canons of knowledge, and those individuals with deeper knowledge of the canon are more cultrually committed. So for the RH carts to have parallel lines, says that whether folks have low or high SCS score, their domain knowledge (which will come from 2 opposing knowledge bases, one or both of which is cultural), must nevertheless remain about the same. Clearly one issue here is that we are using OSI as a proxy for domain knowledge anyhow, but at least we are using it on both sides and there is clearly amplified polarization on the LH chart. You say that SCS and OSI interact, and this is a problem I guess; if the OSI scores you estimate for the SCS buckets at equivalent percentiles may *not* tell us the true domain knowledge, then as you say we may need to ask a different question. Per above, as things currently stand my proposal is about 'half right', not very satisfactory for anyone!

February 28, 2016 | Unregistered CommenterAndy West

I guess I ruled out the bizarre possibility that those with less SCS could have more domain knowledge, which would certainly assist the lines to go flat. Surely not?

February 28, 2016 | Unregistered CommenterAndy West

@Andy West "I guess I ruled out the bizarre possibility that those with less SCS could have more domain knowledge, which would certainly assist the lines to go flat. Surely not?"

Those with less SCS could very well have more of what you'd call domain knowledge, while those with more SCS could well have more of what you'd call cult indoctrination -- but that's because your POV is upside down.

February 28, 2016 | Unregistered CommenterHal Morris


Not sure I follow about the "core values." The information inthe linked post was meant to show how much *more* one can explain about the extent and contours of conflict over risk if one doesn't confine oneself to the blunt measure of right-left but instead uses "core values."

People in US arent' very "political" in general. Most can't name their congress representative.

But they are very culturally attuned. If there is a politcal issue--or even a *factual* one -- that becomes infused w/ antagonostic cultural meanings, cultural worldviews will ferret out variance much more effectively than left-right, which can easily leave you underestimating just how deep divisions run.

s Wildavsky surmised, cutlural affinitites are the explanation for how people who are pretty much clueless about politics reliably figure out what "positons" fit their identity. Also, I'd say, how people who are clueless about science figure out what's known by it.

February 28, 2016 | Registered CommenterDan Kahan


>>'The information in the linked post was meant how much *more* one can explain about the extent and contours of conflict over risk if one doesn't confine oneself to the blunt measure of right-left but instead uses "core values." '

Yes, and I think you have 'more'. Very useful. Just, not enough more.

>>'...cutlural affinitites are the explanation for how people who are pretty much clueless about politics reliably figure out what "positons" fit their identity'

Which is why we must widen any search to the full range of cultural affinities that may be in play, which may include those from full-on cultural entities (such as religion), or from lesser yet still strong affinities such as race, also finding the order of precedence in all their alliances (e.g. 43% of Hispanic Catholics are very concerned about climate change, but only 17% of White Catholics, e.g. white mainline protestants split fairly evenly on Obama / Romney, but white evangelical protestants 5 times to 1 for Romney).

It's a much bigger map than either blunt l/r or even l/r + e/c & h/i reveals. For instance some folks are probably more religious than they are Republican. Yet some folks only conveniently align to some religious beliefs for the sake of Republican identity. Some folks are probably more climate calamitous than they are Democrat. Yet some folks only conveniently align to the certainty of climate calamity for the sake of Democrat identity.

February 29, 2016 | Unregistered CommenterAndy West

well we manage to understand each other pretty well considering you speak UK (or equivalent of Canadian) & I American (equivlaent of Australian)

February 29, 2016 | Registered CommenterDan Kahan


Slowly, and despite occasional lapses into Martian and Jovian, yes.

February 29, 2016 | Unregistered CommenterAndy West

PostPost a New Comment

Enter your information below to add a new comment.

My response is on my own website »
Author Email (optional):
Author URL (optional):
Some HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <code> <em> <i> <strike> <strong>