A note on the perverse effects of actively open-minded thinking on climate-change polarization

Abstract

This research note presents evidence that political polarization over the reality of human-caused climate change increases in tandem with individuals’ scores on a standard measure of actively open-minded thinking. This finding is at odds with the position that attributes political conflict over facts to a personality trait of closed-mindedness associated with political conservatism.

Actively open-minded thinking (AOT) refers to a reasoning disposition that is often thought to have political significance. It consists in the motivation to seek out, engage, and appropriately weigh evidence opposed to one’s strongly held beliefs (Baron, 2008; Stanovich and West, 1997). Some surmise that a deficit in this disposition associated with ideological conservatism (Price et al., 2015; Jost et al., 2003; Taber and Young, 2013) might be the source of political polarization over facts that admit of empirical proof—for example, that human activity is causing the temperature of the earth to increase (Jost et al., 2013; Nisbet et al., 2013). ¹

We will call the position that AOT has this significance for politics the “AOTp thesis.” The point of this research note is to call attention to some evidence that fails to conform to this position. That evidence shows that partisans who score highest on a standard measure of AOT are in fact the most polarized on the reality of human-caused climate change.

The evidence comes from a nationally representative US sample (N = 1600) assembled to aid in construction of the “Ordinary Science Intelligence” assessment, a measure of science comprehension (Kahan, 2016). ² For purposes of validating that measure, subjects responded to a standard seven-item AOT battery (Baron et al., 2015; Haran et al., 2013) (α = 0.61). The subjects also responded to a five-point liberal-conservative and a seven-point partisan-identification measure, which we combined to form a political-outlook scale (“Left_right,” α = 0.78). The data set also contained responses to standard survey questions on human-caused climate change. ³

Subjects’ AOT scores were positively correlated with acceptance of human-caused climate change and negatively correlated with right-leaning political outlooks (Figure 1). These findings are consistent with AOTp. ⁴

Figure 1.

Relationships of AOT to political outlooks and acceptance of climate change.

The more definitive test, however, involves the impact of subjects’ political outlooks conditional on their AOT scores. If polarization over the reality of human-caused climate change is a consequence of a deficit in AOT among conservatives, then one would expect the conservatives lowest in AOT to be substantially more skeptical of climate change than those highest in AOT. Likewise, if an ideological asymmetry in AOT drives partisan conflict over climate change, then the gap between partisans ought to narrow as partisans’ AOT scores go up (Figure 2).

Figure 2.

AOTp prediction. The AOTp thesis posits that a deficit in AOT among ideologically conservative individuals accounts for their resistance to counter-attitudinal evidence on climate change and thus drives political conflict on this issue. If this is true, then one would predict that those conservative individuals who are nevertheless relatively high in AOT would be more inclined to credit evidence of human-caused climate change, narrowing the partisan gap on the reality of this phenomenon.

These results were not observed in the data. As subjects’ AOT scores went up, their acceptance of human-caused climate change increased only if they held left-leaning political outlooks. Among right-leaning subjects, higher AOT scores were associated with slightly less acceptance (Table 1).

Table 1.

Regression analysis.

	b
LR	−1.40  (−17.49)
AOT	0.42  (6.26)
LR x AOT	−0.49  (−6.58)
constant	−0.07  (1.17)
N	1547
LR χ2	575.3(3)
pseudo R2	0.33

AOT: actively open-minded thinking; LR: left_right

N = 1547. Outcome variable is agreement that there is “solid evidence” of recent global warming due “mostly” to “human activity such as burning fossil fuels.” Logit coefficients with z-statistic denoted parenthetically. Left_right and AOT are centered at zero through standardization to promote interpretation (Friedrich, 1982). Pseudo R² calculated by squaring the Pearson correlation between predicted and observed values. Bold denotes that the predictor, the Likelihood Ratio χ², or the pseudo R² is significant at p < 0.05.

As a result, political polarization grew as subjects’ AOT scores increased (Figure 3). For subjects with a low AOT score (−1 SD), the predicted probability that a liberal Democrat would accept that human activity is causing global warming (58%, ± 6%) ⁵ was 39 percentage points (± 7) higher than the predicted probability that a conservative Republican would (19%, ± 3%). For partisans with a high AOT score (+1 SD), there was a 73 percentage point (± 6%) difference (87%, ± 3% vs. 14%, ± 4%).

Figure 3.

Impact of AOT on acceptance of climate change conditional on right-left political outlooks.

These are not the patterns one would expect to see under AOTp. We can think of three explanations.

One is that AOTp is simply false. AOTp is one variant of a more general claim that an asymmetry in critical reasoning explains political conflict over contested policy-relevant facts (Jost et al., 2013). While proponents of this asymmetry thesis have long pointed to correlations between ideology and self-report measures of reasoning style (e.g. Jost et al., 2003), they have not produced evidence that such correlations explain partisan divisions over such facts, much less observable differences in information processing relating to formation of factual beliefs, in representative samples of the general public. On the contrary, experimental studies have found that there is a basic symmetry in manifestations of politically biased information processing (Nisbet et al., 2015; Kahan, in press)—one that is not mitigated but instead magnified by objective measures of cognitive proficiency (Lewandowsky and Oberauer, 2016). The evidence we have presented, while based on a self-report measure of reasoning style, adds further weight to the case against the asymmetry thesis.

Another, related explanation is that AOT in fact enhances the proficiency of reasoning aimed at forming identity-congruent beliefs. Individuals’ personal beliefs have no meaningful impact on their exposure to societal risks like climate change. Those beliefs, however, do activate affective stances that convey individuals’ membership in and loyalty to important affinity groups—information that will be used by other group members to judge their social competence.

Under these conditions, it is expressively rational for individuals to attend to evidence in a manner that conduces to formation of and persistence in beliefs characteristic of their cultural groups (Stanovich, 2013; Kahan, in press, a). Studies suggest that individuals use their critical reasoning dispositions for that purpose. These individuals display greater directional bias in their information processing not because they are more partisan or are more vulnerable to heuristic substitutes for conscious, effortful information processing (Lodge and Taber, 2013; Taber and Young, 2013) but because they are simply better at screening information for identity-congruent inferences (Kahan, in press, b).

The results of this study are consistent with the inference that individuals will use the reasoning dispositions measured by a standard AOT scale in the same way. As relatively liberal individuals make gains in the form of cognitive proficiency measured by the scale, their opinions become progressively more aligned with the view that predominates in the group. The same is so for more conservative ones, although the effect is admittedly less dramatic, possibly due to the relative obviousness of the position that corresponds to their group identity or relatedly to a floor effect in the measure of conservative skepticism about human-caused climate change. The net result is that subjects highest in AOT are in fact the most polarized, just as individuals highest in numeracy, cognitive reflection, and science comprehension are (Kahan et al., 2012; Kahan, 2016).

Third and finally, the results of the study could be understood to suggest that the standard measure of AOT included in the data we analyzed is not valid. AOT is supposed to evince a motivation to resist “my side” bias in information processing (Stanovich et al., 2013). Thus, one might naturally expect the individuals highest in AOT to converge, not polarize all the more forcefully, on contested issues like climate change. Because our evidence contravenes this expectation, it could be that the AOT scale on which our results are based is not faithfully measuring any genuine AOT disposition.

We do not ourselves find this last possibility convincing. Again, the results we report here are consistent with those reported in many studies that show political polarization to be associated with higher scores on externally validated, objective measures of cognitive proficiency such as the Cognitive Reflection Test, numeracy, and science literacy (Lewandowsky and Oberauer, 2016; National Research Council, 2016; Kahan, in press, 2016; Kahan et al., 2012). Because such results do nothing to call these measures into doubt, we do not see why our results would cast any doubt on the validity of the AOT scale we used, which in fact has also been validated in other studies (e.g., Haran et al., 2013; Baron et al., 2015; Mellers et al., 2015).

Instead we think the most convincing conclusion is that the disposition measured by the standard AOT scale, like the dispositions measured by these other cognitive-proficiency measures, is one that has become tragically entangled in the social dynamics that give rise to pointed, persistent forms of political conflict (Kahan, in press, b). As do other studies, ours “suggest[s] it might not be people who are characterized by more or less myside bias, but beliefs that differ in the degree of myside bias they engender” (Stanovich and West, 2008: 159). “Beliefs” about human-caused climate change and a few select other highly divisive empirical issues are ones that people use to express who they are, an end that has little to do with the truth of what people, “liberal” or “conservative,” know (National Research Council, 2016; Kahan, 2015). ⁶

Obviously, our findings, themselves open to more than one reasonable interpretation, make only an incremental contribution to the larger body of knowledge in this area. We report them so that researchers can give them the weight they believe they are due and, more importantly, take them into account in designing still further studies that can help to expand understanding of the role of critical reasoning dispositions in biased political information processing.