Reasoning about climate change

Contributions

• Direct experimental evidence that deliberation does not amplify partisan bias in climate change argument evaluation, contrary to the motivated System 2 reasoning (MS2R) framework.
• First causal test of the belief-coherence account of deliberation: when people reason more carefully, their judgments align more closely with their actual prior beliefs (not driven by partisan identity per se).
• Demonstrates that reasoning magnifies the effect of prior beliefs primarily for less-familiar arguments, suggesting deliberation operates via improved comprehension and alignment with existing knowledge rather than partisan motivated reasoning.
• Identifies the role of argument familiarity in moderating deliberation effects: prior beliefs have a stronger interaction with deliberation for novel, unfamiliar arguments.

Method

Two pre-registered studies with American MTurk participants (Study 1: N≈1,000; Study 2: N≈1,270). Both employed experimental conditions manipulating System 1 (intuitive) versus System 2 (deliberative) processing:

1. Intuitive-response condition: Participants memorized a 4×4 dot pattern (cognitive load), then answered within 28 seconds while viewing the pattern, minimizing deliberative capacity.
2. Deliberative-response condition: Participants evaluated arguments without cognitive load or time pressure, allowing full deliberation.

Stimuli: Six climate change arguments (3 pro-climate, 3 contra-climate), presented in randomized order via slider ratings (0=completely disagree, 100=completely agree). Arguments were content-balanced and drawn from procon.org, framed in politically neutral language with no partisan cues.

Key measures:

• Partisanship: Single-item political preference question (1=Strongly Democratic, 6=Strongly Republican).
• Prior beliefs about climate change: In Study 1, single question on human-caused climate change; in Study 2, averaged questions on causation and severity.
• Cognitive reflection: Cognitive Reflection Test (CRT) to measure disposition toward deliberation.

Analyses: Linear mixed-effect models with crossed random effects for participant and argument. Models included prior belief, partisanship, argument type (pro/contra), response condition (intuitive/deliberative), and all relevant interactions.

Results

Separability of partisanship and prior beliefs: Correlation between partisan affiliation and climate belief was moderate (r=0.34 in Study 1, r=0.44 in Study 2), confirming they are distinct constructs—being Republican does not necessarily mean disbelief in climate change.

Main finding—no MS2R effect: Neither study showed the three-way partisan × argument type × deliberation interaction predicted by MS2R. There was no significant increase in partisan polarization under deliberation (Study 1: no significant three-way interaction; Study 2: similarly non-significant interaction p=.790).

Support for belief-coherence account: A significant three-way interaction between prior belief × argument type × condition was observed:

• For contra-climate arguments: Believers showed lower agreement in the deliberative condition (Study 1: b=−8.1, p<.0001; Study 2: b=−8.52, p<.0001), while deniers showed higher agreement after deliberation (Study 1: b=13.23, p=.008; Study 2: b=10.1, p=.030).
• For pro-climate arguments: No significant change with deliberation for either believers or deniers.

This asymmetry emerged because participants had greater prior exposure and familiarity with pro-climate arguments; deliberation magnified effects primarily for less-familiar (contra-climate) arguments.

Argument familiarity as moderator: A follow-up analysis (N=4,733) measuring participants' rated familiarity with each argument revealed a strong negative correlation (r=−0.64, p=.002) between argument familiarity and the strength of the deliberation × prior-belief interaction: deliberation magnified prior-belief effects most for less-familiar arguments. This suggests deliberation aids comprehension and alignment with prior knowledge on novel claims.

Partisan modulation (secondary analysis): While main effects of partisanship were evident (Republicans and deniers rated contra-climate arguments higher), partisanship did not moderate the effect of deliberation once prior beliefs were accounted for—the causal effect of deliberation on agreement operated through belief coherence, not partisan identity.

Notes

This paper makes a crucial methodological and theoretical contribution by experimentally disentangling partisanship from prior beliefs—a critical control absent in most prior work linking cognitive sophistication to climate denial. The finding that deliberation increases coherence with prior beliefs (rather than partisan identity) is important for understanding why earlier studies found greater partisan polarization among more cognitively sophisticated individuals; the effect may reflect not motivated reasoning but rather better-informed belief updating given different underlying priors. The asymmetry between pro- and contra-climate arguments (deliberation effects only for the latter) is well-explained by familiarity and is a valuable mechanistic insight often missing from macro-level correlational studies.

Limitations: (1) The sample, while quota-matched to national demographics, is ideologically balanced by design (nearly equal Republicans and Democrats), which may underestimate real-world partisan asymmetries in climate belief; (2) Participants were recruited from the US, limiting generalizability to other political cultures; (3) Climate belief was measured with only 2–3 questions per study, which, while deliberate (capturing causation and severity), could be expanded; (4) The arguments were presented in an abstract, survey context, not a realistic information environment where source credibility, messenger identity, and competing narratives shape evaluation. Future work examining how social context, source signals, and real-world information streams influence the relationship between reasoning, prior beliefs, and climate argument evaluation would be valuable.

The authors appropriately note that "rational" (Bayesian) reasoning about climate change does not imply people will become climate believers if their prior beliefs are incorrect—in the real world, prior beliefs reflect exposure to different information streams and possibly different scientific literacy. This underscores the importance of long-term educational interventions and accurate information distribution over appeals to "just think harder."