More than scientists: How message and messenger attributes influence viewers’ climate change intentions

PLOS One | https://doi.org/10.1371/journal.pone.0331672 September 10, 2025 10 / 20 viewers responded differently to particular features of the messages or the messengers. To address this, we conducted hierarchical linear modeling (HLM) analyses in which political orientation was entered as a Level 2 predictor of both the intercept and the beta coefficients associated with each Level 1 variable. Significant Level 2 effects—also known as cross- level interactions—indicate that the strength of the relationship between a Level 1 predictor (e.g., affiliation language) and the outcome (motivation) differs depending on the value of the Level 2 variable (political orientation). All significant cross- level interactions were visualized using HLM’s graphing module. Political orientation significantly predicted the intercept from the Level 1 equation, indicating that left-leaning viewers were generally more motivated to act after viewing the MTS videos than their right-leaning counterparts. We then conducted separate Level 2 analyses for each conceptual grouping: the LIWC summary variables, LIWC drive variables, and messenger attributes. A summary of results is provided in Table 3. Our first set of Level 2 analyses exam- ined the LIWC summary variables—analytical thinking, clout, authenticity, and emotional tone. None of these variables exhibited significant cross-level interactions, suggesting that their effects on viewer motivation were not moderated by political orientation.

Our second set of analyses focused on the LIWC drive variables. Two significant cross-level interactions emerged: affiliation and achievement. As predicted, affiliation-related language (e.g., references to community and collective respon- sibility) was more motivating for left-leaning viewers, while achievement-related language (e.g., references to success or human ingenuity) was more motivating for right-leaning viewers. These patterns are illustrated in Fig 1,Fig 2. Finally, we tested whether political orientation moderated the effects of messenger attributes. No significant cross-level interactions were found, indicating that the effects of messenger characteristics—such as being older, male, attractive, and in a natural setting—were consistent across the political spectrum. Table 2. HLM Level 1 Analysis: Effect of Message/Messenger Attributes on Viewers’ Climate Change Motivation. Variable Coefficient SE t df Intercept (mean Motivation) 2.71 .05 59.60** 580 LIWC Summary Variables Analytical thinking −.0001 .0006 −0.19 2,901 Clout −.0002 .0007 −0.30 2,901 Authenticity .0002 .0005 0.88 2,901 Positive emotional tone −.0005 .0005 −1.005 2,901 LIWC Drive Variables Affiliation .005 .048 0.91 2,900 Achievement −.00006 .005 −0.004 2,900 Power .011 .014 1.09 2,900 Reward focus .020 .016 1.27 2,900 Risk focus −.038 .014 −2.72** 2,900 Messenger attributes Sex (male 1, female 2) −.054 .027 −2.02* 2,740 Age .007 .002 4.43** 2,740 Attractiveness .299 .023 13.20** 2,740 With familymembers −.040 .057 −0.70 2,740

In nature .10 .027 4.16** 2,740 *p < .05, **p < .01. Coefficients were computed using HLM’s restricted maximum likelihood algorithm and are interpreted as average unstandardized beta weights. Random effects for each predictor were assessed one at a time, with all other predictors fixed. Degrees of freedom varied across analyses as a function of the number of predictors in the model, and number of valid cases. For the messenger analyses, video presentations that included both male and female presenters in the same message were dropped resulting reduced degrees of freedom. https://doi.org/10.1371/journal.pone.0331672.t002