Visual art inspired by climate change—An analysis of audience reactions to 37 artworks presented during 21st UN climate summit in Paris

stronger the reflections on the artwork, the more support for climate policy is reported. Men reported less reflection about the artwork and lower perceived quality of the artworks, whereas environmental attitudes were strongly and positively related to support for climate policies, but also to positive emotions. Does the relation between reflection on the artwork and support for climate policy depend on a person’s environmental attitude? To test the assumption that the induced reflection about the artwork might have a stronger relation to policy support depending on differences in environmental attitudes, we tested a model where support for climate policies was regressed on reflection about the artwork, envi- ronmental attitudes, and the interaction term between the two (controlled for gender). For the calculation of the interaction term, the variables were grand mean-centered first to avoid false multicollinearity. Fig 3 displays the results of this analysis. The exploration of the interaction term indicated a negative interaction between the two variables, showing that the relation between reflections on the artwork and support for climate policies is stronger for people with weaker environmental attitudes than for people with stronger attitudes. See S4 Table for all model parameters. Do different artworks relate to different states in the psychological variables?–A two-level model In the next step, the same model as in Fig 2 was specified, but this time as a two-level model. Explicitly modelling level 2 variance in dependent variables in the model on the artwork level

(which means for all people answering at the same artwork). This allowed examination of whether different artworks elicited systematically different responses among their visitors. For all variables but the support for climate policies, there was meaningful variance among art- works. Then a simplified model was constructed where only the relevant variables on the per- son level were included and allowed to correlate. This was done to reduce model complexity for the following analyses. Afterward, we examined the artwork features that related signifi- cantly to the five psychological variables by testing all recorded features stepwise. The model presented in Fig 4 is the result of this endeavor. The first result that can be extracted from Fig 4 is that the emotional responses have the most variation between the artworks (the Intra Class Correlations, ICC, indicate that 23% of the variance in positive emotions and 20% of the variance in negative emotions are between artworks). For the other two variables, the variance between artworks is lower. The variance between artworks in positive emotions can almost completely be explained by them being colorful or black/white (black and white artworks were related to less positive emo- tions), if they made use of the peak shift principle (if yes, less positive emotions were reported), if something personal to identify with was displayed (which related to more positive emo- tions), if they were large in size (which was negatively related to the amount of positive emo-

tions), if they depicted known objects (which was also related to less positive emotions), if they presented information about the artwork (which related to more positive emotions), and finally if they depicted destruction (not surprisingly related to less positive emotions). The variation between artworks in negative emotions triggered can be explained by their size (the bigger, the more negative emotions), if they were polarizing in their design (if yes, then negative emotions were more likely), and if they triggered perceptual problem solving (which also enhanced negative emotional reactions). PLOS ONE Visual art inspired by climate change — An analysis of audience reactions PLOS ONE | https://doi.org/10.1371/journal.pone.0247331 February 19, 2021 11 / 18