Affective Language Processing: Language-driven Evaluation of Character Affect in Morally Loaded Narratives

Abstract

Research has shown that the valence of emotional linguistic stimuli is reflected in corrugator supercilii (‘frowning muscle’) activation, which can be measured using facial electromyography. But in a larger context, the corrugator response may be additionally affected by emotional evaluation, i.e., how we feel about some emotion. An earlier study from 2019 found that descriptions of moral characters experiencing negative emotions elicited substantially more corrugator activity than positive emotions, thus reflecting valence, but they found no effect of valence for adjectives describing immoral characters. The authors therefore suggested that both mental simulation of word valence and moral evaluation may drive corrugator activation, e.g., mentally simulating ‘angry’ activates the corrugator and ‘happy’ relaxes it, whereas evaluation may elicit relaxation for both moral-positive and immoral-negative stories. This is based on the idea that good people deserve good things and bad people deserve bad things. More research was needed to corroborate the existence of these two effects opposing each other. This current study therefore used the same narratives but with an additional task that requested participants to explicitly judge the behavior with a rating task. This was hypothesized to increase their tendency to evaluate what they read and subsequently change the net result that the postulated two effects should have. Critically, we expected more frowning for bad people experiencing positive emotions than negative emotions since bad people feeling good should be seen as very unfair. However, this is not what we found, and our study in fact replicated the findings from 2019: a null result of valence for adjectives describing immoral characters and a large valence-based effect for adjectives describing moral characters. Surprisingly, we did find the expected evaluation-based pattern in responses on an exploratory task where we asked participants to rate the fairness of every narratives’ ending. Finally, we found preliminary evidence that a person’s attitude towards justice, i.e., their preference or desire for morally good people being rewarded and morally bad people being punished, as a personality trait may modulate the effect that morality and valence have on their corrugator responses.