Divisive negative discourse biases social experience: a live experiment at a massive public event

Abstract

Linguistic choices, crucially including negatively valenced words and divisive messages, can bias people's feelings, thoughts, and judgments. However, these phenomena have been typically captured with small groups in controlled settings, casting doubt on their robustness and ecological validity. Here we examined whether such effects hold in a massive public gathering. During a large TEDx event (n = 3139), participants engaged in an interactive musical game and then evaluated their perception of (active and vicarious) enjoyment and (ingroup and outgroup) performance through surveys that manipulated (a) the initial framing ('divisive' or 'communal') and (b) the questions' valence ('positive', 'neutral', 'negative'). Results showed that negatively valenced words reduced enjoyment and performance ratings, particularly under divisive framings. Active enjoyment also decreased under communal framings. These results were corroborated upon adjusting for sociodemographic variables. Briefly, linguistic manipulations of affect immediately altered a crowd's perception of enjoyment and performance. These insights extend psycholinguistic models and contribute to discussions on public communication.

Keywords: Language and linguistics; Psychology.

Fig. 1. Study design.

The experiment began with A the crowd's arrival at the rock arena, and B the handing of sealed surveys to each attendee. Then, (C) participants were divided into two groups to partake in a musical activity requiring them to D perform a famous passage from either “Hey, Jude” or “We Will Rock You”. Next, E participants completed a survey tapping on active and vicarious enjoyment as well as ingroup and outgroup performance. Different versions of the survey were distributed, varying in their introductory framing (‘divisive’ vs. ‘communal’) and questions’ valence (‘negative’ vs. ‘neutral’ vs. ‘positive’). Finally, surveys were F collected, G preprocessed, and H analyzed via 2×3 ANOVAs and group-wide regressions (adjusting for sociodemographic factors).

Fig. 2. ANOVA results.

Significant group-level results based on two-way ANOVAs, considering framing (divisive, communal) and word valence (negative, neutral, positive). Participants reported A lower active enjoyment when faced with valenced wordings, B lower vicarious enjoyment when faced with negative words under a divisive framing, C lower ratings of their own group’s performance when faced with negative words, irrespective of framing, and D lower ratings of the opposing group’s performance when faced with negative words under the divisive framing. Filled markers indicate divisive framing, while open markers indicate communal framing. Brackets indicate significant pairwise differences in interaction effects.