Mapping the Spiral of Silence: Surveying Unspoken Opinions in Online Communities

TL;DR

This paper investigates how spiral of silence operates in online communities, arguing that online data can distort public opinion due to self-censorship by individuals who perceive themselves as in the minority. It introduces a human-plus-AI workflow to generate contextually relevant controversial topics for specific communities using LLM prompts, and then measures willingness to express opinions via a survey of Reddit users analyzed with linear mixed-effects models. The results show that 72.6% of minority-perceiving individuals remain silent and that minority opinions are posted at roughly half the rate of majority opinions, with perceived inclusivity mitigating this effect and increasing minority expression by up to about 2.3x under certain conditions. These findings highlight distortions in social-media-based inferences of public opinion and offer practical implications for moderators and platform designers aiming to foster more inclusive discussions and better representation of minority views.

Abstract

We often treat social media as a lens onto society. How might that lens be distorting the actual popularity of political and social viewpoints? In this paper, we examine the difference between the viewpoints publicly posted in a community and the privately surveyed viewpoints of community members, contributing a measurement of a theory called the "spiral of silence." This theory observes that people are less likely to voice their opinion when they believe they are in the minority--leading to a spiral where minority opinions are less likely to be shared, so they appear even further in the minority, and become even less likely to be shared. We surveyed active members of politically oriented Reddit communities to gauge their willingness to post on contentious topics, yielding 627 responses from 108 participants about 11 topics and 33 subreddits. We find that 72.6% of participants who perceive themselves in the minority remain silent, and are only half as likely to post their viewpoint compared to those who believe their opinion is in the majority. Communities perceived as being more inclusive reduce the magnitude of this effect. These results emphasize how far out of step the opinions we see online may be with the population they purport to represent.

Figures (3)

• Figure 1: We generate controversial viewpoints specific to a provided community. Using the community name, description, and rules, we prompt an LLM to propose controversial topics. For each topic, we enumerate statements representing different stances on the topic. Finally, the generated topics and viewpoints are validated by community members to ensure they are relevant and fall within the norms of what is accepted in the community.
• Figure 2: Across all topics, the average percentage of incongruent viewpoints likely to be shared is $28.3\%$. On the left, we report the percentage of incongruent viewpoints willing to be shared (i.e., Share Likelihood$>$ 4) out of all incongruent viewpoints for the given topic. On the right, we do the same for congruent viewpoints.
• Figure 3: An increase in community inclusion has a greater positive impact on the likelihood of sharing incongruent viewpoints compared to congruent ones. We visualize the marginal effects of community inclusion on opinion expression for incongruent and congruent viewpoints. For every one standard-deviation unit increase in perceived inclusion, the likelihood of sharing increases by $0.63$ for incongruent viewpoints and $0.28$ for congruent viewpoints.