How online toxicity silences the majority and fuels power users

In online political discussions, a small fraction of highly active users typically generates the vast majority of content while most people silently observe. A recent experiment reveals that polite and constructive environments encourage silent readers to post their first comment, but the most active users are driven to post more when discussions become toxic. These findings were published in the journal Science Advances.

Online spaces are central to modern political discourse, yet they suffer from a stark participation divide. A vocal minority often dominates the conversation, creating a distorted picture of public opinion. Researchers refer to these highly active individuals as power users, while the vast majority who consume content without posting are known as lurkers.

This uneven participation raises concerns about a spiral of silence. This is a theory where citizens withhold their views if they believe those views are unpopular. In the modern attention economy of social media, extreme or hostile content often gains the most visibility.

If only a specific type of user feels comfortable producing this content, online discussions may look much more polarized than the actual public. A cycle emerges where moderate voices retreat, leaving the platform to those who thrive on conflict.

To understand what drives people into these distinct roles, researchers need to observe communities as they form. Lead author Lisa Oswald, a researcher at the Max Planck Institute for Human Development, alongside colleagues William Small Schulz and Philipp Lorenz-Spreen, designed a study to track exactly how users select into active or silent roles. They also wanted to test whether financial rewards or strict civility rules could level the playing field.

The researchers recruited 520 participants from the United States through advertisements on the social media platform Reddit. After participants completed an initial survey about their political views and online habits, they were randomly assigned to one of six private Reddit communities created specifically for the study.

Over four weeks, the research team prompted the groups with 20 different political issues. They asked participants to discuss topics ranging from immigration and gun control to universal basic income and artificial intelligence.

To test ways to close the participation gap, the researchers applied different conditions to the communities. Two communities served as a control group with standard platform settings. In two other communities, participants received a financial incentive, earning two dollars for every day they wrote a substantive comment.

The final two communities operated under a moderation condition. In these groups, users received frequent reminders to stay civil and respectful alongside automated spam filters.

Participants completed weekly surveys about their experiences and perceptions of the group dynamics. The researchers then categorized the users into two groups. One group consisted of those who never commented, while the other included those who wrote at least one comment.

For the active users, the team tracked the total number of comments they contributed. The results revealed a mirrored psychological mechanism driving both initial participation and heavy posting.

Participants who remained silent were more likely to view the group as polarized and toxic. These users indicated they preferred environments that felt respectful, knowledgeable, and constructive. If a community felt hostile, they chose to simply read the posts and say nothing.

For users who crossed the threshold to write at least one comment, the exact opposite held true. The active users posted more frequently when they perceived the group as polarized, toxic, and lacking in knowledge.

The researchers noted that an abrasive environment suppresses the voices of the silent majority. At the same time, it acts as a catalyst for a less sensitive minority to dominate the conversation. The objective toxicity of the group interacts with a user’s personal sensitivity to hostility to determine their behavior.

The interventions designed to improve the conversation had mixed results. Financial incentives did encourage more people to speak up, narrowing the participation gap between lurkers and power users. Providing a direct reward successfully motivated hesitant users to join the conversation.

The moderation condition failed to change participation levels or improve the quality of the discourse. Reminding users to remain civil did not reduce the participation gap, nor did it lower the toxicity of the comments. The researchers observed that simply asking for polite behavior is unlikely to change the underlying dynamics of online engagement.

The researchers also analyzed how social feedback influenced user behavior. They found that receiving positive votes on comments was associated with increased posting the following day. When participants received negative feedback through downvotes, they were less likely to continue participating.

The experiment also tracked whether engaging in these political discussions changed participants’ minds. Over the four weeks, the research team found no statistically significant differences in political attitude changes between the treatment groups and the control group. Measures of political trust and polarization remained relatively stable regardless of the community rules or incentives.

Descriptively across all groups, participants reported lower levels of political interest and issue knowledge by the end of the study. The researchers suggested this might be due to an intimidation effect. The comments posted during the study were considerably longer and more detailed than typical Reddit posts, which may have led some participants to feel they could not add anything to the conversation.

The study took place in a uniquely controlled environment that differs from organic social media use. The participants were predominantly liberal, meaning the discussions lacked the deep ideological clashes often found in broader public forums. Additionally, the overall toxicity of the comments during the experiment was exceptionally low from the start.

This low baseline of toxicity left little room for the moderation treatment to make a measurable impact. Future research could explore whether strictly enforced rules, such as banning hostile users rather than just reminding them to be civil, might yield different outcomes. Scientists also plan to investigate how these group dynamics function on other platforms with different community structures and algorithms.

The study, “Disentangling participation in online political discussions with a collective field experiment,” was authored by Lisa Oswald, William Small Schulz, and Philipp Lorenz-Spreen.