Polarization not tied to Reddit use but an influx of far-right users that drove the shift
Researchers noted polarization trends in 2016 were remarkably synchronized across groups...the intense increase in polarization in 2016 was disproportionately driven by new and newly political users.
Reddit polarized before the 2016 US presidential election
Reddit underwent a polarizing event that surrounded the 2016 US presidential election. One theory explaining the hyperpartisan state in 2016 was the effect of long-term internet use. These findings quash the idea that Reddit use itself was to blame, with remarkably robust results.
Researchers analyzed 5.1 billion comments by 34.7 users on Reddit over 14 years.1 This included over 10,000 communities or "subreddits" as they are called and an examination of a commenter's age, gender, and US political partisanship via associations.2
The graph shows what the authors detailed in their report: “…during 2016 every active cohort polarized at the same time. The month-to-month polarization trends in 2016 were remarkably synchronized across cohorts…the intense increase in polarization in 2016 was disproportionately driven by new and newly political users.”
New users were the primary drivers of polarization
New users were the primary drivers of the dramatic change in polarization, which suggests long-term Reddit use can't explain it. The sharp increase in polarization was also driven nearly exclusively by right-wing activity.
The authors state, “Examining polarization over time separately for left-wing and right-wing communities reveals a stark ideological asymmetry. Activity on the right was substantially more polarized than activity on the left in every month between 2012–2018.” That can be seen in the graph below.
Extreme views are more common on the right and that may explain some of the left versus right differences
Looking at the partisan score breakdown of users, we see double the number of users on the far-right as we do on the far-left. The opposite happens with moderate groups. Left-leaning moderates were over four times as numerous as the right-leaning moderates. Right-leaning moderates were such a small group that the far-left was larger, which is noteworthy because the far-left was only half the size of the far-right.
While the average user was center-left, the significant number of far-right users compared to moderate-right may mean the contributions of these right-leaning users draw more attention because the average right-leaning user was more extreme than the average left-leaning user.
The moderate right did not outnumber the extreme right in this study.
The moderate left is many times larger than its most extreme faction.
Extreme views make us more vulnerable to misinformation. This study shows far more users on the extreme right than left. Extreme political views increase how likely someone is to see or share misinformation, this may explain the right-leaning skew in misinformation.
Conservatives might not be more susceptible because of conservative views but because of how extreme those conservative views are. Considering there are substantially more extreme users on the right versus moderate right on Reddit,3 right-leaning users would then, on average, be more susceptible to misinformation. This difference in the proportion of extreme partisans might explain some of the studies that find conservatives are more likely to see or share misinformation.
This study may mean dramatic social change can be induced
Perhaps the most consequential find by the study authors relates to how polarization occurred. Researchers asked whether it was old users polarizing over time or new users who were extreme when they arrived.
The research seems to show the latter with new far-right users driving polarization on the platform. More research should be done, but conceivably this study shows how a dramatic social change could be induced in a short time frame by introducing a large number of extreme actors.
Waller, I., Anderson, A. Quantifying social organization and political polarization in online platforms. Nature (2021). https://doi.org/10.1038/s41586-021-04167-x
Distributions of communities along the age, gender, and partisan dimensions, grouped into behavioral clusters found by hierarchical clustering. The x-axis represents community scores transformed into percentiles (for example, a community with an age score greater than 76% of other communities would be positioned at percentile 76), and the color corresponds to a z-score. As a result, the distribution for all communities (top row) is simply the uniform distribution U(0,100), while the distributions for individual clusters illustrate which percentiles are over-or under-represented within the cluster. Raw score (non-percentile) distributions are shown in Extended Data Fig. 4. The dashed line indicates the 50th percentile. Rows annotated with † comprise two or more clusters (Methods, ‘Creating the community embedding’). Source data.