Testing Hypotheses from the Social Approval Theory of Online Hate: An Analysis of 110 Million Messages from Parler
This study analyzes 110 million Parler messages to demonstrate that social approval (upvotes) predicts online hate and toxicity in complex, time-dependent ways, often suppressing immediate negative behavior while increasing it over longer periods, whereas social disapproval (downvotes) moderates these effects by reducing long-term toxicity.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the internet as a giant, chaotic playground where everyone is shouting, whispering, and posting signs. In this digital world, there's a phenomenon called "online hate," where people say mean, hurtful things about others based on who they are. For a long time, scientists thought these hateful posters were just "bad actors"—people with angry personalities who wanted to hurt others for fun. But a newer idea, called the Social Approval Theory, suggests something different: maybe people aren't just being mean to hurt someone; maybe they are being mean to get a reaction. Think of it like a kid in a schoolyard who starts a rumor not because they hate the victim, but because they want their friends to laugh and cheer. In this theory, the "cheers" are digital signals like upvotes (a thumbs-up) or likes. The big question is: Does getting a "cheer" make someone want to say something even meaner next time? This is a crucial question because if we know why people spread hate, we might be able to design better ways to stop it.
This paper is a massive detective story that tries to answer that question by looking at a mountain of data. The researchers, David Markowitz and Samuel Hardman Taylor, decided to investigate a specific social media site called Parler, which was popular between 2018 and 2021. They analyzed a staggering 110 million messages from this platform, which was known for having very few rules about what people could say. They wanted to see if getting upvotes on a hateful message actually made people post more hate later, or if it made them stop.
Here is what they found, and it's a bit more complicated than a simple "yes" or "no."
First, they looked at the timing. If you get an upvote on a hateful post, does it make you post hate immediately? Surprisingly, the answer was no. In fact, right after getting an upvote, people were actually less likely to post another hateful message in their very next turn. It's as if the "cheer" satisfied them for a moment, and they took a break. However, the story changes if you wait a bit. When the researchers looked at what happened over the course of a week, the upvotes started to work like a magnet. People who got upvotes on their hate posts were more likely to post more hate in the following week. It seems the "reward" of the upvote didn't kick in instantly; it took a little time to accumulate in their minds before it made them want to do it again.
But the type of message mattered a lot. The researchers split the data into two categories: Posts (messages broadcast to everyone's feed, like a billboard) and Comments (messages written inside a conversation thread).
- For Posts: The upvotes made people post more hate in the next week, but this effect faded away after a month.
- For Comments: The pattern was even stranger. Upvotes on a hateful comment made people post less hate in their very next comment, but then more hate in the next week. However, if you looked at the long term (three to six months), the immediate effect flipped to predicting less hate. Crucially, though, when looking at the big picture across all users, those who received more upvotes on their hate comments overall tended to produce more hate in the long run, regardless of the time interval. It's like the upvotes gave a short burst of energy to be mean, but the people who consistently got the most approval were the ones who kept posting hate the most over time.
The researchers also looked at toxicity, which is a measure of how rude or disrespectful a message is (not just hate, but just being mean). They found that for comments, upvotes made people more toxic in the short term (the next week), but this didn't happen for posts.
Finally, they asked: What happens if you get both a "cheer" (upvote) and a "boo" (downvote)? The Social Approval Theory suggested that even if you get booed, if your friends cheer for you, you'll keep going. But the data showed something else. When people got downvotes along with their upvotes, the "cheer" didn't work as well. The downvotes seemed to cancel out the motivation to be mean. It's like if you tell a joke, and half the room laughs but the other half groans; you probably won't tell another joke right away. The study suggests that while upvotes can fuel hate in the short term, downvotes can actually dampen that fire, making the connection between getting approval and posting hate weaker.
So, what's the takeaway? The idea that "people just want to be mean to hurt others" isn't the whole story. The study suggests that social approval does drive online hate, but it's a slow-burning fire, not an instant explosion. It takes time for the "cheers" to pile up and make someone want to say something nasty again. And interestingly, getting a mix of cheers and boos might actually help stop the cycle, at least for a while. The researchers are careful to say this is based on what they measured in this specific dataset, and they suggest that more research is needed to understand exactly how these feelings build up over time. But one thing is clear: on the internet, the way we react to each other—whether with a thumbs-up or a thumbs-down—shapes what gets said next.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.