What Researchers Learned by Reading 700,000 YouTube Comments
You’ve probably been warned not to venture into the comments section on YouTube. But a group of researchers in the United States and Canada threw caution into the wind for the sake of science, looking at a total of 774,939 comments posted on the TEDx and TED-Ed YouTube channels.
In their study, the researchers used a computer program to classify comments based on sentiment. (Word has it that after considering over half a million YouTube comments, the computer program has now turned against humanity.)
Each comment was classified as either clearly positive, clearly negative, or neutral. The researchers then looked for patterns in what types of videos tended to elicit more positive or negative comments.
First, they found that the topic of a video played a role in how likely the video was to be met with positive or negative comments. Some topics, like passion and beauty tended to generate more positive comments. Videos addressing challenging topics like pain and cancer led to more comments with negative sentiment. Perhaps more surprisingly, one of the topics that generated the most negative sentiment was college.
The format of the video played a role too. In particular, comments on animated videos were more neutral, with lower levels of both positive and negative sentiment. After all, it’s just not as easy to get worked up about something said by an animation compared with a real person.
Finally, the researchers looked at whether the gender of the speaker had an effect on the sentiment of the comments? Of course, we all know that men and women are treated the same on the internet, right?
Just kidding. What the researchers found was that, overall, talks by men elicited more neutral comments. By contrast, talks by women received more positive but also more negative comments.
A related question the researchers considered was how to identify comments that were not just negative, but downright offensive. As it turns out, spotting offensive comments is much harder than simply labeling comments as unambiguously negative. The researchers found that filtering out negative comments was too heavy-handed an approach since it removed legitimate comments that were negative but not inflammatory.
One takeaway of this study is that the comment section can be a tricky place. Gender, topic, and video format all play a role in determining what types of comments a video is likely to receive. This information, the authors of the study suggest, is useful for “those who encourage or prepare students and scholars to participate online.”
What are your thoughts on all this? Please leave a comment below. Preferably a positive one.