YouTubers love wildlife, but commenters aren't calling for conservation action

(Press-News.org) YouTube is a great place to find all sorts of wildlife content. It is not, however, a good place to find viewers encouraging each other to preserve that wildlife, according to new research led by the University of Michigan.

Out of nearly 25,000 comments posted to more than 1,750 wildlife YouTube videos, just 2% featured a call to action that would help conservation efforts, according to a new study published in the journal Communications Sustainability.

"Our results basically show that people like to watch videos of zoos and safaris and that they appreciate the aesthetics and majesty of certain animals," said author Derek Van Berkel, associate professor at the U-M School for Environment and Sustainability, or SEAS. "But there really wasn't much of a nuanced conversation about conservation." 

Although he didn't expect to see most commenters urging other YouTube users to call their elected officials or to support conservation groups, "I was hoping there might be more," Van Berkel said. "I thought it might be bigger than 2%."

Despite the low number, however, the team believes the report still has an optimistic take-home message.

"The flip side of this is we can and should do better at messaging, and there's a huge potential to do so," said study co-author Neil Carter, associate professor at SEAS. 

While individual YouTube viewers weren't organically calling for conservation action, there was also a notable absence of conservation groups and influencers working to start conversations and sharing actionable information in the comments.

"There's tremendous untapped potential for conservation messaging to be improved," Carter said.

Unlike many other social media platforms, YouTube provided sufficiently accessible, detailed and structured data to provide insights into the digital culture around wildlife conservation, Van Berkel said. And the data was just the starting point.

YouTube's 8M dataset contained information for nearly 4,000 videos that had been classified as wildlife. The researchers trimmed the list by more than half by selecting videos that featured at least one English language comment and that they could categorize into one of seven topic areas. Those included footage from zoos, safaris and hunting. 

The next step was characterizing the comments by the attitudes they expressed. The team arrived at five different categories for these. Expressions of appreciation and concern, both for wildlife and humans, made up four of the categories. The fifth was calls to action.

With the categories and the criteria for each defined, the team created a "gold set" of comment attitudes from 2,778 comments assigned by hand. The researchers then used this data to train a machine learning model to assess more than 20,000 additional comments.

Those steps were painstaking and labor intensive—the team hired additional participants to crowdsource the construction of the comment attitude gold set. But one of the biggest challenges was training the machine learning algorithm on what calls to action looked like when there were so few to begin with, said co-author Sabina Tomkins, assistant professor at the U-M School of Information.

"If the label you're looking for happens far less often than the others, that problem is really hard. You're looking for a needle in a haystack," she said. "The way we solved that challenge was by looking at the models very carefully, figuring out what they were doing."

Tomkins said the effort from the School of Information graduate students who were part of the research team—Sally Yin, Hongfei Mei, Yifei Zhang and Nilay Gautam—was a driving force behind the project. Enrico Di Minin, a professor at the University of Helsinki, also contributed to the work, which was funded in part by the European Union.

 

END