Study finds big crowds hurt live-stream engagement

Most content platforms thrive on user engagement, but a professor at the University of Miami Patti and Allan Herbert Business School has discovered that too much of it can be harmful.

A paper co-authored by the associate dean for research Kevin Hong indicates that at synchronous content platforms such as Twitch and YouTube Live, large group sizes tend to inhibit participation, not enhance it.

“When a lot of people are participating, you frequently have lots of people talking about different things,” said Hong, a Centennial Endowed Chair and professor of business technology. “When that happens, it looks and feels erratic and people disengage from the conversation.

“There is a limit when everyone is trying to shout in the channel,” Hong noted. “It’s like someone is taking an online class, and every student is trying to ask a question and shout at the teacher in the Zoom chat. It’s probably not going to work very well!”

Hong’s paper, “Lost in the crowd: How group size and content moderation shape user engagement in live streaming,” draws on data from 7,074 Twitch playbacks and their associated chat histories, which took years to complete.  The research was recently published in Information Systems Research, a quarterly, peer-reviewed academic journal focused on the management of emerging information technology.

Lost in the Crowd was co-authored with researchers from Pennsylvania State University, University of South Florida, University of Illinois at Chicago, and Ohio State University. The paper posits that, along with creating cognitive bedlam, large live-stream crowds often spawn an emotional polarity that further diminishes the desire of participants to continue. 

“Previous research had focused on asynchronous platforms like Yelp, Reddit, Wikipedia,” Hong said. “What’s quintessential about these asynchronous platforms is that one person’s participation or consumption does not really affect other people’s consumption. And prior research has consistently found that large group sizes are associated with a higher level of engagement.

“Researchers have been studying asynchronous content platforms for the last 20 years,” Hong continued. “The synchronous platforms have only started to emerge in the past few years, so there’s limited understanding of that. Obviously, it doesn’t seem like group size is going to have a positive effect here—it can be negative.

“How do we manage that negative effect from large group sizes?”                        

What seems promising are bots, particularly those leveraging artificial intelligence. AI bots are already used as moderators by some live-streaming platforms. 

“Human moderators are useful at managing smaller surges in group sizes, but they can’t deal with 500 people calling each other names and talking at the same time,” Hong said. “The bot moderators, even though their functions are limited at this moment, are getting smarter every day. And what the AI bots really excel at is scaling fast and keeping costs low."

“So, there’s a lot of potential for AI agents—AI-powered bots—to sort of moderate those conversations and make them more coherent.”