Majority of social media posts about the metaverse show positive sentiment

A team of researchers from the United Arab Emirates recently conducted a study to discover what people thought about the metaverse on social media.

The team curated a dataset of 86,565 posts on the X social media platform (formerly Twitter) and used machine learning to divide them into three categories representing sentiment: positive, neutral and negative.

According to the researchers, the distribution showed that posts expressing positive sentiments about the metaverse took up the lion’s share of the data with 45,506 (53% of the total). The researchers identified 28,663 posts displaying neutral sentiment, making up 33% of the total, and the final 14% (12,396 posts) showed negative sentiment.

An example of a positive post (called “tweets” in the paper) given in the paper was “good morning everyonenlet (sic) keep building the metaverse.” At the other end of the spectrum, an example given of a post labelled as negative was “correct me if im wrong but isnt the metaverse just going to be like an mmorpg where you have to do your job surrounded by irritating npcs if so ill stick with irl interactions please.”

The team states, in their paper, that positive and negative sentiments followed patterns, but neutral sentiment didn’t:

“The positive tweets contain a lot of words supporting metaverse adoption including ‘need’, ‘love’, ‘right, ‘future’, and ‘new’. In contrast, the negative tweets besides containing words such as ‘bad’, ‘crazy’, and ‘don’t’, also contained many offensive words. No distinct pattern of words is evident in neutral sentiment tweets.”

They also write that this “offers early promising signs for the adoption of metaverse technology.”

Image source: Hayawi, et. al., 2024.

Previous studies using machine learning to deduce metaverse sentiment from social media posts have reached around 88% accuracy in benchmarks — meaning more than 1 in 10 posts are mislabeled or misinterpreted by the models.

The U.A.E. team’s models reached as high as 92.6% accuracy across a dataset containing more than 85K posts, making this study one of the most detailed analyses of public sentiment for metaverse to date. It bears mention, however, that this is preprint research and as such may not have been through the peer review process.

Related: Scientists use WiFi signals to track human movement for the metaverse

Future research will focus on other social media platforms, including Meta and Reddit, as well as integrating offline and traditional sentiment analysis such as scientific surveys to create a more comprehensive picture.

The researchers also admit the current research is limited by the terminology used to conduct the study, namely that the scope of their work was limited only to posts containing the word “metaverse” and, as such, social media discussions about the topic but not containing the keyword were omitted from consideration.