Map of emotions on social media in the U.S. and UK on April 24. Credit: Expert System and Sociometrica, April 24, 2020
An Italian-based artificial intelligence company is regularly analyzing social media posts about the coronavirus for their emotional content.
Why it matters: Classifying tens of thousands of posts by their emotional tone provides a snapshot of how people feel about the pandemic. Spoiler alert: not great!
How it works: Expert System specializes in semantics and natural language reading, a branch of AI involving computer systems that attempt to make sense of written language.
- In doing so, a computer can rapidly analyze vast amounts of the written word — like, for example, a day's worth of social media posts about the COVID-19 pandemic.
For the past few weeks, Expert System has been collecting English language social media posts each day that feature frequently used hashtags like #coronalockdown and #covid19. Its AI can extract the emotional content of those posts, which is then analyzed and interpreted by Sociometrica.
- On April 24, "fear" had become the single most widespread emotion, displacing "sadness."
But, but, but: Such negative feelings have been declining over the past 10 days, from 62.4% to 45.5%. At the same time, neutral and positive feelings are on the rise, with particular growth around posts showing "hope."
- Also increasing in intensity is "health fanaticism," which Expert System defines as "a feeling of fear and anxiety around certain aspects of health and an emphasis on defending the health of one’s own body."
The bottom line: A natural language AI can tell you what you probably already know: the pandemic is terrible, but if you squint hard enough, there's light at the end of the tunnel.
Go deeper: Isolation can be bad for mental health