{"id":132728,"date":"2020-05-01T10:55:52","date_gmt":"2020-05-01T14:55:52","guid":{"rendered":"https:\/\/www.bates.edu\/news\/?p=132728"},"modified":"2020-05-01T11:02:59","modified_gmt":"2020-05-01T15:02:59","slug":"can-peoples-tweets-help-find-coronavirus-outbreaks","status":"publish","type":"post","link":"https:\/\/www.bates.edu\/news\/2020\/05\/01\/can-peoples-tweets-help-find-coronavirus-outbreaks\/","title":{"rendered":"Can people’s tweets help find coronavirus outbreaks?"},"content":{"rendered":"

March 11, 2020, saw a triple whammy of bad news.<\/p>\n

Just during the evening, the NBA suspended its season; President Donald Trump issued an initially confusing order suspending some travel from Europe to the United States; and Tom Hanks announced that he and his wife, Rita Wilson, had contracted the coronavirus.<\/p>\n

COVID-19, of course, was the backdrop to all this breaking news. And as the pandemic gained momentum at an exponential rate through the month of March, the week centered on Wednesday, the 11th, was when many Americans realized how fundamentally the disease would alter their daily lives.<\/p>\n

The next day, Chris Danforth \u201901, an applied mathematics professor at the University of Vermont, and his team saw that their \u201chedonometer,\u201d an instrument that quantifies, in real time, national happiness based on millions of Twitter posts, had recorded a remarkable drop in mood \u2014 a low that lasted for the next several weeks amid the losses of jobs, freedom of movement, and loved ones.<\/p>\n

\u201cIn the entire history of our instrument, over a decade, we\u2019ve never seen an event that effectively persists in our collective mood for more than a day or two,\u201d Danforth says. \u201cSince March 12, the mood has been dramatically depressed on Twitter.\u201d<\/p>\n

\"\"<\/a>

Chris Danforth \u201901 directs the Computational Story Lab at the University of Vermont. (Courtesy of Chris Danforth)<\/p><\/div>\n

Danforth, who directs the Computational Story Lab at Vermont, has spent his career finding out what social media \u2014 on a scale of millions of posts in dozens of languages and countries \u2014 can tell us about human emotion.<\/p>\n

Using a different method from the hedonometer, he and social psychologist Andrew Reece \u201901 have found that people suffering from depression post photos on Instagram whose colors are cooler and darker than those of non-depressed people.<\/p>\n

And, in analyzing current data during the coronavirus pandemic, he\u2019s also seeing evidence that his work analyzing social media might just identify viral outbreaks.<\/p>\n

The hedonometer gauges mood primarily using Twitter, which Danforth says offers the most publicly available posts. Each day the software crunches about 50 million Tweets, 10% of the total number of posts on the site.<\/p>\n

When they first created the hedonometer, Danforth and his team surveyed speakers of languages from English and Spanish to Indonesian and Arabic, asking them to rate thousands of words on a scale of happiest to saddest.<\/p>\n

As a result, the hedonometer can take the words and phrases (and emojis) in each tweet, assign a value to them based on the mood they express, and average them together into a single measure of happiness, on Twitter, for that day.<\/p>\n

\"\"<\/a>

This gargantuan graphic shows the average happiness on Twitter from the launch of the hedonometer in 2009 to the present day (each dot represents one day). Chris Danforth \u201901 says the instrument has recorded more variability in mood since the 2016 presidential election. (Courtesy of Chris Danforth)<\/p><\/div>\n

Hedonometer readings reveal patterns in the national consciousness and in daily life. In normal times in the United States, posts are happiest on the weekends and saddest on Tuesdays, Danforth says. Posts made in urban parks are as happy as those made on Christmas.<\/p>\n

National moods as measured by the hedonometer hew to world events, at least some of the time, Danforth has found. Bad news \u2014 a school shooting, a celebrity death, a natural disaster \u2014 is associated with a drop in happiness, though only for about a day.<\/p>\n

Using other instruments, Danforth\u2019s team can also measure the amount of attention Twitter users pay to particular topics, like the coronavirus.<\/p>\n

From this kind of analysis, it\u2019s clear that attention doesn\u2019t always follow the news, Danforth points out ruefully. Americans on Twitter were talking about the coronavirus a lot in January, when China was taking momentous steps to stop the spread in that country. That\u2019s clear from the number of times Danforth\u2019s instruments detected the word \u201cvirus.\u201d<\/p>\n

\n

\u201cEarly March was a really hard time to be on Twitter when you\u2019re a scientist.\u201d<\/p>\n<\/blockquote>\n

Then, attention \u201cdecayed\u201d for about six weeks, Danforth says \u2014 people weren\u2019t tweeting \u201cvirus\u201d as much. It was during those six weeks that the coronavirus was blooming in the United States, as scientists at Northeastern demonstrated this month.<\/p>\n

Danforth sees the silence on Twitter as a bellwether for how people, from John and Jane Doe up to top federal officials, were failing to notice the scale of the problem.<\/p>\n

\u201cEarly March was a really hard time to be on Twitter when you\u2019re a scientist,\u201d Danforth says. \u201cThe world\u2019s on fire, and nobody\u2019s paying attention.\u201d<\/p>\n

\u201cThat\u2019s what\u2019s so diabolical about this particular virus,\u201d he adds. \u201cIf you have one person testing positive simultaneously to 10,000 people having it, there\u2019s no contact tracing to be done. The cat\u2019s out of the bag.<\/p>\n

\u201cThat six-week period will cost tens of thousands of people their lives because we didn\u2019t respond soon enough. That decay in our attention was visible just in our use of the word \u2018virus.\u2019\u201d<\/p>\n

All this led up to that week of March 11, when the hedonometer measured a national mood that was the lowest in over a decade. And it has persisted \u2014 weeks of just Tuesdays, or a natural disaster every day.<\/p>\n

\"\"<\/a>

This graph of late 2019 and the first months of 2020, each dot representing a day, shows how moods in the United States took an unprecedented nosedive after the coronavirus pandemic hit. (Courtesy of Chris Danforth)<\/p><\/div>\n

\u201cThe Boston Marathon bombing in 2013 was one of the saddest days we\u2019ve experienced,\u201d Danforth says. \u201cOur entire month has been below the day of the Boston Marathon bombing. That\u2019s a pretty remarkable change in the daily content on Twitter.\u201d<\/p>\n

Danforth describes the hedonometer alternately as a \u201cDow Jones\u201d and a \u201cthermometer\u201d for the national mood. But his team\u2019s ability to analyze millions of Twitter posts might have an impact beyond that.<\/p>\n

In identifying words related to coronavirus in various countries, Danforth noticed something curious. Countries in which, at a given point, there was a surge of descriptive words about the pandemic \u2014 words like \u201cpandemic,\u201d \u201ccoronavirus,\u201d \u201cWHO\u201d \u2014 tended to see higher recorded COVID-19 case counts three weeks later.<\/p>\n

Countries that didn\u2019t have any uptick in virus-related language at that time didn\u2019t have a later surge in COVID-19 cases. And once an outbreak had clearly taken hold in a country, people\u2019s language shifted to phrases about responses to it: \u201csocial distancing,\u201d \u201cflatten,\u201d \u201ccurve,\u201d \u201clockdown,\u201d \u201cventilators.\u201d<\/p>\n

\n

\u201cHuman beings are really good at acclimating.\u201d<\/p>\n<\/blockquote>\n

In Italy, that surge in pre-virus tweets came in February. In the United States, in March.<\/p>\n

It\u2019s not clear why this correlation exists \u2014 why a country would be tweeting a lot about the coronavirus before testing makes clear that an outbreak has taken hold, or why a country with lower case counts wouldn\u2019t be tweeting as much. But Danforth thinks that \u201cthis firehose of tweets\u201d can help health officials identify outbreaks as they are happening.<\/p>\n

\u201cIn places where there aren\u2019t tests, you can use it to optimize and prioritize resources,\u201d he says.<\/p>\n

For Danforth, using the hedonometer and other instruments to measure emotion during the pandemic has reinforced one other aspect of human nature. Since moods on Twitter plummeted after March 12, they have been steadily rising again, even though they stayed well below average for weeks.<\/p>\n

As April turns to May, the hedonometer\u2019s readings have crept back up to \u201cnormal.\u201d<\/p>\n

\u201cWho knows what\u2019s going to happen?\u201d Danforth says of the pandemic. But \u201chuman beings are really good at acclimating. The instrument is back to roughly where it was before.\u201d<\/p>\n\n","protected":false},"excerpt":{"rendered":"

Chris Danforth \u201901 of the University of Vermont, who studies emotions through millions of social media posts, noticed a big dip in national mood \u2014 and thinks Twitter can prove useful for public health officials. <\/p>\n","protected":false},"author":1005,"featured_media":132788,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_hide_ai_chatbot":false,"_ai_chatbot_style":"","associated_faculty":[],"_Page_Specific_Css":"","_bates_restrict_mod":false,"_table_of_contents_display":false,"_table_of_contents_location":"","_table_of_contents_disableSticky":false,"_is_featured":false,"footnotes":"","_bates_seo_meta_description":"","_bates_seo_block_robots":false,"_bates_seo_sharing_image_id":0,"_bates_seo_sharing_image_twitter_id":0,"_bates_seo_share_title":"","_bates_seo_canonical_overwrite":"","_bates_seo_twitter_template":""},"categories":[7],"tags":[12135],"class_list":["post-132728","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-alumni","tag-covid-19"],"_links":{"self":[{"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/posts\/132728","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/users\/1005"}],"replies":[{"embeddable":true,"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/comments?post=132728"}],"version-history":[{"count":8,"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/posts\/132728\/revisions"}],"predecessor-version":[{"id":132801,"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/posts\/132728\/revisions\/132801"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/media\/132788"}],"wp:attachment":[{"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/media?parent=132728"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/categories?post=132728"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.bates.edu\/news\/wp-json\/wp\/v2\/tags?post=132728"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}