By Ephrat Livni
Before Facebook and Twitter existed to explain us to ourselves, we knew the weather made us moody because we were gloomy on rainy days and cheery when the sun shone. Until social media was invented, however, there was no scientific way to measure the connection between moods and meteorology on a large scale, to see if that link manifests in our expressions, or measure the effects of the weather compared to other big events.
Now, there’s already a rich and (relatively) long record of human expression in the form of posts we wrote, and these can be associated with locations and dates, for which there are also meteorological records. These data allowed environmental economist Patrick Baylis, a postdoctoral scholar at Stanford University, to lead what his team of researchers says is the largest-ever investigation into the relationship between meteorological conditions and the sentiments people express. (The paper, covered in MIT Technology Review, is prepress and has not yet been peer reviewed).
The team sought to answer three questions:
Is there a link between weather conditions and sentiments in human expressions?
If so, is that connection robust enough to manifest even when excluding discussions about the weather itself?
Finally, how significantly does weather influence sentiment compared to other events that alter expressions, like natural disasters or major breaking news?
To answer these questions, the research team algorithmically analyzed 3.5 billion tweets and Facebook posts written in the US between 2009 and 2016, seeking expressions of sentiment both directly and indirectly related to the weather. In other words, they didn’t just look at posts about a snowstorm, say, but also analyzed what people wrote when they weren’t explicitly considering climatic conditions, in order to see how weather might be inadvertently affecting what people wrote.
Researchers determined whether people were expressing positivity or negativity using a sentiment-classification tool called the Linguistic Inquiry Word Count, commonly used by social scientists to assess the feelings behind writing in social media posts. Those feelings were noted, and the collected posts were associated with specific times, places, and climatic conditions.
Meteorological data considered included daily maximum temperature, temperature range, precipitation measures, cloud cover, and relative-humidity statistics, all measures considered central to the relationship between weather and emotional states in previous, smaller studies. The researchers write that they were able to find “substantial evidence that less ideal weather conditions relate to worsened sentiment.”
Specifically, they noted that cold temperatures, hot temperatures, precipitation, narrower daily temperature ranges offering little shift from steadily unpleasant conditions, humidity, and cloud cover were all associated with worsened expressions of sentiment, even when excluding weather-related posts.
The weather’s a real big deal to us, it seems. Freezing conditions influenced people’s negative expressions as much as earthquakes and terror attacks, the researchers found. To the extent that written expressions serve as a valid proxy for underlying emotions, they now believe there is evidence “that the weather may functionally alter human emotional states.”