Twitter isn't impressed with what we're doing about climate change

The Wire
May 26, 2015

English is a happy language. At least the 10,000 most used English words are positively biased, according to a 2012 study conducted by mathematicians from the University of Vermont. To reach their conclusion, they used a tool they’d built called the hedonometer – an evaluator of happiness in natural (English) language.

The hedonometer works by assessing how happy or unhappy a particular corpus of text is based on the words used in the corpus. The happiness score of each word is set before the assessment begins: in the case of the English language, 50 participants drawn from Amazon’s Mechanical Turk marketplace rated each word on a scale of 1 (least happy) to 9 (most happy). Then, the hedonometer tracked how much they were used and in the presence of which other words.

Since its building in 2011, the tool has also been used to identify individual happiness expressed in literary corpora as well as which the happiest cities are.

Now, a part of the Vermont group, together with a researcher from the University of Adelaide, has put the hedonometer together with one other – and far more brazen – gauge of human opinions and sentiments, to understand how the Twitterverse might feel about climate change. Specifically, they collated 1.5 million tweets that mentioned “climate” between September 14, 2008, and July 14, 2014, and adjudged their collective happiness against the happiness of some 100 billion general tweets from the same period.

Although the mathematicians found that not all tweets that mentioned “climate” at least once were about climate change (being about the social climate, for example), they also calculated that their exclusion didn’t change the happiness index of the overall set. However that was in hindsight.

Their results, submitted to the arXiv pre-print server on May 14, 2015, are accompanied by three telling charts.

1. Sadder than happier


The average happiness index of the 100 billion tweets logged in the six years was 5.99, its daily variation indicated by a red dotted line in the chart. And the average happiness index of tweets about the climate was almost always below this line (ending up with an average of 5.84).

Though there were occasional outliers – in the form of spikes – there were also more negative outliers than positive ones. So, people on Twitter aren’t happy about the climate.

The mathematicians note that “the week of October 28, 2012 appears as one of the saddest weeks of climate discussion on Twitter” as this was the week “when Hurricane Sandy made landfall on the east coast of the US”.

2. “Science” v. “denial”


Relative to the set of 100 billion unfiltered tweets, which words made tweets mentioning “climate” happier or sadder? The mathematicians built a so-called word-shift graph, which ranks words by their influence over tweets’ sentiments.

The words on the left contributed to sadness and those on the right, to happiness. Moreover, yellow bars indicate happy words, purple bars indicate sad words, and the up/down arrows indicate if they were used more or less frequently.

So, it’s reassuring to see the increased use of “science” contributed to making climate-related tweets happier. At the same time, the words whose increased use made tweets sadder were “threat”, “pollution”, “denial” and, toward the bottom, “poverty”.

The nature of denial of climate change has focused not on refuting the warming of Earth itself but on the aspect of humans having caused it. Thus, if climate-deniers are persisting on Twitter, it’s not clear how they could’ve contributed to the sadness. On the other hand, “denial” and “denying” featuring on the sad side of things signals that the persistence of denialism is a cause of distress (which you could say is a kind of reassurance).

Aside: the more frequent use of “hell” saddened tweets more than the more frequent use of “heaven” made them happier.

3. Three happiest days


On April 30, 2012, April 9, 2009, and December 28, 2008, the happiness indices were the highest for tweets mentioning “climate”, clocking 6.36, 6.27 and 6.27 respectively. What happened on these dates? From the paper…

  1. April 30, 2012 – “On this date, Twitter users were reaching out to Brazilian Dilma [Rousseff] to save the Amazon rainforest”
  2. April 9, 2009 – “Twitter users were discussing the release of a new book called Sustainable Energy Without the Hot Air by David JC MacKay. Also on this date, users were posting about a Climate Prize given to a solar-powered cooker in a contest for green ideas”
  3. December 28, 2008 – “This is due in part to a decrease in the word “no”, and an increase in the words “united”, “play”, and “hopes”. On this day, there were “high hopes” for the US response to climate change”

Similarly, the three saddest days were…


  1. October 9, 2008 (5.29) – “Topics of conversation in tweets containing “climate” include the threat posed by climate change to a tropical species, a British climate bill, and the US economic crisis”
  2. April 4, 2010 (5.37) – “Popular topics of conversation on this date included a California climate law and President Obama’s oil-drilling plan”
  3. August 6, 2011 (5.38) – “A topic of conversation on this date was the Keystone XL pipeline, a proposed extension to the current Keystone Pipeline”

Blurry reflections

Even if the averaged outlook seems gloomy, there exists a demonstrable appreciation of positive action – even if the action is small. This is evident especially when Twitter users caused a spike in happiness on April 9, 2009, tweeting about the solar-powered cooker (the Kyoto Box, made from cardboard and able to boil or bake food).

Another example is the Forward on Climate Rally, which brought together nearly 50,000 people in Washington DC on February 17, 2013, to call on President Barack Obama to veto the Keystone pipeline bill. According to the authors, on the day, “the happiness of climate tweets increased slightly above the unfiltered tweets during this event, which only occurs on 8% of days.” The bill was eventually vetoed.

However, the paper isn’t perfect. There are two factors skewing results – both having to do with locations. First: In the six years during which the mathematicians parsed the tweets, the volume ballooned from 1 million tweets/day to 500 million tweets/day. The underlying expansion in the user base will have been accompanied by a shift in the demographics as well, including more people who are less likely to have tweeted about the climate than others. In the absence of user demographics, thus, the more accurate weighted average of the happiness index remains out of reach.

Similarly, the second factor is that results largely concern American sentiments. The mathematicians don’t mention that they’ve drawn tweets put out by users located in the United States, yet the outcomes and the saddest and happiest dates are dependent on American opinions. This makes a weighted average of the happiness index that accounts for country-wise differences all the more meaningful.

Despite these pitfalls, the paper provides a blurry reflection of the popular perceptions of climate change. Given the affinity for positive action, in fact, the conclusions provide ample reason to believe people on Twitter don’t think enough is being done to tackle climate change. The period of 2008 to 2014, when the tweets were tracked, was also the time in which faith in the Kyoto Protocol and its subsequent amendments declined, and the disappointing Copenhagen summit happened in 2009.