Source-linked AI summary
Climate change sentiment on Twitter: An unsolicited public opinion poll
Emily M. Cody, Andrew J. Reagan, Lewis Mitchell, Peter Sheridan Dodds, Christopher M. Danforth
TL;DR
Climate change information is discussed across media, but newspaper coverage can be biased and scientific findings can be difficult for the public to understand. This paper analyzes climate tweets with the Hedonometer and finds that their sentiment varies with climate events and news, while the discussion appears dominated by activists rather than deniers.
Problem
Newspaper coverage may dismiss or sensationalize climate change, while scientific findings can be inaccessible to the general public, motivating analysis of social media discussion.
Method
The study examines Twitter messages containing “climate” and analyzes word frequencies and sentiment patterns around climate news, events, and natural disasters.
Results
Tweets containing “climate” are less happy than all tweets, and their discussion appears dominated by climate change activists rather than climate change deniers.
Takeaways & Limitations
Twitter provides a resource for examining public attention and sentiment surrounding climate change news, events, and natural disasters.
Takeaways & Limitations
Because the analysis includes only tweets containing “climate”, it may miss climate-related events and background discussion not linked to that word.
Abstract
from arXiv · showhide
The consequences of anthropogenic climate change are extensively debated through scientific papers, newspaper articles, and blogs. Newspaper articles may lack accuracy, while the severity of findings in scientific papers may be too opaque for the public to understand. Social media, however, is a forum where individuals of diverse backgrounds can share their thoughts and opinions. As consumption shifts from old media to new, Twitter has become a valuable resource for analyzing current events and headline news. In this research, we analyze tweets containing the word "climate" collected between September 2008 and July 2014. Through use of a previously developed sentiment measurement tool called the Hedonometer, we determine how collective sentiment varies in response to climate change news, events, and natural disasters. We find that natural disasters, climate bills, and oil-drilling can contribute to a decrease in happiness while climate rallies, a book release, and a green ideas contest can contribute to an increase in happiness. Words uncovered by our analysis suggest that responses to climate change news are predominately from climate change activists rather than climate change deniers, indicating that Twitter is a valuable resource for the spread of climate change awareness.
INTRODUCTION
Climate change information is widely discussed, but traditional media can distort findings or make scientific evidence difficult for the public to understand. Twitter offers a many-to-many forum for diverse users to communicate opinions and provides an underanalyzed source for studying climate discussions.
- Motivation: Traditional media may sensationalize or dismiss climate change, while scientific papers can be difficult for the general public to understand.Journalistic personalization, dramatization, and novelty biases are identified as possible sources of distortion; scientific communication is constrained by paywalls, formal language, and technical results.
- Social media: Social media shifts communication from one-to-many broadcasting toward many-to-many exchange among people with diverse backgrounds.This format allows users to communicate and form their own opinions, while exposing climate change issues as potential public concerns.
- Social media: Twitter has supported analyses of social phenomena, public health, disasters, and sentiment, making it a useful setting for climate-opinion research.Prior applications include earthquake reporting, influenza detection, Hurricane Sandy activity, disaster-related interaction changes, and sentiment analysis.
- Study focus: The study applies the Hedonometer to tweets containing “climate” to examine climate-change opinions and sentiment over time.The authors collected roughly 1.5 million tweets through Twitter’s gardenhose API and compare climate-related discussions with broader Twitter activity.
METHODS
The study measures happiness in climate-related tweets using the Hedonometer, compares climate tweets with unfiltered tweets, and examines selected dates and events through word shifts and time series.
- Sentiment measurement: The Hedonometer calculates collection-level happiness from the happiness scores of individual words.Word ratings were obtained from 50 Mechanical Turk participants across 10,222 frequently used English words.
- Sentiment measurement: Average happiness is measured for tweets containing “climate” from September 14, 2008 to July 14, 2014 at daily, weekly, and monthly timescales.Words rated between 4 and 6 are omitted, and “climate” itself has a score of 5.8 and is excluded from the happiness calculation.
- Data scope: A random sample of 1,500 tweets estimates whether tweets containing “climate” concern Earth’s climate or other meanings of “climate”.The authors calculate happiness both for the full sample and after removing non-Earth-related tweets.
- Comparative analysis: Word shift graphs compare climate tweets with a random 10% reference sample of all tweets by ranking words that contribute to happiness differences.The full reference collection is called the “unfiltered tweets”.
- Event analysis: The analysis examines Hurricane Irene, Hurricane Sandy, a Midwest tornado outbreak, and the Forward on Climate Rally using happiness time series and word shifts.These events span August 2011 through February 2013.
RESULTS
Climate-related Twitter discussion became more frequent in raw counts but less frequent relative to overall activity, and its tweets were generally less happy than unfiltered tweets. Happiness varied around climate news and events, with word usage linking lower happiness to negative and disaster-related language.
- Twitter activity: Raw counts of “climate” tweets increased while their relative frequency decreased over the study period.Relative frequency divides the daily “climate” count by the daily total of the 50,000 most frequently used words; raw activity grew from approximately 1 million to 500 million tweets per day overall.
- Happiness over time: Climate tweets were consistently less happy than all tweets, although isolated daily outliers exceeded the unfiltered baseline.Longer weekly and monthly averaging reduced the significance of these outliers; one example involved repeated use of “progress” on March 16, 2009.
- Happiness over time: October 28, 2012 was among the saddest climate-discussion weeks, and October 2012 was among the saddest months, coinciding with Hurricane Sandy’s landfall.The figure presents happiness by day, week, and month against the average happiness of all tweets.
- Word shifts: 5.99 versus 5.84: unfiltered tweets had higher average happiness than climate tweets in the word-shift comparison.The shift reflects less frequent use of positively rated words and more frequent use of negatively rated words in climate discussions.
- Word shifts: “Fight”, “crisis”, “threat”, “pollution”, “denial”, “tax”, and “war” contributed to lower happiness, while “love” contributed most to the overall change.“Disaster” and “hurricane” were relatively more frequent in climate tweets, linking the discussion to natural-disaster mentions.
- Word shifts: Climate tweets used relatively less profanity, while “heaven” and “hell” appeared in connection with a Mark Twain quote about climate.Eight of 97 non-Earth-related tweets in the manual sample referenced the quote.
- Keyword comparisons: Tweets containing “climate” had previously been found to have ambient happiness similar to tweets containing “no”, “rain”, “oil”, and “cold”.The study uses these keywords as comparison terms for further climate-related analysis.
Climate Related Keywords
Keyword and hashtag comparisons reveal distinct climate-discussion communities on Twitter: skeptics more often use “global warming” and related denialist hashtags, while “climatechange” tweets emphasize environmental impacts and action.
- Keyword comparisons: Tweets containing “global warming” use more negatively rated words and profanity than tweets containing “climate”.They use words such as “fraud,” “blame,” and “freezing” more often, while “science” and “energy” appear less frequently.
- Hashtag communities: The “globalwarming” hashtag is often used by deniers, with increased negative and cold-weather language suggesting sarcastic usage.Words including “fraud,” “lie,” “snow,” and “freezing” increase relative to tweets containing “climate”.
- Hashtag communities: The “climaterealists” hashtag is associated with more words such as “fraud,” “lies,” “wrong,” and “scandal,” and fewer words related to climate action.Words including “fight,” “pollution,” “combat,” and “threat” occur less frequently.
- Hashtag communities: The “agw” hashtag represents users more opposed to anthropogenic climate change, whereas “climatechange” tweets emphasize environmental impacts and combating climate change.“agw” increases terms such as “fraud” and “conspiracy,” while “climatechange” increases “green,” “energy,” “environment,” “pollution,” and “fight”.
- Comparison with prior work: With the exception of “globalwarming,” the keyword analysis largely agrees with earlier findings, while more skeptics use “global warming” than “climate”.The “globalwarming” hashtag may also be used by activists, so the keyword and hashtag patterns are not identical.
Analysis of Specific Dates
Daily Hedonometer analysis links unusually happy or unhappy climate tweets to specific events and topics. The analysis identifies climate-related issues on Twitter but may miss some events because climate tweets form a small fraction of all tweets and include background noise.
- Scope and limitation: Climate tweets represented a very small fraction of unfiltered tweets, limiting the analysis’s ability to capture everything on Twitter.The authors note that background noise may make some events difficult to analyze.
- Saddest dates: On October 9, 2008, negative climate-tweet language concerned climate-change threats to a tropical species, a British climate bill, and the U.S. economic crisis.The date was among the saddest according to the Hedonometer analysis.
- Saddest dates: On August 6, 2011, “don’t” and “stop” contributed most to decreased happiness during discussion of the proposed Keystone XL pipeline extension.The analysis connects changes in happiness with the climate-related issues being discussed that day.
- Overall pattern: Per-day analysis shows that important climate-change issues appear on Twitter with different happiness levels depending on unfolding events.The study therefore examines specific natural disasters and a non-weather climate-related event in subsequent analyses.
Natural Disasters
Natural disasters concentrate attention around climate-related tweets and coincide with lower collective happiness. During Hurricane Sandy, disaster-related discussion decayed quickly, while climate-related discussion persisted much longer.
- Disaster events: Natural disasters focus collective attention on climate-related conversations, including Hurricane Irene, Hurricane Sandy, and a Midwest tornado outbreak.The analysis examines these three disasters using happiness time series and word shift graphs.
- Scope boundary: The event analysis captures only hurricanes and tornadoes mentioned alongside the word “climate,” excluding disasters such as Hurricane Arthur and the March 2012 tornado outbreak.The authors therefore bound the analysis to climate-linked disaster discussion rather than all disaster-related Twitter activity.
- Disaster events: Hurricane Sandy produced the largest peak in tweets containing “hurricane” alongside “climate” during the study period.The peak occurred in October 2012, when Hurricane Sandy made landfall.
- Discussion decay: During Hurricane Sandy, “hurricane” had successive half-lives of 1.57, 0.96, and 1.56 additional days, whereas “climate” had half-lives of 8.19, 22.58, and 84.85 days.The unequal half-lives indicate rapid decay for direct hurricane discussion and much slower decay for climate discussion; the decay is not exponential.
- Sentiment response: Natural disasters caused a happiness dip driven more by increases in negative words than decreases in positive words.Climate tweets used more negative words than tweets without “climate” during the disasters.
Forward on Climate Rally
The Forward on Climate Rally was associated with a slight increase in happiness among climate tweets, an uncommon pattern in the broader time series. Positive terms appeared alongside negative rally-related language, while the event focused on the Keystone pipeline bill.
- The rally aimed to persuade the U.S. government to act against climate change, with the Keystone pipeline bill as a particular focus.
- Climate-tweet happiness increased slightly above unfiltered-tweet happiness during the rally.
- This above-baseline increase occurs on only 8% of days in the broader comparison.
- Although “protestors,” “denial,” and “crisis” were present, the rally also introduced “live,” “largest,” and “promise.”
- President Obama eventually vetoed the Keystone pipeline bill.
CONCLUSION
The study explores sentiment in climate-related tweets surrounding climate news, events, and natural disasters. It finds that these tweets are less happy than all tweets and that discussion appears dominated by activists rather than deniers.
- Tweets containing “climate” are less happy than all tweets.
- Climate-related word shifts suggest discussion is dominated by climate change activists rather than climate change deniers.
- Frequent appearances of “science” and “scientists” in word shifts strengthen the conclusion that Twitter discussion largely agrees with scientific consensus.
- The study provides a general exploration of sentiment surrounding climate tweets in response to disasters, news, and events.