Scientists Build New Tool to Combat Climate Lies
New data set helps fight online climate misinformation.
Scientists have created a new, specialized dataset to help identify and combat the spread of incorrect or misleading information about climate change. This moves us closer to systems that can alert people to science denial.
For decades, certain groups have spread false ideas about climate change. These groups use tactics similar to those once used by the tobacco industry to confuse the public. Researchers aimed to close a gap in available data, building a resource specifically for climate change falsehoods.
Building the Dataset
The team collected articles from 15 "Climate Change Counter Movement" organizations that actively spread misinformation. They also gathered content from trusted news sources like The Guardian and BBC. This allowed them to create a large collection of both true and misleading articles.
The dataset includes 1,168 articles for training, with an average length of about 559 words.
Characteristics of Misleading Articles
The study highlights that misleading climate articles often share features with "fake news." They use dramatic language and enticing headlines, much like clickbait.
"The rise of this misleading information is part of a carefully crafted strategy by climate change counter movement (CCCM) organizations," the authors state. "These organizations use a formula consisting of a narrative structured around the principle ingredients of disinformation, misinformation, propaganda and hoax, sprinkled with the stylistic elements of sensationalism, melodrama, clickbait and satire..."
Impact and Future Steps
This new dataset is a crucial step for computer programs that learn to spot patterns, known as Natural Language Processing (NLP). It gives these programs the raw material needed to understand and flag misleading content. The goal is to develop applications that can automatically detect or warn users about climate change misinformation.
One limitation is that some organizations had more articles than others, which could introduce a slight bias. However, this new dataset is a significant step forward in battling the unique challenge of climate misinformation. The next step is to use this dataset to train artificial intelligence to recognize and flag these misleading messages when they appear online.
This research provides a vital weapon in the fight for truth in the digital age.
arXiv:2004.14907v1 [cs.CL] 30 Apr 2020