Narratives of climate change
Published:
How do we talk about climate?
Background: how is climate change talked about? This project started as a curiosity about the kinds of stories that get told about climate. I wanted to know how to distinguish true information about the science and the effects of climate change from that which is false. As I’ll explore below, factfulness is more of a spectrum.
This dataset is complied by Google researchers and I am using Huggingface models to classify statements based on sentiment. Secondly, I wanted to know more about the kinds of topics that make up the climate discourse and make these findings into a tool for others to use too. So I labelled the dataset into a handful of commonly talked about climate topics (bears, sea levels, warmings) and then fine tuned a BERTopic model on it. This new model is called BERTClimate, check it out!
How do different claims compare in their sentiment? First off, most claims are classified as negative, across all 4 truthfulness categories. Claims that are supported by facts are the most negative (70%) compared to 60% for the rest.
Most of the coverage is about warming, and then CO2 and extreme weather:
Claims about fires don’t have enough evidence, which is to the case for other topics. Strikingly, most topics are not talked about in a way that is factually supported. For all topics, there is almost an even split between “supports” and “refutes” or not being enough evidence for the claims:
Unsurprisingly, most of the topics have a negative label:
This was a short exploration of the language of climate change. A few things stuck out to me which would make for interesting follow up questions. First, in the “warming” topic category, a lot of the claims were related to modeling and predictions. Given climate is so complex, building models is key but often not well understood. I would like to think more about that: how science in particular gets talked about. This is also related to topics of public policy around climate change.