What is the moral of the story around climate change?

I’m very excited to announce a new paper that was authored by my student Haiqi Zhou and that was recently published as part of the first workshop on NLP meets Climate Change.

In this study, we explore the use of Large Language Models (LLMs) such as GPT-4 to extract and analyze the latent narrative messaging in climate change-related news articles from North American and Chinese media. By defining “narrative messaging” as the intrinsic moral or lesson of a story, we apply our model to a dataset of approximately 15,000 news articles in English and Mandarin, categorized by climate-related topics and ideological groupings.

Our findings reveal distinct differences in the narrative values emphasized by different cultural and ideological contexts. As can be seen in the figure below (top), Chinese sources emphasize promoting “international cooperation” and “sustainable development,” while North American sources concentrate on “addressing” and “adapting” to climate change, reflecting a potentially stronger sense of crisis, but also individualism. Additionally, the Chinese emphasis on “promoting” solutions contrasts with a North American emphasis on reacting to and accepting consequences. The developmental framing of Chinese-language news as a whole suggests a far more proactive stance than the North American one.

Within North America (middle), Conservative outlets place a much stronger emphasis on markets, investments, and economic and financial issues. This reflects the 2019 Pew Research Center survey that shows Conservative Republicans being skeptical towards climate policies — a majority (62%) of this group says these policies hurt the economy. We also note that Conservative media view climate change through the “adapting” lens, while Liberal media emphasizes a lens of “addressing.” Conservative messaging focuses on climate change as a process of change, whereas Liberal media tends to view the problem as something that must be “solved.”

This work should hopefully demonstrate the potential LLMs introduce in understanding and influencing climate communication, offering new insights into the collective belief systems that shape public discourse on climate change across different cultures. Our hope is that this knowledge can be used to shape climate messaging to reach more resistant audiences and identify the most pro-active stance for changing human behavior.