How image search algorithms shape views on climate change in different countries

Editors' notes

This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:

fact-checked

peer-reviewed publication

trusted source

proofread

The images that turn up in internet search results for the term "climate change" vary dramatically by country and tend to reflect prevailing views, according to a new study published in Nature Climate Change.

In Argentina, where intense wildfires have burned in recent years and surveys show some of the highest levels of climate concern, top internet search results are likely to include images of burning homes. In Estonia, where surveys show relatively low levels of concern about climate change, internet users are more likely to see images of icebergs, polar bears, or scientific charts that may make climate change seem like a distant or abstract phenomenon.

"If the world wants to curb climate change, it will need a global response that cannot happen when people are not well informed. More relevant and objective search outputs are key to that informing process," said senior study author Madalina Vlasceanu, an assistant professor of environmental social sciences at the Stanford Doerr School of Sustainability and director of the Climate Cognition Lab.

Vlasceanu and colleagues analyzed images displayed in Google Image Search in 49 countries in response to search terms such as "climate change." To gauge the impact of these images, they then assessed people's emotional responses and support for action after viewing them.

They found that the images displayed in the search results had higher ratings for emotional response and need for action in nations where people already had more concern about climate change—but not necessarily in countries facing the most intense climate risks and impacts. This suggests image search algorithms contribute to a sentiment propagation effect, where images reflect pre-existing sentiments.

Better search, better solutions

In a follow-up experiment, the researchers showed a variety of climate-related images displayed by Google in nations with either high or low climate concern to a new sample of nearly 900 participants to evaluate how people are influenced by such images.

They found that people who saw images that are frequently served up in high-concern nations reported feeling more threatened by climate change, supportive of climate policies, and inclined to take actions such as purchasing an electric vehicle or installing solar panels.

"This suggests to us that climate images depicting objective reality can drive broad changes in climate sentiment," said lead study author Michael Berkebile-Weinberg, a postdoctoral scholar at Columbia Business School who worked on the research with Vlasceanu while she was a faculty member at New York University.

The authors say their findings point to a need for algorithms that accurately depict climate impacts. "Internet search algorithms play an important part in informing the world about the risks of and solutions to our accelerating global climate crisis," Vlasceanu said. "If we have better image search, we can inspire people to take appropriate actions, commensurate with the degree of the threat."

Collaboration across disciplines will be important to bring about the global-scale efforts necessary to address climate change through policy innovation, educational programs, infrastructure development, or algorithmic deployment, Vlasceanu said.

Key collaborators will include behavioral scientists; developers of image search tools and generative AI; climate scientists; sustainability-focused engineers; educators; and representatives in government and law who set and enforce environmental, energy, and natural resource policy, and others.

New home, next steps

At Stanford, Vlasceanu is now expanding on the Nature Climate Change study and others to explore the behavioral, social, political, and cognitive barriers to progress on climate change and sustainability.

"My students and I are investigating how the building blocks of collective cognition—the thinking of large groups of people—can be leveraged to address important societal issues, like the way in which depicting climate change as a function of subjective sentiments rather than the objective reality can undermine societal action," Vlasceanu said.

Her work on image search has opened new research questions about how algorithms in other processes, from news gathering to generative AI, may shape society's impressions of climate risk. Her lab is also interested in how societal structures, such as policies, educational programs, infrastructure, and entertainment, shape the assessment, feasibility, and adoption of climate solutions.

More information: Michael Berkebile-Weinberg et al, Internet image search outputs propagate climate change sentiment and impact policy support, Nature Climate Change (2024). DOI: 10.1038/s41558-024-02178-w

Journal information: Nature Climate Change

Provided by Stanford University