Advancing climate insights with social media data

With more than twenty researchers and experts from diverse fields and institutions across Europe and beyond, the IIASA Expert Workshop on Using Social Media Data and Digital Platforms in Climate and Sustainability Research showcased cutting-edge work and facilitated lively discussions on topics like the role of social media data and digital traces in analyzing low-carbon behavior changes, societal readiness for climate solutions, and the role of artificial intelligence in tracking and driving pro-environmental behaviors. Here’s a look back at some of the highlights:

Using social media data in climate research

The workshop kicked off with a welcome session and an introduction to the workshop objectives, led by the organizer Sibel Eker and Sustainable Service Systems Research Group Leader, Bas van Ruijven. They underscored the growing relevance of social media data in understanding public attitudes toward climate solutions and the shift towards a low-carbon society. Through individual presentations, participants shared insights on the value of social media as a lens into collective mindsets and driver of societal change.

The first session focused on harnessing social media data to analyze different social phenomena, such as public perception and migration patterns in climate contexts. Andrea Ghermandi from the University of Haifa discussed ways in which social media data is uniquely positioned to help environmental sustainability research, such as capturing human-nature interactions at scale. For example, using digital footprints, he was able to monitor jellyfish outbreaks along Israel’s Mediterranean coast, as well as socio-ecological change over time. He was also able to monitor the co-construction of meaning and values, for instance, using generative AI and LLMs to characterize emotions and relational values in texts associated with visits to a popular nature park. Next, Finn Muller-Hansen from the Mercator Research Institute on Global Commons and Climate Change (MCC) showcased how twitter data could be used to gauge technology perceptions. In particular, his findings reveal that while the attention to Carbon Dioxide Removal (CDR) has been on the rise, geoengineering has suffered a decline over the last years, indicating that CDR has become largely independent from the geoengineering discourse. Furthermore, he finds that the sentiments in CDR Tweets were mostly net positive and has increased over time. Carmen Loschke from the eko-Institut in turn explored whether the energy crisis has polarized citizens in Germany on the energy efficiency policy. While she finds a clear increase in public awareness of energy savings, the discourse is highly polarized. Acting Migration and Sustainable Development Research Group Leader, Dilek Yildiz, then demonstrated how Facebook audience size data can help cover the data gaps in inter-country migration flows, showing that traditional and social media data can be integrated.

Bringing traditional social science into the digital age

Following these sessions, the conversation shifted to integrating conventional social science methods with digital data sources. In particular, a comprehensive psychology model called COM-B was put under the spotlight. This is a model that accounts for capability, opportunity, and motivation for a certain behavior and integrates this with (softer and harder) intervention types such as guidelines, marketing, and legislation, among others. Sara Constantino from Stanford University shared her conclusions from behavioral experiments, providing evidence for social tipping points, highlighting that the “how” and “why” tipping occurs remains unclear, and that even a trivial link between group identities and choices can destroy potential for tipping.

Hazel Pettifor from Oxford explored how lifestyles could be characterized using a universal framework, into a common “unit” which would allow a fair comparison across countries and cultures.

Sabine Pahl and Linda Steg from the University of Groningen then led a discussion on bridging the gap between theory and large-scale data-experiments focusing on if social media data genuinely reflects real-world attitudes and behaviors, and whether such data lends itself to causal analysis. Participants debated the strengths and limitations of social media data, probing ways to integrate it effectively with conventional methodologies via iterative approaches, where empirical studies first guide social media analysis for cross-validation, followed by larger scale social media analysis guiding more detailed empirical approaches in field studies and survey design. The session concluded with reflections on the benefits and potential biases inherent in social media research and how cross-platform analysis and meta-analysis could potentially overcome these.

Exploring climate action and norms online

Day two opened with an engaging look at social media’s role in influencing climate action and social norms. Xinran Wang from the University of Groningen gave a presentation on how COP26 was perceived and discussed on Twitter. She identified three types of actors in the online discourse – victims (disproportionately affected individuals), harmful actors from government and industry who are not taking the necessary actions, and helpful actors, for example, from NGOs and research institutions, and then advocated for the removal of institutional barriers to capitalize on the socio-cultural support for ambitious climate policy which was evident from her research.

Next, Mary Sanford from CMCC talked about how initiatives like Veganuary influenced perceptions and barriers to climate action on twitter, partitioning the network of re-tweets into sources (news, activists, promotional) and their audiences (retweeters), concluding that veganism, especially as climate action, becomes heavily politicized and that this can measured by combining network-based approaches with identification of actors in online discourses.  

I then showcased how machine learning and LLMs can be used to understand the climate discourse on Reddit. I found that most topics on this platform fall under climate solutions, causes, catastrophic impacts, and societal responses. In terms of possible solutions, tree planting, nuclear energy, and electric vehicles (EVs) were among the top five most discussed topics on climate Reddit. Sentiment analysis and LLMs based detection of approval and disapproval revealed that while solutions such as EVs, nuclear reactors and geoengineering have some of the highest approval rates of all topics, they rank among the highest in disapproval, signaling strong but mixed feelings among the Reddit population.

Finally, Nathalie Van Raemdonck from VU Brussels spoke about how the design of platforms such as Instagram and Twitter shape social norms around climate. Through qualitative interviews, she investigated instances of social norm enforcement finding that the climate discourse on Twitter is sensitive to group loyalties with big in-group and out-group distinctions on man-made climate change, while for Instagram she found that cross interactions with out-groups are few. Her findings exemplify how these platforms foster and challenge collective values. These sessions illustrated social media’s powerful role in shaping narratives and potentially influencing behavior at scale.

The role of lifestyle apps and digital tools in tracking climate behavior change

The final sessions focused on other digital platforms. Victoria Spaiser from the University of Leeds, for example, discussed her eight-week field experiment with a custom-built smartphone app to monitor climate-related lifestyle choices and drive sustainable behavior. While overall no effect was observed between the control and test groups in terms of moral interventions, differences emerged over time in specific lifestyle choices. Heating usage, for instance, decreased with moral interventions, hinting at the usefulness of such an approach for domains where it is easy for people to change their behavior.  

Fabian Dablander from the University of Amsterdam shared his data-driven insights on the popular climate action app, Aworld – the official platform in support of ACTNOW, the UN’s campaign for individual action on climate change and sustainability – which aims to track sustainable behavior. Lastly, Indira Sen from the University of Konstanz introduced Glassdoor – a website where current and former employees can anonymously review companies – as a platform that can be used to understand the internal sustainability efforts of companies.  

Next steps

The workshop’s closing breakout sessions allowed participants to brainstorm future projects and collaborations. In smaller groups, attendees tackled questions around developing monitoring systems to track social change, assessing societal readiness for climate solutions, and enhancing AI-driven interventions. We outlined possible next steps, emphasizing the need for expanded data access.

With the event wrapped up, participants left with new ideas, potential future partnerships, and a renewed confidence in the potential of social media data and AI for impactful climate research at scale. Stay tuned as we work to expand the insights generated in this workshop into a perspective piece in the future!

Note: This article gives the view of the author, and not the position of the Nexus blog, nor of the International Institute for Applied Systems Analysis.