How do we react to social media posts about climate change? This was the starting point of an international study featuring INESC TEC collaboration, analysing nearly two million posts on the social network X (formerly Twitter) published over 12 years in the Iberian Peninsula.
Shock, anger and irony; these are the common ways citizens express themselves when the topic concerns extreme weather events. Ana Sofia Cardoso and Alípio Jorge, INESC TEC researchers, are co-authors of the study – which was published in Scientific Reports (Nature Group).
The research, which involved institutions from Portugal, Spain, the Czech Republic and France, used advanced natural language processing (NLP) and deep learning techniques, covering the period from 2010 to 2022. The results show that 39% of the posts (often linked to phenomena such as wildfires, droughts and storms) express negative feelings, while 35% are neutral.
Among the most identified emotions are anger, surprise and, to a lesser extent, joy. The frequency of posts associated with anger has increased over the past four years, reflecting growing public concern about the impact of – and responses to (or lack thereof) – climate crisis. Irony, used as a form of criticism or venting, accounted for around a quarter of the content, a trait particularly visible in Portugal and Spain. Peaks in emotional activity were recorded during events like the Pedrógão Grande wildfires (2017), Ophelia (2019), and the 2022 heatwave, which destroyed 57,000 hectares and forced the evacuation of thousands of people.
“Studies like this highlight the importance of monitoring public discourse on social media as a complementary tool for climate planning and crisis response,” stressed Alípio Jorge.
The INESC TEC team played a key role in applying and training BERT-based (Bidirectional Encoder Representations from Transformers) language models. These algorithms are trained on large volumes of textual data in Portuguese, Spanish and English and can be retrained for more specific tasks such as sentiment analysis.
“This work is part of INESC TEC’s ongoing research in NLP, particularly in applying pre-trained language models to specific tasks, as already happens in other contexts like the Citilink project. The methodology we’ve developed is scalable and replicable. It has the potential to be applied in new research and contexts, whether in social sciences, risk communication or public policy support,” added Alípio Jorge.
The researchers mentioned in this news piece are associated with INESC TEC and the Faculty of Sciences of the University of Porto.
News, current topics, curiosities and so much more about INESC TEC and its community!
