Project Publications:
Global Dynamics of Climate Change Imagery: Emotional and Engagement Effects Across Visual Frames on Twitter/X
Author(s): Isaac Bravo, Katharina Prasse, Stefanie Walter, Saffron O'Neill, and Margret Keuper
Date Published: 2025
Abstract:
This study examines how climate change is visually framed on Twitter/X and how these frames influence users’ emotional responses and engagement across languages. Using automated image analysis of over three million images posted between 2019 and 2022, we identify nine dominant visual
frames. “Public Engagement: Politics & Events” was most prevalent, interactive visuals such as memes generated the highest engagement but predominantly negative response in user comments. By contrast, we find more positive responses for solution-focused frames, but lower levels of
engagement. These results underscore the value of automated image analysis in uncovering culturally nuanced patterns in visual climate communication.
Citation:
Bravo, Isaac, et al. "Global Dynamics of Climate Change Imagery: Emotional and Engagement Effects Across Visual Frames on Twitter/X." Science Communication (2025): 10755470251401900.
Download the paper.
DCBM: Data-Efficient Visual Concept Bottleneck Models
Author(s): Katharina Prasse, Patrick Knab, Sascha Marton, Christian Bartelt, and Margret Keuper
Date Published: July 13, 2025
Abstract:
Concept Bottleneck Models (CBMs) enhance the interpretability of neural networks by basing predictions on human-understandable concepts. However, current CBMs typically rely on concept sets extracted from large language models or extensive image corpora, limiting their effectiveness in data-sparse scenarios. We propose Data-efficient CBMs (DCBMs), which reduce the need for large sample sizes during concept generation while preserving interpretability. DCBMs define concepts as image regions detected by segmentation or detection foundation models, allowing each image to generate multiple concepts across different granularities. This removes reliance on textual descriptions and large-scale pre-training, making DCBMs applicable for fine-grained classification and out-of-distribution tasks. Attribution analysis using Grad-CAM demonstrates that DCBMs deliver visual concepts that can be localized in test images. By leveraging dataset-specific concepts instead of predefined ones, DCBMs enhance adaptability to new domains.
Citation:
Prasse, Katharina, et al. "DCBM: Data-Efficient Visual Concept Bottleneck Models." International Conference on Machine Learning. PMLR, 2025.
Download the paper and the code.
I Spy With My Little Eye: A Minimum Cost Multicut Investigation of Dataset Frames
Author(s): Katharina Prasse, Isaac Bravo, Stefanie Walter, and Margret Keuper
Date Published: Jan 3, 2025
Abstract:
Visual framing analysis is a key method in social sciences for determining common themes and concepts in a given discourse. To reduce manual effort, image clustering can significantly speed up the annotation process. In this work, we phrase the clustering task as a Minimum Cost Multicut Problem [MP]. Solutions to the MP have been shown to provide clusterings that maximize the posterior probability, solely from provided local, pairwise probabilities of two images belonging to the same cluster. We discuss the efficacy of numerous embedding spaces to detect visual frames and show its superiority over other clustering methods. To this end, we employ the climate change dataset \textit{ClimateTV} which contains images commonly used for visual frame analysis. For broad visual frames, DINOv2 is a suitable embedding space, while ConvNeXt V2 returns a larger number of clusters which contain fine-grain differences, i.e. speech and protest. Our insights into embedding space differences in combination with the optimal clustering - by definition - advances automated visual frame detection.
Citation:
Prasse, K. et al. (2025) I Spy With My Little Eye: A Minimum Cost Multicut Investigation of Dataset Frames. WACV.
Download the paper and the code.
Towards Understanding Climate Change Perceptions: A Social Media Dataset
Author(s): Katharina Prasse, Steffen Jung, Isaac Bravo, Stefanie Walter, and Margret Keuper
Date Published: December 16, 2023
Abstract:
Climate perceptions shared on social media are an invaluable barometer of public attention. By directing research towards this topic, we can eventually improve the effectiveness of climate change communication, increase public engagement, and enhance climate change education. We propose two real-world image datasets to promote impactful research both in the Computer Vision community and beyond. Firstly, ClimateTV, a dataset containing over 700,000 climate change-related images posted on Twitter and labelled on basis of the image hashtags. Secondly, ClimateCT, a Twitter dataset containing images with five-dimensional annotations in super-categories (i) Animals, (ii) Climate action, (iii) Consequences, (iv) Setting, and (v) Type. These challenging classification datasets contain classes which are designed according to their relevance in the context of climate change. The challenging nature of the datasets is given by varying class diversities (e.g. polar bear vs. land mammal) and foci (e.g. arctic vs. snowy residential area). The analyses of our datasets using CLIP embeddings and query optimization (CoCoOp) further showcase the challenging nature of ClimateTV and ClimateCT.
Citation:
Prasse, K. et al. (2023) Towards Understanding Climate Change Perceptions: A Social Media Dataset. https://www.climatechange.ai/papers/neurips2023/3
Download the paper and the data.
Visualizing Climate Change in the Media: A Systematic Literature Review, Challenges, and Future Research
Author(s): Isaac Bravo, Stefanie Walter, Katharina Prasse and Margret Keuper
Date Published: Annals of the International Communication Association.
Abstract:
Climate change visualizations in the media play a crucial role in conveying information, raising awareness, and motivating action. This study combines systematic and scoping literature review (2005–2024) with content analysis to analyze current research on climate change visualizations of traditional and digital media. The findings show that most existing research concentrates on traditional media, with limited focus on social media and a predominance of studies focused on Western countries. Framing theory emerges as the predominant theoretical framework, especially in qualitative studies. By analyzing and comparing a large corpus of scientific studies, this study identifies predominant topics, methodologies trends, and research gaps, while also highlighting key challenges and implications for future research.
Citation:
Bravo, Isaac, et al. "Visualizing climate change in the media: a systematic literature review, challenges, and future research." Annals of the International Communication Association (2026): wlag011.
Download the paper.
Conference Presentations:
I Spy With My Little Eye: A Minimum Cost Multicut Investigation of Dataset Frames.
More information about the Conference, see here.
I Spy With My Little Eye: A Minimum Cost Multicut Investigation of Dataset Frames.
More information about the Conference, see here.
Detecting Manipulated Visuals: A Computational Approach in the Climate Change Discourse.
More information about the Conference, see here.
Computational Analysis of Manipulated Visual Content in Climate Change Discourse on Twitter.
More information about the Conference, see here.
Analyzing Manipulated Images in the Climate Change Discourse
More information about the Conference, see here.
Towards Understanding Climate Change Perceptions: A Social Media Dataset
More information about the Conference, see here.
Analysing the Role of Social Media in Shaping Public Opinion on Climate Change: A Comparative Study Across Regions Using Automated Image and Text Analysis
More information about the Conference, see here.
Analysing the effects of visual framing on social media in shaping people's emotional engagement on climate change
More information about the Conference, see here.