November, 2025 | Presentation and Publication in EMNLP Findings

From November 4 to November 9, 2025, Shijia Zhou and Dr. Siyao (Logan) Peng presented recent project findings at the Conference on Empirical Methods in Natural Language Processing (EMNLP 2025) in Suzhou, China.

The study “What Media Frames Reveal About Stance: A Dataset and Study of Memes in Climate Change Discourse” is a collaboration between the Computational Linguistics group led by Prof. Dr. Barbara Plank, supported by Saif M. Mohammad, and the Communication Science team represented by Dr. Simon Lübke, Dr. Jörg Haßler, and Prof. Dr. Mario Haim. In this work, the authors adopt an interdisciplinary approach to conceptualize and computationally explore how climate change is expressed through internet memes. They introduce CLIMATEMEMES, the first dataset of climate change memes annotated for stance and media frames.

The paper is now published in the Findings of the Association for Computational Linguistics: EMNLP 2025.

November 12, 2025 | Research stay at GESIS

From October 27 to November 7, Nadja Ozornina was invited to GESIS – Leibniz institute for the Social Sciences as a guest researcher. Together with colleagues from KODAQS: Competence Center for Data Quality in the Social Sciences, she developed a tool for multilingual topic modeling and had the opportunity to present the KLIMA-MEMES project at the colloquium of the Computational Social Science department. The research stay contributed to the discussion on applying computational methods to analyze multimodal data in the context of climate change and outlined new avenues for collaboration. 

July 22, 2025 | Talk at IC2S2

From July 21 to 24, Nadja Ozornina participated in the 11th International Conference on Computational Social Science (IC2S2) in Norrköping, Sweden. At the conference, she presented findings from her ongoing study, „Making Sense of Images: A Literature Review on Approaches to Integrating Visual Data into Clustering for Social Science,“ which is part of her dissertation project supervised by Prof. Dr. Mario Haim. In this study, Nadja conducted a systematic literature review to identify key concepts, methodological approaches, and challenges in image clustering. Her work contributes to the development of analytical strategies for handling large-scale multimodal datasets from social media, which include both textual and visual content.

July 15, 2025 | Talk at IAMCR 2025 in Singapore 

„Communicating Environmental Justice: Many Voices, One Planet” was the theme of this year’s International Association for Media and Communication Research (IAMCR) conference, which took place from July 13 to 17. During the conference at Nanyang Technological University (NTU) in Singapore, Nadezhda Ozornina presented findings from the KLIMA-MEMES project and joint research with Simon Lübke, Mario Haim, and Jörg Haßler. The presentation, entitled “Fighting the Climate Crisis with Humor: Analyzing Climate-Related Internet Memes on Reddit,” presented the results of an analysis of internet memes from online platform Reddit. The study  demonstrates that memes about climate change more frequently address the consequences of climate change than its causes or appropriate countermeasures. The analysis also revealed that the memes primarily contained aggressive humor that negatively mocks others. Thus, our research shows that meme authors often use humor to assign responsibility for the climate crisis.

June 17, 2025 | Publication on Climate Communication in the Hybrid Media System

How do political actors and media outlets discuss climate change on social media? This very question is the focus of our latest publication, which was recently published in the journal Media and Communication. In the article, Simon Luebke, Nadezhda Ozornina, Mario Haim, and Jörg Haßler analyze posts on TikTok, Facebook, Instagram, and YouTube created by climate stakeholders and media organizations in Germany during the 2023 UN Climate Change Conference (COP28) in Dubai.

The paper titled ‘Climate Communication in the Hybrid Media System: Media and Stakeholder Logics on Social Media’ examines the extent to which media organizations and political actors follow different communicative logics when addressing climate change. Given today’s hybrid media landscape, where various actors interact on social platforms, we explore whether these distinct logics are beginning to converge. Over a 16-day period surrounding COP28, we collected and analyzed all social media posts from high-reach media outlets, as well as from individuals and institutions that influence climate discourse and policy in Germany. In total, we manually coded and systematically analyzed 1,050 posts using a standardized manual content analysis.

Our findings reveal clear distinctions. Media outlets tend to frame climate change as a political issue, emphasize its negative consequences, and focus slightly more on individual figures (personalization). In contrast, political actors are more likely to use strategic frames, discuss the climate crisis beyond the context of COP28, and address specific policy measures. Interestingly, posts from both groups receive higher engagement (likes) across platforms when they adopt a populist framing. That is, when they criticize current climate policies and express skepticism about the necessity of climate protection measures.

The article is part of a special issue on “Journalism in the Hybrid Media System” and can be accessed (open access) via the link below. 

June 12 – 16, 2025 | KLIMA-MEMES at ICA25 in Denver

From June 12–16, the 75th annual conference of the International Communication Association (ICA) took place in Denver. The KLIMA-MEMES project was featured in two presentations that showed research by Simon Lübke, Nadezhda Ozornina, Mario Haim, and Jörg Haßler about climate communication on social media.

The first paper, titled „Fragmentation of Digital Climate Change Discourse?,“ presented a comparative study of how key actors in the German climate discourse communicate on Facebook, Instagram, TikTok, and YouTube. The authors observed clear differences between political actors, such as parties, politicians, and NGOs, and climate journalists. While journalistic content tends to be event-driven and personalized, emphasizing negative aspects of climate change, stakeholders treat climate change as a broader, ongoing issue independent of events like climate summits and frame it in ways that align with their own interests.

The second paper, titled „The Rise of Climate Change Skepticism in Germany,“ examined the climate communication strategies employed by the far-right political party Alternative für Deutschland (AfD). The authors compared climate-related content from all German political parties represented in the Bundestag during the UN Climate Change Conference (COP28) in Dubai. One notable finding was that the AfD communicated about climate change more frequently than previous research suggested. The party primarily expressed skepticism about the urgency of the climate change issue and questioned the necessity of climate protection policies.

June 15, 2025 | CVPR: Presenting Diff2Flow in Nashville

From June 11–15, the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) took place in Nashville – and our colleagues from the KLIMA-MEMES project were there to present a poster. Johannes Schusterbauer, Ming Gui, Frank Fundel, and Björn Ommer presented their research titled „Diff2Flow: Training Flow Matching Models via Diffusion Model Alignment.“ In their work, they explore a central question in the field of generative AI: How can we transfer the knowledge of a trained diffusion model to a faster flow-matching model for image generation? To tackle this, the team proposes a novel framework called Diff2Flow. It allows this transfer to happen without the need to retrain the entire model, making generative AI both faster and more efficient.

Abstract: Diffusion models have revolutionized generative tasks through high-fidelity outputs, yet flow matching (FM) offers faster inference and empirical performance gains. However, current foundation FM models are computationally prohibitive for finetuning, while diffusion models like Stable Diffusion benefit from efficient architectures and ecosystem support. This work addresses the critical challenge of efficiently transferring knowledge from pre-trained diffusion models to flow matching. We propose Diff2Flow, a novel framework that systematically bridges diffusion and FM paradigms by rescaling timesteps, aligning interpolants, and deriving FM-compatible velocity fields from diffusion predictions. This alignment enables direct and efficient FM finetuning of diffusion priors with no extra computation overhead. Our experiments demonstrate that Diff2Flow outperforms naïve FM and diffusion finetuning particularly under parameter-efficient constraints, while achieving superior or competitive performance across diverse downstream tasks compared to state-of-the-art methods.

June 10, 2025 | Workshop Media & Climate Change in Boulder

Ahead of the ICA Annual Conference in Denver, a one-day workshop on Media & Climate Change took place at the University of Colorado Boulder. The event was jointly organized by the Munich Science Communication Lab (MSCL) at LMU, the Media and Climate Change Observatory at the University of Colorado Boulder, and Denison University’s Sustainability & Environmental Studies Program.

During the workshop, Simon Lübke presented the interdisciplinary research project KLIMA-MEMES, offering insights into several areas of the project’s work. Among other topics, his presentation focused on the role of internet memes in climate communication—exploring how visual humor, irony, and user-generated content shape public engagement with climate change across digital platforms.

The workshop brought together researchers from communication studies, environmental science, and media analysis to discuss current challenges and new approaches to climate communication. We’re grateful to have contributed to this vibrant exchange—and to share perspectives from the KLIMA-MEMES project on how digital culture intersects with the climate crisis.

June 4, 2025 | Guest Lecture at BU College of Communication

During a research stay at Boston University, Simon Lübke was invited to present a talk at the College of Communication. Hosted by Prof. Betsi Grabe—who serves as an international expert on the KLIMA-MEMES project—Simon introduced the goals and methods of KLIMA-MEMES and shared our latest findings on climate change skepticism in Germany.

The session brought together students and doctoral researchers from the College of Communication to discuss how skepticism toward climate policy manifests in different national contexts. Participants compared our German data with trends they’ve observed elsewhere, enriching the conversation with global perspectives. In addition to the exchange about climate change communication, we held a mini–workshop on effective presentation techniques and the key expectations for academic talks. The interactive format allowed attendees to give feedback on delivery style and slide design, fostering a practical dialogue on scientific communication.

We send our sincere thanks to Prof. Grabe for her warm hospitality and to everyone who joined us for their insightful questions and constructive feedback.

May 26 – 28 | Talks at re:publica 25

From May 26 to 28, 2025 our colleagues from the KLIMA-Memes project attended re:publica25, the festival of the digital society. Björn Ommer and Simon Lübke  were both speakers at the conference. Björn Ommer had the honor of opening the event with his keynote speech, „Generative AI and the Future of Intelligence.“ On the conference’s largest stage, he discussed why AI is becoming a critical enabling technology and its implications for our society. He also explained why sovereignty in the AI age requires a new mindset. Simon Lübke gave a presentation at the event as well. In his talk, „Vote for Me, I’m Authentic!,“ he discussed the importance and impact of authenticity in politics and presented findings from his research. We thank the organizers for giving us the opportunity to present our work at the event.

May 8, 2025 | PRUF Symposion 2025 

Simon Lübke presented the results of an analysis of the 2025 federal election campaign on TikTok at the Institute for German and International Party Law and Party Research (PRUF) symposium. Entitled „Mit TikTok in den Bundestag?!“, the presentation was based on a content analysis of the parties‘ TikTok profiles during the final phase of the election campaign. This analysis was conducted with Anna-Katharina Wurst and Jörg Haßler. Initial results show that the Left Party and the Alternative for Germany (AfD) had the highest average number of likes and shares. Migration and the economy were also the most frequent topics on TikTok. Additionally, the analyses revealed that right-wing parties engage in negative campaigning more frequently than left-wing parties and that their content focuses heavily on individual politicians.

April 16, 2025 | COMPTEXT 2025 Conference

Nadezhda Ozornina, Markus Baumann, and Johannes Schusterbauer presented findings from the KLIMA-MEMES project at the COMPTEXT 2025 conference in Vienna. Their talk, titled “Where Does the Bias Hide? Exploring the Applicability of Multimodal Models for Clustering Multilingual Social Media Data,” was based on an interdisciplinary collaboration between communication scientists and computer vision researchers. It investigated how multimodal data, such as internet memes, could be clustered across languages. A challenge is that the content must be grouped by both language and content across languages. To accomplish this, the researchers used the Contrastive Language-Image Pretraining (CLIP) approach and compared different combinations of text input and their effects on clustering. Their results show that the model works best when fed AI-generated descriptions of content to find thematically coherent, cross-language clusters.

Abstract: 

Where Does the Bias Hide? Exploring the Applicability of Multimodal Models for Clustering Multilingual Social Media Data.
The rise of multimodal content on social media underscores the need for automated text-image clustering in the social sciences (Peng et al., 2023). As global issues are discussed across multiple languages simultaneously, applying multimodal models like CLIP beyond single-language contexts becomes important. However, using these models for multilingual tasks presents a challenge of language bias, remaining underexplored in text-image settings. This study examines the applicability of multimodal models for multilingual social media data, focusing on language bias in text-image clustering. We apply the CLIP model to a multilingual dataset of 65,685 Instagram images with OCR-generated texts collected during COP28. Building on prior research in multilingual analysis (e.g. Reber, 2019), we combine image embeddings with text embeddings from either (a) the original texts or (b) English translations by Google. The k-means clustering results are then compared. The analysis reveals biases in the multimodal clustering approach using CLIP, particularly in its tendency to cluster data by language rather than content. Contrary to expectations, machine translation only marginally mitigates these biases, which become more pronounced when images are weighted more heavily for classification. This suggests that biases are rooted not only in the textual but also in image features, particularly in how the visual representation of texts differs across languages. We explore potential techniques for mitigating language bias, such as inpainting to neutralize text-based visual elements and vision-language models for generating more language-agnostic textual descriptions of images. The findings highlight the potential and limitations of these approaches, offering new avenues for multimodal analysis.

March 28, 2025 | Workshop on Coordinated Desinformation

Researchers from the KLIMA-MEMES project attended a workshop hosted by the bidt and the BLM on coordinated disinformation on March 28, 2025. At the event, experts from research, industry, politics, and civil society discussed findings on how to detect and combat disinformation on the internet.

Simon Lübke gave a short presentation on the topic “Countering ‚climate ideology‘: the role of climate change skepticism in the German climate change discourse on social media”. The talk presented findings on the proportion of climate change skepticism on platforms such as Facebook, YouTube, Instagram, and TikTok. The results show that climate change skepticism tends to play a lesser role in Germany than in other countries. In addition, the analyses show that institutional actors in the German discourse mainly use more subtle forms of skepticism that express doubts about the necessity of climate change measures.

March 20, 2025 | KLIMA-MEMES Talks at DGPuK 2025 Conference

The 70th annual conference of the German Communication Association (DGPuK) took place from March 19 to 21 at the Free University of Berlin. The conference theme was „The Public and its Values“. The KLIMA-MEMES project had two talks at the conference. First, Simon Lübke presented a joint paper with Nadja Ozornina, Mario Haim and Jörg Haßler comparing climate change communication by politicians and journalistic media on different social media platforms. They conclude that the communication of political actors and the media differs in the focus and topics of climate change.

Second, the same team of authors presented a paper on the AfD’s climate communication in social media. The talk presented by Simon Lübke showed that climate change had a comparatively high priority in social media communication during the COP28 in Dubai and that the party shared climate change skeptical positions on social media.

January 20, 2025 | COP29: Data collection completed

The KLIMA MEMES project analyzes visual social media content on climate change. As part of the project, we have once again collected data from different social media content to analyse communication on the topic and from different political actors. The context of the data collection was the UN Climate Change Conference 2024 in Baku, Azerbaijan (COP29). A similar sampling strategy was used last year to collect data on COP28 in Dubai. The UN climate conferences are global events where attention to climate change is particularly high, and the number of climate change stories reaches its highest level each year. This year’s event in Germany coincided with the collapse of the national government. Whether this had an impact on the prominence of the events in Baku and the discussions on climate change will be examined later in the project.

November 27, 2024 | Research talk by Saif Mohammad

On 27 November 2024, Dr Saif Mohammad from the National Research Council Canada and an international expert in the project gave a lecture entitled „NLP for Affective Science: A Window into Emotions through Language and Computation“ at the MaiNLP research lab at LMU. In his talk, Dr Mohammad discussed core theories of affect and emotion, word-emotion associations, and approaches for tracking emotional arcs. 

We would like to thank Saif for his inspiring talk and for sharing valuable insights into his research.  

November 26, 2024 | Project workshop with international experts

On 26 November, we held our second project workshop in Munich, together with our international experts Betsi Grabe (Boston University) and Saif Mohammad (NRC Canada). The meeting, which brought together all working groups, focused on the exchange and evaluation of (inter)disciplinary analyses of climate change communication on social media.

In the first part of the workshop, the working groups presented their approaches and results so far, as well as their goals for the next phase of the project. The second part of the workshop focused on assessing the interdisciplinary collaboration within the project so far and exploring further opportunities for cross-group collaboration.

We would like to thank all participants of the workshop and especially Betsi Grabe and Saif Mohammad for their valuable input.

November 26, 2024 | Lecture by Betsi Grabe

On 26 November, Prof. Betsi Grabe from Boston University and an international expert on the project gave a lecture on „Taking Images Seriously“ at the LMU. The lecture was part of the Political Communication course taught by Jörg Haßler at the Department of Media and Communication. In her talk, Betsi Grabe provided insights into the dynamics of US election campaigns over the past 20 years, focusing in particular on patterns in the visual self-representation of the leading candidates.

We thank Betsi for providing insights into her research agenda and the inspiring talk.

November 15, 2024 | Memes on collapse of Germany’s Government

On November 6, the governing coalition in Germany, consisting of the SPD, the Greens, and the FDP, collapsed after Chancellor Olaf Scholz initiated the dismissal of Finance Minister Christian Lindner, causing the FDP to withdraw from the government. Following these events, numerous memes about the situation appeared on various social media platforms. In subsequent media interviews, researchers from our KLIMA-MEMES project provided scientific insights into the motives behind the creation of memes, their content and their impact.

The individual articles with Bayerischer Rundfunk (BR), Westdeutscher Rundfunk (WDR) und Norddeutscher Rundfunk (NDR) can be accessed via the links below:

October 11, 2024 | LMU Climate School

As in the previous year, the KLIMA-MEMES project was once again part of the LMU Climate School in 2024. At the interdisciplinary event on October 11, Simon Lübke led a workshop on news media and climate change. In the workshop, participants discussed the basics of journalistic reporting on climate change. They learned how to recognize and identify reporting patterns in climate change reporting. The workshop also provided insights into current trends in the climate debate were discussed, including constructive journalism and the role of internet memes and humor.

October 3, 2024 | ECCV 2024

Johannes Schusterbauer and Björn Ommer took part in the 8th European Conference on Computer Vision ECCV 2024 in Milan from September 29 to October 4. Among other things, they presented the project paper entitled “FMBoost: Boosting Latent Diffusion with Flow Matching” and were selected for an oral presentation. In this paper, they present a method that can be used to achieve complex image processing models with less computational effort.

Abstract: Visual synthesis has recently seen significant leaps in performance, inparticular due to breakthroughs in generative models. Diffusion models have been a key enabler as they excel in image diversity. This, however, comes at the prize of slow training and synthesis which are only partially alleviated by latent diffusion. To this end, flow matching is an appealing approach due to its complementary characteristics of faster training and inference but less diverse synthesis. We demonstrate that introducing flow matching between a frozen diffusion model and convolutional decoder enables high-resolution image synthesis at reduced computational cost and model size. A small diffusion model can then provide the necessary visual diversity effectively, while flow matching efficiently enhances resolution and details by mapping the small to a high-dimensional latent space. These latents are then projected to high-resolution images by the subsequent convolutional decoder of the diffusion approach. Combining the diversity of diffusion models, the efficiency of flow matching, and the effectiveness of convolutional decoders, achieves state-of-the-art high-resolution image synthesis at 1024 ² pixels with minimal computational cost. Cascading our model optionally boosts this further to 2048 ² pixels. Importantly, our approach is orthogonal to recent approximation and speed-up strategies for the underlying model, making it easily integrable into the various diffusion model frameworks.

September 25, 2024 | ECREA 2024 Conference

Jörg Haßler and Simon Lübke attended the 10th ECREA European Communication Conference in Ljubljana last week, from September 24 to 27. During the conference, Jörg Haßler presented their joint work with Simon Lübke, Nadezdha Ozornina, and Mario Haim titled „The use of humor styles in memes on climate change“ in the panel „Public understandings of climate change.“ The presentation discussed various humor styles used in internet memes and shared findings from our project’s analyses on the use of humor in content related to COP28.

September 20, 2024 | forsa / Paul Lazarsfeld Scholarship for Nadezdha Ozornina

Nadezdha Ozornina has been awarded the forsa / Paul Lazarsfeld Scholarship 2024 by the Methods Division of the German Communication Association (DGPuK). Our colleague from the KLIMA-MEMES project was honored with this award for her work „Comparing Multilingual Topic Models“ at the joint annual conference of the Methods and Digital Communication Divisions of the DGPuK in Hamburg. The Paul Lazarsfeld Scholarship recognizes outstanding students or graduates in journalism and communication studies for particularly demanding or innovative research and final theses in the field of methods. The entire project team congratulates Nadezdha on this achievement.

September 19, 2024 | Interview with Jörg Haßler on the political communication by German parties on TikTok

The political communication of the Alternative für Deutschland (AfD) on social media is often cited as one of the reasons for the party’s recent electoral successes. Specifically, the party’s content on TikTok is highlighted to explain the high voter turnout among younger demographics. In an interview with Deutschlandfunk, the project leader of KLIMA-MEMES, Jörg Haßler, analyzes the AfD’s communication strategy on TikTok. He points to the heightened emotionalization of AfD content as a key factor explaining the higher average reach of AfD posts compared to those of other parties. These and other insights from Jörg Haßler can be found in the interview titled “ „Soziale Medien – Wie demokratische Parteien auf TikTok nachziehen“ by Tobias Krone.

September 19, 2024 | Conference of the DGPuK Methods Division

Nadja Ozornina took part in the joint annual conference of the DGPuK Methods and Digital Communication Divisions in Hamburg from September 18 to 20. As part of the conference, she presented a joint paper with Simon Lübke and Mario Haim and entitled “More than the sum of its parts: Challenges and approaches of automated analysis of multimodal social media content”. The paper deals with the analysis of multimodal social media content, i.e. content that combines multiple modalities such as text and image to construct meaning. The presentation discussed the methodological challenges that exist when analyzing multimodal content and presented various approaches to solving them.

Abstract: In the context of datafication, social media posts are an important source for automated analyses in communication science. Multimodal content, which combines different modalities such as text, images and sound, is playing an increasingly important role. However, analyzing this content poses methodological challenges for researchers, as images and texts have to be processed differently and it is often unclear how the information from the different modalities can be linked. Current research methods in communication science are mostly unimodal, while an overview of existing multimodal approaches and their characteristics is lacking. The article identifies three different groups of approaches for analyzing multimodal content, which differ in terms of the number of concepts examined and the respective intermediate analytical steps. Various application examples for the three approaches are presented and the advantages and disadvantages of the individual methods for different questions are discussed.

September 16, 2024 | Interview on pro-Russian cartoons and memes

In an investigative report by NDR, WDR and Süddeutsche Zeitung, internal documents from the Moscow-based Social Design Agency (SDA) were analyzed, providing insight into Russian disinformation campaigns in Germany. The documents show that Germany is one of the Kremlin’s preferred targets for spreading disinformation. In this context, the SDA disseminates numerous cartoons and internet memes on social media that specifically convey fake news and pro-Russian narratives. Our colleague from the KLIMA-MEMES project, Simon Lübke, analyzed some of the cartoons from the data leak for the Süddeutsche Zeitung. He also provides a short assessment of the conditions under which this content can be effective.

The article with the research results and Lübke’s statements can be accessed via the following link.

August 16, 2024 | ACL 2024 Conference

Shijia Zhou, Siyao Peng, and Barbara Plank from the KLIMA-MEMES Project attended the 62nd Annual Meeting of the Association for Computational Lunguistics (ACL 2024) from August 11-16 in Bangkok to present results from several research projects. Among them was a project presentation entitled “CLIMATELI: Evaluating Entity Linking on Climate Change Data” in which they use the entity linking approach to understand who and what is mentioned in different climate change text content.

Abstract: Climate Change (CC) is a pressing topic of global importance, attracting increasing attention across research fields, from social sciences to Natural Language Processing (NLP). CC is also discussed in various settings and communication platforms, from academic publications to social media forums. Understanding who and what is mentioned in such data is a first critical step to gaining new insights into CC. We present CLIMATELI (CLIMATe Entity LInking), the first manually annotated CC dataset that links 3,087 entity spans to Wikipedia. Using CLIMATELI (CLIMATe Entity LInking), we evaluate existing entity linking (EL) systems on the CC topic across various genres and propose automated filtering methods for CC entities. We find that the performance of EL models notably lags behind humans at both token and entity levels. Testing within the scope of retaining or excluding non-nominal and/or non-CC entities particularly impacts the models‘ performances.