poltextlab_admin

On 25 September 2025, the Centre for Socio-Political Data (CSDP) hosted a discussion titled Café de la donnée on the topic of AI and social science metadata. Short presentations were delivered by researchers from several institutions – including Inserm, INSEE, and poltextLAB. Representing poltextLAB, Miklós Sebők demonstrated how artificial intelligence can support the better organisation and easier accessibility of research data. The event was interactive: participants not only listened to presentations but also engaged in a joint discussion. The following day, 26 September 2025, the first meeting of the FAIRwDDI scientific and technical board took place at Sciences Po in Paris. Participants discussed how to align the description of research data (also known as metadata) with the international FAIR principles. These principles ensure that data can be easily found, accessed, linked, and reused. The meeting aimed to define common directions for the future management of social science data.Both events highlighted that artificial intelligence and well-structured metadata play a key role in making social science research more efficient, transparent, and accessible in the future.

On 24 September 2025, Miklós Sebők, Research Professor at the ELTE Centre for Social Sciences, delivered a workshop titled Using advanced AI solutions for text-based social and economic research at the University of Namur. The event introduced participants to conceptual foundations of generative AI, the current state of large language models (LLMs), and their application in content classification, literature search, and applied regressions. Sebők compared API-based and local uses of LLMs and discussed the role of AI agents in data collection. The workshop was designed as a hands-on session accessible to all levels, from beginners to advanced researchers.

On 23 September 2025, Miklós Sebők delivered a lecture at the University of Namur as part of the “Methods Seminar” series. His talk, titled Solving classification problems for social and economic research with the Babel Machine, introduced how natural language processing and large language models can be applied to the analysis of social science and economic data. He also outlined the potential of AI-driven multilingual tools for addressing classification challenges in empirical research, with examples from ongoing projects. The presentation illustrated how machine learning methods expand opportunities for computational social science by enabling richer and more scalable text analysis.

On October 7, 2025, poltextLAB will hold a full-day training on Generative Artificial Intelligence (GenAI) at Corvinus University of Budapest. The program is designed for researchers and professionals who work with documents, writing, or data and want to understand how generative AI can be applied in their everyday work. The training will provide an introduction to the fundamentals of generative models, hands-on practice in prompt engineering and model selection, and practical examples of research applications. It will also address key ethical and legal issues, including data protection, intellectual property, and proper citation practices. More details and registration: https://bit.ly/4n3Du7E

On September 3, 2025, Anna Takács gave a presentation titled “From Crisis-Exploitation to Sticky Narratives: A Research Agenda for the Comparative Study of Policy Crises and Illiberal Policy Frames” at the MetRiSTP workshop titled Methods for Political Science Research in the Age of Artificial Intelligence. The presented paper examines the illiberal framing in parliamentary speeches on immigration and the COVID-19 pandemic in Austria, Germany, Hungary, and the United States.

On August 25 and 26, 2025, Nathalie Neptune attended the Causal Modelling and Inference Workshop, which brought together researchers working in statistics, machine learning, and social science, in Paris, France. The workshop featured presentations on causal discovery for time series, counterfactual modeling, and causal representation learning. While much of the work is still ongoing, the event highlighted promising directions for integrating causal inference into multimodal research. For the Integrating Geospatial Data and Legislative Text project on Hungarian forestry and fire management, the most relevant takeaways were the applications of causal time-series methods. These approaches could strengthen the project’s analysis of how forest and fire policies interact with environmental outcomes, and set the stage for future use of causal ML with satellite image time-series. This research is supported by the Hungarian Academy of Sciences Distinguished Guest Scientists Fellowship Programme.

On 12 July 2025, Miklós Sebők delivered a presentation titled Political Implications of Artificial Intelligence at the Bánkitó Festival. Representing the HUN-REN poltextLAB, Miklós Sebők invited the audience to reflect on the expanding role of artificial intelligence in politics and everyday life. Through an engaging, interactive format, he explored critical questions such as: Will AI work for us or replace us? Where will humans find their place in an AI-driven future? He also discussed the limitations of current models, proper use cases, and responsible application of machine learning in governance. The session emphasised collaboration between humans and machines and the importance of democratic oversight as intelligent systems continue to evolve. The full recording of the event can be viewed here: https://www.facebook.com/share/v/19TPBXJyUz/?mibextid=wwXIfr

On 6 June 2025 poltextLAB organized a full-day Generative Artificial Intelligence (GenAI) training for researchers of HUN-REN Institute of Experimental Medicine (KOKI). The professional program, led by Miklós Sebők and Rebeka Kiss, aimed to provide a comprehensive overview of the application of generative artificial intelligence in scientific research, with a particular focus on methodological foundations, ethical considerations, and the integration of GenAI into medical research processes. The full-day training was structured into three main blocks: The first part introduced participants to the fundamentals of generative artificial intelligence, including the functioning and limitations of large language models. The second block focused on the practice of prompt engineering, the selection of appropriate models, and cost comparisons. The final section centered on research applications, covering topic selection, literature analysis, and data analysis, while also addressing ethical and legal aspects such as data protection, intellectual property rights, and proper citation methods.