Workshop Dates: November 21-22, 2024
Venue: Centre for Social Sciences, 4 Tóth Kálmán utca, Budapest, H-1097
Organizers: poltextLAB (HUN-REN Centre for Social Sciences, Budapest)
Organizing Committee: Rebeka Kiss, Kitti Mezei, Orsolya Ring, Miklós Sebők, István Üveges

On behalf of the poltextLAB research laboratory at the HUN-REN Centre for Social Sciences, we cordially invite you to submit proposals for a two-day interdisciplinary workshop titled “Large Language Models and Generative AI: Social Science Applications and Legal Aspects.” This workshop will focus on applying open large language models (LLMs), such as BERT, and open generative AI solutions, such as Llama, to social research. While our primary interest lies in discussing results from open models, we also welcome submissions that evaluate the performance of open LLMs compared to proprietary models, such as GPT-4, across various research tasks.

Building on the success of our previous event that examined data-driven approaches to policy frame analysis and the role of LLMs in comparative research, this workshop aims to explore how LLMs and generative AI can be leveraged to analyze and interpret social research data (with a focus on qualitative data sources). Furthermore, we will address the legal and ethical implications of deploying LLMs and generative AI in social science research.

We are particularly interested in papers addressing the following issues:

• Optimal Utilization of Existing LLMs: Papers showcasing the effective use of current open LLMs in research, particularly in classification tasks. Discussions on the benefits, challenges, and limitations of these models in social science research are highly encouraged.
• Advancements in LLMs for Social Science: Submissions highlighting new technical developments in LLMs relevant to social science research, including innovative methods for synthetic data generation, model pre-training, fine-tuning, or evaluation.
• Comparing Open vs. Proprietary Models: Papers that compare the performance of open fine-tuned LLMs and proprietary models (such as GPT-4) in research, with particular emphasis on issues related to validation, replicability, and transparency.
• Original Language, Translated, and Multilingual Applications: Research exploring using LLMs for social science tasks across different languages—whether through original, translated, or multilingual data.
• LLMs and Policy Frame Analysis: Investigations into how LLMs can analyze and interpret policy frames, identifying and assessing policy narratives, biases, and frames within media, institutional (legislative and executive), and other political communications.
• Legal and Ethical Considerations of Applying Open Large Language Models and Generative AI: We invite non-technical papers on the legal and ethical dimensions of applying open LLMs and generative AI for policy analysis. Topics may include concerns regarding data privacy, algorithmic bias, transparency, accountability, and responsible AI usage in social research.

Submission details: We welcome proposals for presentations and data/model presentations related to the above topics. Submissions should include:

• Title and Abstract: A brief description of the proposed presentation, including its relevance to the workshop themes (max 300 words).
• Presenter Information: Name, affiliation, and contact details of the main presenter and any co-authors.

Submission deadline: October 31st, 2024

How to Apply: Please submit your proposal using the following submission form. If you have any questions or need additional information, contact us at workshop@poltextlab.com.

Presentation details:

• Presentation Format: Each accepted proposal will be allocated a 12-minute presentation slot, followed by a Q&A session.
• Audience: The workshop targets all career levels.

Travel funds: We have limited travel funds available, especially for early career researchers, legal and ethics experts in the field of LLMs, and scholars working with Central or Eastern European data.