Recent advances in generative AI, particularly Large Language Models (LLMs), are reshaping how scientific knowledge is created, managed, and communicated. These technologies are evolving into agentic AI systems: autonomous agents capable of planning, reasoning, and acting across multiple stages of the scientific process, from hypothesis generation and experimental design to data analysis and dissemination. This shift offers transformative opportunities to accelerate discovery and reduce researcher workload, but also raises profound challenges related to governance, trustworthiness, and adaptability. However, these systems also disrupt established norms of research data management (RDM), particularly around reproducibility, transparency, and accountability.

This Dagstuhl Seminar examines how RDM infrastructures, practices, and governance models must evolve to support the responsible use of agentic AI in science and industry. A central focus of the seminar is the role of semantic technologies and neuro-symbolic AI in enabling machine-actionable, explainable, and trustworthy research processes. In particular, neuro-symbolic approaches offer promising foundations for combining data-driven learning with symbolic reasoning, structured knowledge, provenance, and domain constraints, thereby supporting more robust and interpretable AI behavior in research settings.

The seminar is guided by two central research questions: (RQ1) : Which tasks in the (FAIR) Research Data Management cycle can be responsibly delegated to agentic AI systems, and under what conditions? (RQ2:) How can we ensure that outputs generated by agentic AI systems are trustworthy, explainable, and accountable to human researchers and institutions?

We will organize discussions along multiple topics, which cover three main pillars: (P1) Trust refers to the need to ensure transparency, traceability, and reproducibility in AI-supported research and development. (P2) Responsibility involves the creation of robust governance frameworks, the definition of ethical norms, and the development of clear authorship conventions. (P3) User-Centric Adaptability focuses on the design of AI systems that respond to the diverse needs of researchers and institutions. The topics discussed will comprise:

• FAIR and Structured Knowledge for Agentic AI Towards Machine-Actionable Research: We will assess current gaps in interoperability, semantic frameworks, and provenance tracking, explore emerging methods for linking FAIR and agentic AI infrastructures, and discuss how structured knowledge can enhance transparency, accountability, and responsible research.
• Trustworthiness in Agentic AI for Research: We will examine how trustworthiness can be ensured in agentic AI systems, incl. data quality, bias mitigation, explainability, and reproducibility, and how neuro-symbolic approaches can support verifiable and interpretable decision-making.
• User-Centric Agentic AI for Supporting Research: We will discuss how user-specific and contextual information can be represented, created, and utilized for agentic AI applications to improve the user experience and personalize services.
• Application of Agentic AI-driven Research: We will discuss, how AI-driven methods can support researchers across the scientific process and which ethical guidelines, best practices, and researcher training are needed to ensure that generative AI, supported by semantic and neuro-symbolic techniques, strengthens the integrity, reliability, and social value of scientific research.

The seminar aims to bring together researchers from AI, databases, Semantic Web, research infrastructures, libraries, and industry to develop a shared research agenda and a roadmap for trustworthy, FAIR-aligned, explainable, and adaptive RDM in the age of agentic and neuro-symbolic AI.

Creative Commons BY 4.0

Maria-Esther Vidal, Sandra Geisler, Cinzia Cappiello, Annette ten Teije, and Tarek Richard Besold

This seminar qualifies for Dagstuhl's LZI Junior Researchers program. Schloss Dagstuhl wishes to enable the participation of junior scientists with a specialisation fitting for this Dagstuhl Seminar, even if they are not on the radar of the organizers. Applications by outstanding junior scientists are possible until September 11, 2026.