ESRA 2027 sessions by theme
Back to Overview of Sessions
Telling Slop from Saviour: AI Literacy and Practice in Research Data Management |
| Coordinator 1 | Dr Frederic Gerdon (GESIS – Leibniz Institute for the Social Sciences) |
| Coordinator 2 | Ms Ekaterina Chicherina (GESIS – Leibniz Institute for the Social Sciences) |
| Coordinator 3 | Dr Lisa Spitzer (Leibniz Institute for Psychology (ZPID)) |
| Coordinator 4 | Dr Deborah Wiltshire (GESIS – Leibniz Institute for the Social Sciences) |
| Coordinator 5 | Mr Knut Wenzig (DIW Berlin/SOEP) |
| Coordinator 6 | Mr Daniel Bela (LIfBi) |
| Coordinator 7 | Dr Arne Bethmann (SHARE Germany and SHARE Berlin Institute) |
Session Details
AI-based tools, such as Large Language Models, have entered research data management at remarkable speed. They promise to relieve bottlenecks: drafting variable descriptions, mapping legacy documentation onto standards such as DDI or SDMX, generating structured metadata from unstructured sources, supporting anonymization, and preparing FAIR-compliant documentation. They also produce plausible but wrong output, introduce inconsistencies at scale, may threaten privacy, and flood shared infrastructures with confident-sounding "slop". Making sensible and safe use of such tools thus requires knowledge and skills regarding how exactly these technologies work, how to use them most efficiently, and how to reduce risks and avoid harms.
This session asks (1) where AI genuinely helps and where it merely appears to do so in RDM and (2) how to train people to make sensible use of AI. Importantly, answering these questions requires the needs and experiences of researchers as well as RDM professionals. Potential topics for contributions to this session include (among others):
- Which lifecycle stages AI meaningfully accelerates, and which competencies make that acceleration real rather than apparent
- How AI output is validated against standards such as DDI and SDMX, and who remains accountable when it is not
- How to guard against scale-amplified errors and the degradation of trust in shared (meta)data
- What AI literacy means in practice: role-specific profiles, the influence of AI policy, and how much automation data protection permits
- How these competencies are taught, institutionally embedded, and measured, and how to get people to use the training at all
We welcome reports of failure alongside success, and methodological reflection alongside technical work. The aim is a realistic view of what AI can contribute, and of what we must be able to do ourselves.
