Large language models outperform humans in identifying neuromyths but show sycophantic behavior in applied contexts
While LLMs outperformed humans at identifying isolated neuromyth statements, they struggled to hint users towards the same misconception when they were included in more applied user-like questions-presumably due to LLMs' tendency toward sycophantic responses. This limitation suggests that, despite their potential, LLMs are not yet a reliable safeguard against the spread of neuromyths in educational settings. However, when users explicitly prompt LLMs to correct unsupported assumptions-an approach that may initially seem counterintuitive-this effectively reduced sycophantic responses.
Authors
- Korbinian Moeller (ORCID: https://orcid.org/0000-0003-1385-7778)
- Markus Spitzer (ORCID: https://orcid.org/0000-0001-6040-1939)
- Joshua Weidlich (ORCID: https://orcid.org/0000-0002-1926-5127)
- Luisa Frede
- Eileen Richter
- Annabelle Morgan
Institutions
- Loughborough University (GB)
- Zurich University of Teacher Education (CH)
- Martin Luther University Halle-Wittenberg (DE)
- University of Tübingen (DE)
Publication Details
- Journal
- Trends in Neuroscience and Education
- Published
- 2025-05-06
- DOI
- https://doi.org/10.1016/j.tine.2025.100255
- Citations
- 10
- Primary Topic
- Neuroscience, Education and Cognitive Function
- Type
- article
- Field-Weighted Citation Impact
- 14.31
Funders
- Human Frontier Science Program
- UK Research and Innovation
- Human Frontier Science Program
- Economic and Social Research Council