R0043/2026-04-01/Q001/SRC06
Marburg University study on acquiescence bias in LLMs — surprising counter-finding
Source
| Field |
Value |
| Title |
Acquiescence Bias in Large Language Models |
| Publisher |
arXiv |
| Author(s) |
Daniel Braun |
| Date |
September 2025 |
| URL |
https://arxiv.org/abs/2509.08480 |
| Type |
Research paper |
Summary
| Dimension |
Rating |
| Reliability |
Medium-High |
| Relevance |
Medium |
| Bias: Missing data |
Low risk |
| Bias: Measurement |
Low risk |
| Bias: Selective reporting |
Low risk |
| Bias: Randomization |
N/A -- not an RCT |
| Bias: Protocol deviation |
N/A -- not an RCT |
| Bias: COI/Funding |
Low risk |
Rationale
| Dimension |
Rationale |
| Reliability |
Peer-reviewed arXiv submission with transparent methodology (37,975 question variations, 5 models, 3 languages) |
| Relevance |
Directly tests whether "acquiescence bias" (a survey methodology term) applies to LLMs; relevant to vocabulary mapping |
| Bias flags |
Single-author study from one institution; findings require replication |
| Evidence ID |
Summary |
| SRC06-E01 |
Counter-finding: LLMs show "no" bias (opposite of acquiescence), complicating the vocabulary mapping |