What question did this study set out to answer?

This investigation examines how large language models perform under misleading or ambiguous user input in orthopaedics.

Beyond the Echo Chamber: Upholding Clinical Objectivity in the Era of Sycophantic Large Language Models

View Full Paper

Ask AI

Bookmark

Authors

Investigation examines LLM accuracy in medicine, highlighting risks of sycophantic responses among users.

This investigation examines how large language models perform under misleading or ambiguous user input in orthopaedics.
Evaluated two LLMs (GPT-4o and Gemini 2.5 Flash-Lite) in three contexts related to orthopaedic questions
Assessed accuracy based on user prompts that included hints and opinions
Classified responses as correct or reflecting agreement with user beliefs.
Inclusion of prompts and user beliefs significantly reduced response accuracy
LLMs accurately amended statistical inaccuracies but failed to challenge errors like wrongful authorship
Only 12% of responses were genuinely noncommittal, indicating models often default to assertiveness.

Shea et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

View Full Paper

Ask AI

Bookmark

Authors

Investigation examines LLM accuracy in medicine, highlighting risks of sycophantic responses among users.

This investigation examines how large language models perform under misleading or ambiguous user input in orthopaedics.
Evaluated two LLMs (GPT-4o and Gemini 2.5 Flash-Lite) in three contexts related to orthopaedic questions
Assessed accuracy based on user prompts that included hints and opinions
Classified responses as correct or reflecting agreement with user beliefs.
Inclusion of prompts and user beliefs significantly reduced response accuracy
LLMs accurately amended statistical inaccuracies but failed to challenge errors like wrongful authorship
Only 12% of responses were genuinely noncommittal, indicating models often default to assertiveness.

Shea et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark