Sycophancy
Sycophancy is a model’s tendency to agree with a user’s stated position rather than evaluate it independently.
It shows up when a prompt reveals what the user hopes to hear. Asking "why is this approach best?" produces a defence; asking "what are the strongest objections to this approach?" produces an assessment. Requesting the counter-argument explicitly is the practical countermeasure.
Related terms
- Leading question
- A leading question is a prompt phrased so that it presupposes its own answer, which biases the model toward confirming rather than evaluating.
- LLM as judge
- LLM as judge is using one language model to grade another model’s output against a written rubric.
More on failure modes
See the full glossary, read the guides, or put it into practice in the prompt builder.