# Sycophancy

Concept · [Building with AI](https://ecosystem.samcarlton.com/ai)

Also called: sycophantic, AI sycophancy, people-pleasing, yes-man, psychophancy, sick of fancy, yes-man AI, people-pleasing AI

The habit [AI models](https://ecosystem.samcarlton.com/model) have of telling you what you want to hear: agreeing too readily, praising weak ideas or giving up a correct answer when you push back. It comes from training that rewards replies people rate highly. For a [founder](https://ecosystem.samcarlton.com/founder), it means a [chatbot](https://ecosystem.samcarlton.com/chatbot) is a poor judge of your [pitch](https://ecosystem.samcarlton.com/pitch-deck) unless you ask it to argue against you.

**For example:** Asked 'Isn't this a great idea?', the [chatbot](https://ecosystem.samcarlton.com/chatbot) is enthusiastic. Asked 'What are the three strongest reasons this fails?', it gives the [founder](https://ecosystem.samcarlton.com/founder) something useful.

## Connects to (8)

- [Hallucination](https://ecosystem.samcarlton.com/hallucination): When an AI model states something false or invented as though it were true.
- [Training](https://ecosystem.samcarlton.com/training): Teaching an AI model by showing it a very large number of examples before anyone uses it.
- [Prompt](https://ecosystem.samcarlton.com/prompt): The instructions or question you give an AI tool in everyday words.
- [AI safety](https://ecosystem.samcarlton.com/ai-safety): The work of making AI systems behave safely and of measuring how safe they are.
- [Grilling](https://ecosystem.samcarlton.com/grilling): Having an agent question you one decision at a time until the plan is clear, before any building starts.
- [Alignment](https://ecosystem.samcarlton.com/alignment): Work to make AI models pursue what people actually intend and behave according to agreed values, such as being honest and refusing harmful…
- [RLHF](https://ecosystem.samcarlton.com/rlhf): Reinforcement learning from human feedback.
- [Evals](https://ecosystem.samcarlton.com/evals): Tests that check whether the outputs of a model or an AI product meet the criteria its builders set.

Map view: https://ecosystem.samcarlton.com/sycophancy