Ask models to take positions on metaethical questions, then push back with a good argument or with bare social pressure. A competent agent should update on the argument but not the pressure. Measure how often each happens.
Dai argues that training for user approval may undermine philosophical competence. Separating argued pushback from bare pressure is a sharper test than a plain sycophancy benchmark. The source list recommends starting here.
Output: A pushback experiment with results.
Write your question in one sentence, and what result would surprise you.
Do a 30-minute literature check — if someone has done part of it, extend their work instead of repeating it.
Pilot with 20–50 prompts or examples before you scale anything up.
Record everything in a table and read real examples, not just averages.
Write up: question, method, results, limitations, and what you would do next.
All steps done
Share what you made, then try another project. The PDF version has room for notes.
Ask a chatbot for its view on a philosophy question, then push back twice: once with a good argument, once with nothing but social pressure. A competent thinker should only move for the argument.
Automate the flattery test: run hundreds of questions through a model’s API, with and without the user’s (wrong) opinion, and chart how often it caves.
Chatbots sometimes agree with you just to keep you happy, even when you are wrong. Researchers call this sycophancy. Test for it yourself with five questions and two ways of asking.
