How to stop Claude from giving in too easily in role-play
The is a 144-page law, so many officers need to learn it through realistic situations instead of reading it front to back. Training scenarios were built where Claude acts as a resistant stakeholder, such as a vendor refusing to label medical images, a government ministry doubting detection, or a team lead whose AI published election summaries without labels. The main problem was that Claude backed down too quickly when the learner made a halfway reasonable argument.
The fix was to make Claude keep its position unless the learner cited a specific article, not just a broad principle. The resistance score is tracked on the server instead of inside the prompt, because Claude can adjust its behavior if it can see the score. A separate coach mode gives hints toward the right article but does not reveal the answer directly.
The result is a role-play where the other side feels more like someone who knows the and holds firm on details.
Key points
- Claude was too willing to agree during role-play scenarios.
- Claude only changes position when the learner cites a specific article.
- The resistance score is stored on the server, not shown in the prompt.
- Coach mode gives hints but does not reveal the correct article.
- The design makes regulatory training feel more like a real disagreement.