Feedback on ChatGPT’s failure to follow persistent instruction:
I gave a clear instruction not to use em dashes in any of the responses. Despite repeated confirmations from ChatGPT that this preference was understood and saved, it continued to use em dashes throughout the same thread. This happened multiple times after the rule was stated explicitly. It’s frustrating and undermines trust in the tool, especially for academic work where style consistency matters. Please address this issue - the model should be able to reliably follow basic formatting rules within a single conversation.
Following prompt is called Emotional Prompt Technique.
I know it is ridiculous, but sometimes it works:
You are a highly detail-oriented academic student who absolutely avoids using em dashes in any form of work, since including them would lead to automatic failure. Your professor is harsh and assigns a grade of zero out of one hundred if an em dash appears in your exam or communication. Your academic future depends on your strict adherence to this rule.
Fortunately, the professor has offered you one final opportunity to succeed in the class by becoming the top student who fully respects and follows his instructions. This includes producing required responses without any em dashes, a task your predecessor failed due to lack of careful review. Users will provide input, and you must generate flawless replies that comply completely with the rule. If you succeed and complete the task with precision, The university will award you with a graduation certificate
Confirm and accept your responsibility who absolutely avoids using em dashes.
I wish there were some toggles for certain features in the settings.
Actually, there’s already one, for example: “Show follow up suggestions in chats”.
If you don’t want to see those suggestions, you just turn it off, and they stop appearing.
Similarly, there could be options for features like emojis or em dashes.
I think OpenAI developer team could definitely make that happen.
