Suggestion to improve the feedback system in OpenAI
Hello, OpenAI team.
I would like to make a suggestion regarding the way you collect user feedback when presenting two possible responses and asking to choose the best. One issue with this method is that many users may only read one of the two options and select one without actually comparing both, which introduces a significant margin of error in the evaluation.
My proposal is that, instead of asking users to choose between two responses, they be asked to rate each one on a scale from 1 to 10. This way, you could obtain a more accurate analysis of which is better and why, reducing the bias of quick selection.
Thank you for your work and for considering this suggestion.