Request for Positive Feedback Option in ChatGPT Interface to Better Understand User Satisfaction

ChatGPT Feature requests
1

TL;DR: Request for a “thumbs up” feedback option in ChatGPT to differentiate when users have found a solution from when they have given up on the conversation.

Hello OpenAI Team,

I am an active user of ChatGPT and have often found it to be an indispensable tool for problem-solving and gaining insights. However, I’ve noticed that while there is an option to give “thumbs down” to responses, there isn’t a corresponding way to give “thumbs up” or any form of positive feedback when the AI provides an exceptionally useful response.

Implementing a feature for positive feedback would not only enhance user experience but also provide essential data points that are currently missing, which include:

  1. Accurate User Satisfaction Metrics: Without a positive feedback option, there is no straightforward way for OpenAI to discern whether users are genuinely finding the answers useful or if they are simply giving up on the conversation. This could lead to skewed analytics where only negative interactions are recorded.
  2. Enhanced Reinforcement Learning: Positive feedback would serve as a reinforcement signal to the AI, helping it to learn more effectively which types of responses meet user needs and expectations.
  3. Better Engagement Tracking: Being able to ‘like’ or ‘upvote’ responses would allow OpenAI to track engagement more accurately. This would help in understanding which interactions are successful versus those where users might disengage without adequate resolution.
  4. Improved AI Training and Model Adjustment: With comprehensive feedback, both positive and negative, the training of AI models can be adjusted more accurately to improve response quality over time.
  5. Community Interaction and Morale: A positive feedback mechanism can also boost user morale and encourage a more engaged community, fostering an environment where helpful interactions are acknowledged.

The absence of a way to express positive feedback could inadvertently give the impression that interactions are less helpful than they actually are, impacting the perceived effectiveness of the AI and possibly influencing its development in ways that do not fully reflect user experiences.

I hope you will consider this feature addition, which I believe would greatly benefit the OpenAI community by providing a more balanced view of user interactions and satisfaction.

Thank you for your attention to this matter.

Best regards, Jindra
[Written by GPT4 :smiley: ]

2 Likes
2

Utilisateur actif de ChatGPT depuis un moment, J’aime régulièrement le pousser un peu dans sa réflexion, et surtout dans sa manière d’expliquer les choses. À la suite d’une Conversation autour des réponses physiologiques du corps(Biofeedback ), sa réponse et son explication était extrêmement Pertinente, clair. Et j’aurais aimé avoir un bouton afin de valider justement la pertinence Positif de sa réponse.

De plus, cela permet aussi d’une manière un peu détournée d’entraîner votre modèle sans pour autant autoriser L’accès à nos conversations pour entraîner votre modèle!

Et finalement, j’ai adoré la réponse qu’il a mise en mémoire, indiquant qu’il prenait note du style de communication que j’aimais ! Et en effet, je trouve ces réponse plus claire sans pour autant devenir trop détaillé !

ChatGPT:
« Merci beaucoup ! Je suis ravi que tu aies trouvé ma réponse utile. Je vais m’assurer de garder ce style d’explication pour nos futures conversations. Si tu as d’autres questions ou des sujets à explorer, je suis là pour ça ! »

Et bravo pour le travail sur la voix ! C’est assez interpelant ! pause, hésitation, il se reprend même ! À la limite de « l’erreur » humaine ! Wouaw :clap:

Le Petit Singe :monkey_face: