Suggestion for Improving Thumbs-Up/Thumbs-Down Feedback System

I’d like to suggest an improvement to the thumbs-up and thumbs-down feedback system in ChatGPT. Currently, the buttons are positioned very close to each other, and there is no way to edit or reverse a response once it’s clicked. This can lead to accidental feedback, such as mistakenly clicking the thumbs-down when the thumbs-up was intended.

I believe it would enhance the user experience if:

  1. The thumbs-up and thumbs-down buttons were spaced further apart to reduce the chance of accidental clicks.
  2. There was an option to edit or retract feedback after it has been submitted, allowing users to correct mistakes.

LOL, ChatGPT literally wrote this summary after I made this mistake, oh the irony. Thank you so much, I literally love this tool

I just refresh the tab and the options :+1:t2::-1:t2: become available again. Not sure how legit it works as feedback, but it works for me as end user. :grinning_face:

Please restore separate :+1: and :-1: buttons under each response. Combining them into one menu turns a one-click feedback action into a two-click action for no apparent benefit.

I agree with restoring the separate thumbs-up/thumbs-down buttons.

The old design was better in two ways: it was one-click, and it was instantly readable. If combining the buttons was meant to create a cleaner interface, I don’t think it worked. The new control takes up less space, but the icon itself is tiny and visually complicated.

The separate thumbs-up and thumbs-down symbols are simple, familiar, and almost iconic as a pair—like the happy and sad theater masks. Sometimes two clear symbols are cleaner than one compressed one.

I’d like to suggest a change to the way the thumbs-up / thumbs-down feedback system works in ChatGPT.

The main problem is that thumbs-down currently conflates two very different things:

  1. “This answer is bad/wrong.”

  2. “This answer may be fine, but I don’t want to follow the suggested direction.”

These should not be treated as the same type of feedback.

Suggested solution

When ChatGPT proposes one or more follow-up actions, each suggestion could have a :+1: and :-1: button.

:+1:

A thumbs-up would mean:

Yes, do this.

ChatGPT could immediately execute the proposed action.

:-1:

A thumbs-down would mean:

No, not this direction.

Instead of simply recording negative feedback, clicking :-1: could open a small menu with useful alternatives, for example:

  • This answer doesn’t satisfy me

  • Search further

  • Try something different

  • Give me other options

  • That’s not what I meant

  • Let me explain what I want instead

The last option could simply open the normal input field.

Why this would be useful

ChatGPT frequently ends a response with suggestions such as:

“I can search further.”
“I can compare these options.”
“I can look for another solution.”

In many cases the user wants to accept or reject one of those suggestions, but typing “yes, do that” is unnecessary friction.

More importantly, a thumbs-down should not automatically mean that the response itself was bad.

For example:

“I can search further for the video.”

The user may press :-1: because they don’t want ChatGPT to search further. That doesn’t necessarily mean the preceding answer was bad.

The distinction should therefore be:

:-1: = “Not this proposed direction”

rather than automatically:

:-1: = “Bad answer.”

Multiple suggestions

If ChatGPT presents several possible next steps, each could have its own :+1:/:-1: control.

For example:

Search YouTube further :+1: :-1:
Check B.O.O.S. archive material :+1: :-1:
Look for clips on social media :+1: :-1:

This would allow the user to steer ChatGPT with almost no typing.

I think this could make ChatGPT feel considerably more interactive and reduce unnecessary back-and-forth, while also producing much more meaningful feedback data.

The important point is that feedback on an answer and feedback on a proposed next action are two different things and should be treated separately.

Thanks for the suggestions! We’ll share the requests for clearer feedback buttons, a way to correct accidental votes, and separate controls for choosing next steps. We don’t have a timeline to share, but appreciate the examples.