AI Summary: emojis limited to 👍 and 👎

[tianrunhe]

Issue - Enhancement

To help us better understand how user like or dislike the AI summary, I think instead of providing the full emoji picker for the comment, we limit it to only 👍 and 👎. Furthermore, we put them along the header, sth like this:

In the backend, we could still use the Reactji Interacted metric and simply just record +1 or -1.

This way it's much more streamline and easier to do analysis and would enable us to iterate our GPT models (e.g., A/B testing for different models).

Acceptance Criteria (optional)

• User can vote 👍🏻 or 👎🏻 for AI comment
• When a vote is cast, Reactji Interacted metric is sent to Segment, with value +1 or -1 respectively

Estimated effort: X points (see CONTRIBUTING.md)

[nickoferrall]

Restricting emojis to only 👍 or 👎 would give us a slightly better understanding of whether users like the summaries, but it makes the UI a bit more busy/cluttered IMO. If we had our own custom model that we were training with user feedback, I think it would be worth doing.

But for gauging whether users want the summaries, I think we should focus on whether they toggle off the feature. We have a design issue (#7882) that's being worked on to add that functionality.

[jordanh]

I'd love to see the AC for this expanded slightly (or perhaps better or a follow-on issue added) that adds the option for the user to say why 👎 was pressed. I think it will help us optimize the engineering of these prompts better.

I think we got lucky when a user wrote in to support with bug #8491 ... it would be nice to catch other such cases more reliably

[github-actions]

Stale issue

[tianrunhe]

I still want to do this, so removing the stale label. Might get to it as a side project some time.

[jordanh]

Iceboxing for now as I believe we need to focus on other things. @tianrunhe if you feel otherwise, please reopen