Discourse AI - AI feedback counters

Some sites are now using AI to monitor customer forums for information to guage server status. Currently the information is just users posting topics and when it is a more serious problem the number of topics increases as well as the number of me too replies for such topics, or me too replies to past event topics.

To make this process more effective if Discourse created template topics for such edited by the site staff that users could only add reactions this might streamline the process.

For example a software application fails because the database crashes. Users would go to the Discourse forum and find the category for reporting, then find the topic and click a reaction. If the reaction is a server down emoji then after a moving average threshold is meet an event occurs. The event should first try to notify an agent to verify the event is real and if so then notify the support staff.

The purpose of this is to keep down the spontaneous topics for such events and to record in one place over time when the reactions are occuring.

For a more advanced version of this

  • The quality of the user creating the reaction could be weighed then some users would just be ignored because the sky is not falling while other would have a large weight because they then to verify the problem before reacting.
  • If the threshold is triggered a banner could be put up that a problem has been identified.
  • Users could opt-in to receive email when an event occurs or maybe even an SMS message.
  • The AI would understand a new problem topic before it is created and redirect the user to use a reaction on the indicated post.

For those with access see the end of this reply for similar.


EDIT (08/25/2023)

Here is a case where this would be useful.