Someone recently liked one of my posts in this topic, which prompted me to re-read it.
Re-reading a discussion from 7 years ago can be a strange and fun experience, both for the things you’ve forgotten you said, and also what the passage of time brings (e.g. I’m not sure that HAWK and Dave were working for Discourse in 2017, and are now, respectively co-CEO and Product Manager. This was also before I started Pavilion). One might say that the “discussion archeology” opportunities that forums afford are one of their strengths.
Relatedly, I’ve been meaning to look into the recent addition of the “hot” topic list (added in this commit), and re-reading the above gave some impetus. The notes in the code provide a useful equation for the hot algorithm:
(total likes - 1) / (age in hours + 2) ^ gravity
There’s some discussion of this addition in the topic announcing it
I want to preface my comments on the hot topic list with the observation that they’re essentially thought bubbles and that I agree with Lindsey’s overview of how the product utility of the Hot list (or any other topic list) should be determined:
That said, here are some things that come to mind out of the intersection of this old discussion and the new Hot topics list:
-
I wonder if “taxonomies” could be optionally used to weight the hot algorithm, i.e. a certain category or tag may get a lot of activity (e.g. likes), but you (the forum admin) consider that “shallower”, or less desirable in some way, than activity in another category or tag. In other words, adding in more site-specific discovery curation.
-
I wonder what a topic list contextualised to user variables (e.g. trust level, the prior topic views of that user) would look like / how it would behave. In other words, adding in more user-specific automated / algorithmic curation.
In other words: