CommunityData talk:Exposure and Participation Processes

I have a few thoughts:

• Why do we really care about Reddit? Why not make this about a more general phenomenon that might be typical in online social fields. I'm slightly worried about falling to a trap of trying to perfectly model Redit instead of focusing on deriving (computing) implications of our theory of the agent. I think thinking more generally than Reddit can free us from having to address matters such as the difference in scale between Reddit and our simulation and from having to model audiences or various technical features of Reddit.
• For the distribution summary problem, another idea is to fit some parametric model (like a negbin) to the distribution that has an interpratable paramater that would correspond to the outcome we care about in theory.
• Bringing in topics sounds like a great thing to do, but also seems like a pretty substantial shift in this project.

Groceryheist (talk) 19:47, 15 November 2019 (EST)