CivilServant supports communities on reddit to conduct your own A/B tests on the effects of moderation practices, and share those results to an open repository of moderation experiments.
Add a way to specifically record the fact and details of experiments, probably in a database model
Decide if you want custom code/scheduled jobs per experiment type, or whether you should handle experiments through a pipeline with conditional logic on what to do under what conditions
Remember to set the permissions requested with PRAW.get_authorize_url