Ran a test of experimental variation 2 today, tomorrow we're going to move into trying out variation 1. This variation essentially removes 1/3 of the system and keeps just the chat, as a way of comparing outcomes from var 2 and 3 with a standard(ish) synchronous chat.