I'm surprised they managed to get useful data within one week from a sample of users chosen at random, since presumably the process involved something like:
- monitor messages in & out for a period to establish a baseline
- skew in messages to +ve vibes for a period and monitor out messages
- skew in messages to -ve vibes for a period and monitor out messages
I realise some FB users will generate sufficient data on this basis within 7 days, but what proportion of a random sample ? It would seem to bias any analysis to the most frenetic FB users, but then maybe those are the valuable ones.