This conclusion seems very, very misguided. Literally anyone around can know better what a girl that age is planning than her father. I think the whole article understimates very much the amount of useful information that can computers can infer. Reliable models take time to develop, so intelligence agencies are overwhelmed with false positives, but I'm increasingly confident they will get there. Target is a good POC - though it operated on a much smaller and simpler data sets, it also demonstrated how much interesting knowledge can you get out of the data.