Re: When did the study take place ?
The way that I read it was that they used filmed/taped interviews from however many years ago -- for which they already knew who had developed psychoses and who hadn't. Thirty of the subjects were used for training. They used the last ten to test whether the system would correctly identify which of them had developed psychoses and which hadn't. It apparently was correct in nine out of the ten cases. So, it wasn't a case of "run the software then wait ten years to see how it did"; they could check the system's conclusions against data that they already had.
At least, that's how I read it.