I was thinking along those lines too, but your solution is a bit too invasive. Instead, what if we analyzed the peripheral data generated by a person and ask t

I was thinking along those lines too, but your solution is a bit too invasive. Instead, what if we analyzed the peripheral data generated by a person and ask them to tag something to it in the app? There's this project called stereomood (see: http://www.stereomood.com/ http://www.stereomood.com/) that's doing something similar. The thing is that people tend to do certain groups of behaviors in certain moods. So, if you can associate the two and then ask for confirmation or a response along those lines. You can subtly check on them without influencing them. If we ask them a question in these lines then we tend to get false feedback. Even if they aren't feeling that way we might trigger a response and make them feel that way subconsciously. Or, even worse we might become an annoyance and that person might stop replying properly to us. Bad data is to expected but it must be minimized. Even cooler would be to mine this data further and create a profile of the person. So, that their psychiatrist/psychologist/counselor can better help them. After all there's only so much you can say in the span of an hour. I wouldn't be surprised that a startup comes up trying to do just that.