It would be interesting to do a sentiment analysis on users to see if users themselves are becoming more negative over time or if negative voices are composing

It would be interesting to do a sentiment analysis on users to see if users themselves are becoming more negative over time or if negative voices are composing more of the message volume. With the degree of data available I have wondered if you could determine a point at which you could suggest users see professional help. Then if you could do such a thing, would it be ethical to do so? Would it be ethical _not_ to do so?