Chat apps encrypt the messages as they are sent over the network preventing you from monitoring them.

This is where correlation would have to occur.

You see your son/daughter upset at 5:30 PM.

Suppose logs show your son/daughter was accessing IPs associated with Facebook/Messenger at 5:28 PM.

One could reasonably infer that something/someone upset them.

How would that work? I can’t see how the surveillance software would distinguish a benign message from a malign one. For example ”you’re so cool” could be either a compliment or an insult, depending on the context.