I believe that's the purpose of flagging the user for a human review as described. There's a big difference between being nice to a kid, and being nice, then asking for location info, pictures, etc which could be grooming behavior.
Here's an example. Earlier today I was on a python forum and a kid was posting about learning Dijkstra's algorithm. So I sent the kid a message praising him for tackling it (because after learning it like 10x, it just won't stick with me). I told him/her that if they really like the algo, they should look into graph databases because pairing the two is extremely powerful, and potentially very lucrative. Kid wrote back that it meant a lot, that parents don't give them a lot of support on the things they like, etc. So I gave some guidance about other applications, told the kid I'd been through some similar things, and to keep doing their thing and let their interests guide them. Benign.
Now, if after the kid told me they were a kid I started asking them for pictures, where they lived, what they liked, etc, that's a bit more concerning, right? Especially since I'm asking those things after the kid told me they were a kid.
In either case, if the conversation was flagged, it would be pretty obvious one is benign, one is questionable and potentially warrants a closer look. If the reviewer looks at my other posts and sees a pattern of behavior, then it's probably safe to raise the red flag on me, you know?