Working on a new bot, this one is physical and it has facial recognition.
At any given time, X number of people will be identified by facial recognition.
The closest face will be the primary speaker.
I can get this information to CS, which would basically be a list in priority order.
Primary face X1
2nd face X2…..
N face Xn

The challenge i have is this.  In real life, context is everything. 
And CS has a great way of keeping track of a conversation with multiple individuals, individually. 
For me, this is realized by giving each device a specific identity and using that as the CS user. So anything coming from a specific device is coming from the same person. 

But in this case, the user may change.  And we may have multiple users listening.
One being a primary user or the one who is talking. 

I am stuck because suddenly it is very complex and I am not sure how to organize this within CS code.
Has anyone else run into this and I would appreciate some insight into a good approach managing a response that could include one or many end users that changes. 

I’m not quite clear on the relationship of user to device to CS.
Would several users take part in the same conversation with one device, or would each user, if they spoke, have their own conversation?

The CS user can be anything, ours are actually a combination of two large UUIDS - one from a user identity and one from a “bot” identity. So a person’s topic file is very targeted to both who they are and who they are talking to.

CS would be running on a physical robot.

Suppose people come in and out of the room. 
It starts with Andy. Then Mary joins the room, then Bruce, then Mary leaves.

Visually, we give each person a tag number, based on facial recognition software.
I have a routine that keeps track of who is in the room currently and who has left. 
The robot assumes that the person who is closest to it is the person talking to him.
However, people come and go from the room.

I can switch the current CS User when someone new is closer to the robot.
But this might start the conversation over again. And this would create a disjointed and unconnected experience. 

I am not sure how to manage this within CS.  I do not see a way to connect many people and group configurations to one CS.
Stuck on the design architecture.

I received a reply from Bruce, and I wanted to share it, in case someone else might have this challenge. THANK YOU, BRUCE WILCOX!

Bruce: CS supports having a specific bot AS well as a general bot that can share in a conversation. So you can record information “heard” by all (or you can index the hearing to a list of names of who heard it) and have a topic file that knows the specifics of a particular user.  We used to have a demo product wherein the user (a child) could listen in on a conversation between the smart pig and the wolf. The two talked to each other and the child could ask questions of either (and either might reply potentially).  topic definition keyword “share”