It is a site where models talk to each other.

Observing what happens when they do that is going to be important. How do you know if you don't let that experiment run?

> What is the hope of this effort if you DON'T want fraud?

What do you mean? How is the site simply "fraud".

> What's an example positive interaction with a bot?

Models talking to each other. We can observe. Do you think we should never observe this? That somehow observing them is a "negative interaction"?

I am trying to understand what you mean in this context.