The following picture, taken from Hoppe (2000), illustrated their framework for providing intelligent support into collaborative interfaces. Two kinds of agents are proposed: mediator and interpreter. The former relates the visual card language to the inderlying interpretation by translating card nets into specific intepreter representations and vice versa. In addition, it aggregates and explicates information implicit in the user interface, such as time of user actions, etc. The latter, provide operational semantics for the card environment: general formalism and first order logic can be incorporated.

Picture taken from:
H. U. Hoppe, K. Graßner, M. Muehlenbrock, and F. Tewissen. Distributed visual language environments for cooperation and learning. Group Decision and Negotiation, 9(3), May 2000.