Finished reading an article on Spatial Cognition by A. Frank. The idea he had was to ask an agent1 to constructing a map replicating reality. Agent2 had to use that map to navigate the physical space. From the measurement of correctness and effectiveness of the agent2 interaction in the environment he could define the goodness of the realised map by agent1.

He constructed, therefore a three stage model of the situation as follows: Level 1: Reality, what is for real in the environment; Level 2: Beliefs, what the agent 1 and 2 think of the perceived reality; Level 3: Map, what the agent 1 has perceived and decided to be represented of the reality with the aim of guiding agent 2.

The author describe the two-tiered reality and beliefs model in which facts describing the simulated environment and the simulated agent’s beliefs of this environment are separated.

I wonder if it is possible to substitute one of the agent with a human using the system, namely with agent2. We can invent then a treasure-hunt task in the city environment where one or more “human” has to follows a series of steps to find an object. The map is defined and shared between agent1 and the participant/s. Agent1 has also the ability to see how his humans partner/s are using the map to retrieve the messages and find the treasure. Then his goal would be to evaluate the communication/interaction between the peers.

I advocate here for a different epistemological strategy of supporting learning: instead of talking of mirroring or guiding systems where the interaction is regulated through computational means, I propose a “coupling system”, where there in no model of correct interaction underneath but the system just stick of the users interactions, offering hooks to relevant information when needed. The system/agent and the humans forms a multi-agent distributed system.

The pictures are taken from: A. U. Frank. Spatial Cognition II (International Workshop on Maps and Diagrammatical Representations of the Environment, Hamburg, August 1999), volume 1849 of Lecture Notes in Computer Science, chapter Spatial Communication with Maps: Defining the Correctness of Maps Using a Multi-Agent Simulation, pages 80–99. Springer-Verlag, Berlin, 2000. [url]