2 Criteria for a Good Model of Agent Learning

2.3 Examinability of models learnt by the agents

Agent models should be examinable, in the sense that the state of an agent's learning should be open to common-sense interpretation. If this is not the case then we would be able to analyse only the agent's behaviour rather than the process by which it was generated. While this may be realistic for the modelling of agents who model other agents, this is unhelpful from our point of view as modellers seeking to understand the effects of specific approaches to agent learning and expectations formation. The internal states of the agents can be made opaque to other agents without sacrificing our knowledge of those states. Furthermore, if we can choose the extent to which an agent is aware of the internal states of other agents, then we can model and contrast this aspect as well.

Modelling Learning as Modelling - 23 FEB 98

[Next] [Previous] [Up] [Top] [Contents]

Generated with CERN WebMaker