7 Towards a Specification of a Framework for Modelling Economic Learning by Modelling

7.5 Action selection

Finally the agent has to find an action, represented by a single point of all the possibilities, i.e. find

. This may not be possible so a default action must be also designated. If the nature of M is suitably direct and the space PS is also a language closely related to L, then it may be possible to deduce x from

and G.

The agent has to do this expending as little effort as possible. Sometimes the goal statements will interact with the model statements and the present data, as when implementing a minimax strategy. At other times the goal statements will be independent, in which case the agent needs to find: which is easier. The task of finding an action when the goal specifies a unique action (as in maximizing situations), will usually be computationally onerous for realistic agents.

Modelling Learning as Modelling - 23 FEB 98

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

Generated with CERN WebMaker