next up previous contents
Next: The game tree Up: Example 1: Risking misinterpretation Previous: Plan library   Contents

Demonstrations

In the series of demonstrations that follow, the game tree is taken one piece at a time, showing how the strategy of the system varies with the belief model, and finally comparing the system with each of the fixed strategies for the problem. Two parameters are important. First, the intention rule at level 3 is varied to represent the level of risk in asking for the car-spanner. This is because the second agent is expected to use the rules at this level in recognising the first agent's plan, and in particular, to decide which spanner to lend to him. This variable is called $ p$. Second, since the sampler is employed in the demonstrations (see Section 3.4.8), the stereotype error with which the agent has estimated $ p$ is also varied. This is represented by $ n$, the number of dialogues on which the belief model has been trained. A range of values, 2, 8, 32, and 128, was used for $ n$. In the demonstrations, the first agent enters the dialogue intending to use the car-spanner.



Subsections
next up previous contents
Next: The game tree Up: Example 1: Risking misinterpretation Previous: Plan library   Contents
bmceleney 2006-12-19