We have to maintain both simplicity and reliability in the planning operators. If we can identify those details of the world Dynamics likely to become relevant within our distribution of problems and structure our operators accordingly, we may achieve the goal to certain extent.

Problem generator: Produces planning problem according to some fixed but unknown probabilities.

The planner takes in the problem, along with a library of operators, and output plans, which, when applied to the domain, result in some set of experiences. In this framework, the operators are designed by a domain expert and are immutable.

The domain expert designs the operators, they are kept simple in order to keep the planning phase manageable. Hence there is less specificity of the operators.

The planning process is supplemented by a new operator design module:

1. Explanation sub module: It is used to associate observed world Dynamics with a consistent causal model, calling on explanations from the knowledge base

2. Calibration mechanism: It associates precise numerical functions to the qualitative causal structure.

3. Publication module: Assesses the capabilities of our operators against the distribution of planning problems, and produces a minimal sufficient set of operator definitions for use by the planner.