Problems in Building an Instructable Production System.

The Instructable Production System project is exploring the incremental growth properties of production systems PSs by constructing a generally intelligent problem solving system by gradual external instruction. The present task domain is an abstract job shop, in which finished goods are made from raw materials. We start with a Kernel a small PS of about 200 productions which has the basic capabilities to grow by instruction 1 process a restricted natural language 2 form productions from its input 3 impose PS control conventions on them and 4 perform basic manipulations in its environment. We take the basic computational and representational adequacy of PSs for AI programs as established. This note presents some immediate difficulties that are expected. These derive from the instructional situation 1 The instructor can observe the system in the environment and can communicate with it freely, but cannot examine its internal structure directly 2 Interaction with system is in an external language, analogous to natural language 3 The initiative for interaction is mixed 4 Instruction may be on any topic--specific tasks, general properties of tasks, the language of communication, possible errors, how to plan and explore, etc and 5 Knowledge and system structure gained through instruction accumulates over the life of the system. The current approach uses means-ends analysis as the basic philosophy of both problem-solving and instruction.