System designers must choose between capturing knowledge themselves and defining knowledge acquisition tools. Experts often have little patience for unfamiliar detail. Failure to get them directly involved in knowledge capture can limit the rate at which a system's ability grows.
Much of Sinapse's knowledge is coded in MMa notation for defining trees or as MMa procedures. This required that encoders know MMa well, and made it difficult to prevent some low-level implementation details of Sinapse from leaking into the components. Attempts to get wave propagation experts, who were expert parallel machine programmers, directly involved in the knowledge encoding task were not very successful because of the high cost of learning the system. The system designers have had to substitute for the experts when encoding the knowledge, making the designers a bottleneck in the acquisition process.
Getting the experts involved requires that they be taught basic principles of knowledge representation and domain analysis. The system designers should spend their energy building knowledge acquisition tools, to allowing the experts to easily find out what the system already knows, and revise that knowledge, using notational schemes which are familiar [#!Marcus88a:SALT-ka-for-propose-and-revise-systems!#], [#!Boose90a:knowledge-acquisition-foundations!#].