Capturing and implementing undocumented knowledge
Electric Power Research Institute (EPRI) takes on the problems presented by the loss of experienced workers.
The knowledge elicitor should be at least somewhat familiar with the domain. The elicitor(s) may have the required familiarity through previous experience, or he/she may be given time to be bootstrapped into the domain prior to knowledge elicitation. There are several reasons why such domain knowledge is necessary. Most importantly, it permits the knowledge elicitor to understand specialized domain terminology, be able to ask intelligent questions, and have some recognition of the specific areas to probe further to obtain the needed knowledge.
Most knowledge elicitors need guidance regarding the valuable domain knowledge to be captured. This may not be necessary if the elicitor is extremely familiar with the domain and the knowledge that needs to be captured. Without this depth of knowledge, however, he/she may need to rely on someone else for direction.
In many cases, the expert has a wide range of expertise, some of which is unique and some that is also known by others. It should be noted, however, that some workers who are identified as experts may say, “I don't have any valuable knowledge; other people know what I know.” It is not uncommon for an expert not to realize he/she has valuable knowledge not known by anyone else.
Experts are usually extremely busy because they are the ones assigned the most demanding and difficult tasks, and may be consulted by others needing access to their unique knowledge. It may be difficult for the knowledge elicitor to have much time with the expert. Therefore, the time must be used wisely to capture the knowledge that is most valuable and not available to others.
In any event, the expert's manager or other people familiar with the situation should be queried regarding the specific knowledge to elicit. They will have an understanding of the knowledge areas that are important. They will be able to identify the most valuable and needed information that should be collected and subsequently made available to others.
Select Elicitation Methods
Knowledge elicitation efforts usually take place in stages, and the nature of the knowledge is a major consideration in selecting the appropriate elicitation methods. The first stage is for the elicitor to develop an understanding of the general knowledge of importance available to the expert. Methods are available to develop a high-level description or overview of the expert's valuable knowledge, e.g., the concept mapping method. The description created by applying the method can be reviewed with the expert and his/her manager to select areas to drill down to the levels at which the most valuable undocumented knowledge is held.
Following selection of the specific areas of importance, the elicitor may drill down to a deeper level of expertise applying the same methods used to create the high-level overview of the expert's knowledge. Alternatively, another method may be selected that is more appropriate for the nature of the knowledge. For example, if the knowledge is based in large part on significant events occurring in the past, then an interview approach focusing on critical incidents or critical decisions may be appropriate.
Other approaches may be more suited to knowledge that relates to operations and maintenance processes and equipment. Such knowledge may be elicited with the help of simulations and scenarios using mock-ups or actual equipment. The simulations and constructed scenarios method and the think-aloud problem-solving method encourage the expert to describe what he/she is doing and thinking about as he/she performs the simulated or actual tasks. Video or audio recordings and photographs may be taken at appropriate times during the elicitation sessions, edited, indexed, and made available to others when access to the expertise would prove beneficial.
It may be desirable to drill down to a more detailed level of knowledge at certain points during the elicitation process. For example, the expert may report that he/she senses almost unconsciously that something is in alignment, and that one “thing” can be inserted into another. If this capability to perform the action more quickly and better than anyone else has high value, then an unstructured interview approach might be applied to ferret out the important cues that are present. The elicitor may ask about visual, auditory, and tactual cues that are being used, possibly at almost an unconscious level.
Consider Storage, Presentation, and Use
Previous researchers working in the field of expert systems and knowledge management have observed the existence of a knowledge acquisition bottleneck. The knowledge elicitation methods, applied appropriately in the context and situation, can alleviate serious knowledge acquisition bottleneck problems.
Despite such reduction of knowledge acquisition bottlenecks, however, care must be exercised to facilitate the subsequent steps of knowledge storage, presentation, and use. The very methods that can alleviate bottlenecks in knowledge acquisition can create time and effort barriers for subsequent stages of the process.
For example, methods such as structured and unstructured interviews that rely on audio recording of elicitation sessions can create a transcription bottleneck. Transcription, editing, and reviewing audio records of interview sessions are time-consuming activities. Techniques to minimize the editing required to format knowledge for use by others include careful and selective audio recording and, for certain kinds of knowledge capture, use of video recording.
Computer speech recognition might be considered as an approach for avoiding the transcription bottleneck. At this time, the technology is not yet advanced enough to make this approach feasible. Both the elicitor and expert would need to train the speech recognition system in their respective voice patterns, and technical terms not in the speech recognition lexicon would need to be entered prior to the elicitation session. Speech recognition technology is moving ahead rapidly, however, and it may help reduce the transcription problem in the future.