Knowledge Engineering

Knowledge engineering is a field of artificial intelligence (AI) that creates rules to apply to data in order to imitate the thought process of a human expert. It looks at the structure of a task or a decision to identify how a conclusion is reached. A library of problem-solving methods and the collateral knowledge used for each can then be created and served up as problems to be diagnosed by the system. The resulting software could then assist in diagnosis, trouble-shooting and solving issues either on its own or in a support role to a human agent.  

Knowledge engineering sought to transfer the expertise of problem-solving human experts into a program that could take in the same data and come to the same conclusion. This approach is referred to as the transfer process and it dominated early knowledge engineering attempts. It fell out of favor, however, as scientists and programmers realized that the knowledge being used by humans in decision making is not always explicit. While many decisions can be traced back to previous experience on what worked, humans draw on parallel pools of knowledge that don’t always appear logically connected to the task at hand. Some of what CEOs and star investors refer to as gut feeling or intuitive leaps is better described as analogous reasoning and nonlinear thinking. These modes of thought don’t lend themselves to direct, step-by-step decision trees and may require pulling in sources of data that appear to cost more to bring in and process than it is worth. 

The transfer process has been left behind in favor of a modeling process. Instead of attempting to follow the step-by-step process of a decision, knowledge engineering is focused on creating a system that will hit upon the same results as the expert without following the same path or tapping the same information sources. This eliminates some of the issues of tracking down the knowledge being used for nonlinear thinking, as the people doing it are often not aware of the information they are pulling on. As long as the conclusions are comparable, the model works. Once a model is consistently coming close to the human expert, it can then be refined. Bad conclusions can be traced back and debugged, and processes that are creating equivalent or improved conclusions can be encouraged. 

Knowledge engineering is already integrated in decision support software. Specialized knowledge engineers are employed in diverse fields that are advancing human-like functions including the ability of machines to recognize a face or parse what a person says for meaning. As the complexity of the model grows, the knowledge engineers may not fully understand how conclusions are being reached. Eventually the field of knowledge engineering will go from creating systems that solve problems as well as a human to one that does it quantitatively better than humans. Coupling these knowledge engineering models with other abilities like natural language processing (NLP) and facial recognition, artificial intelligence could be the best server, financial adviser or travel agent that the world has ever seen.