Knowledge modelling

Our work is about knowledge: specifically how people model their own knowledge. Our knowledge modelling cycle has four steps.

•Elicitation: Getting knowledge from human beings

•Representation: Choosing the best format to show that knowledge

•Testing: Testing the representation for correctness, completeness and accuracy

•Transfer: Passing on knowledge via media, teaching and training

Each step interacts with the others. For example, some representations make testing easier because they make significant absences more visible.

Knowledge Modelling draws together theories, concepts and methods from a wide range of disparate fields. This is how it happened…

In the mid-1980s, Gordon and his colleague Peter McGeorge were researching expert systems – software that’s meant to mimic how experts solve complex problems such as the diagnosis of disease.  But expert systems weren’t working as expected.

It had initially been assumed that experts, of all people, would think rationally and logically, but it turned out they didn’t.  Also, experts’ thinking showed particular patterns of error.  Gordon and Peter set out to discover how experts really did think.

They soon found that the obvious fields – computer science and psychology – didn’t offer the methods they needed, so they searched domains such as mathematics, anthropology, sociology and market research, and collated a range of methods for eliciting information from experts.

It became clear that there were different types of knowledge, and that different elicitation methods mapped on to them.  This prompted Gordon and his colleague Neil Maiden to develop the ACRE (ACquisition of REquirements) framework. There’s more about the ACRE framework here. Gordon then went on with Jo Hyde to develop the Verifier method for identifying expert errors.

Gordon applied Verifier to a problem that had eluded experts for decades – the Voynich Manuscript. Sue Gerrard applied it to autism – another long-standing theoretical challenge.  Both pieces of research resulted in papers published in leading peer-reviewed journals.

There’s a lot of knowledge about knowledge, but it’s scattered across many different domains.  Hyde & Rugg has brought it all together for the first time in the Knowledge Modelling Cycle that integrates the elicitation, representation, testing and transfer of knowledge.

The mechanisms underpinning knowledge are complex, but the techniques we use to elicit, represent, test and transfer it are simple, inexpensive, and easy to learn.   We’ve presented summaries of some of them on our website.  We hope you find them useful.

A sample of the Knowledge Modelling Handbook, by Gordon Rugg & Sue Gerrard