Conclusion and Justification
There is a widely accepted observation that while experts cannot always explain in advance of a situation what will happen, they can always defend their conclusions when presented with a given case. This observation lies at the very heart of Erudine’s Behaviour Manager.
Instead of drawing up a list of rules in advance to describe the behaviour of a system, and then building the system from this rigid list, the Erudine approach lets experts construct system behaviour through situations, teaching the software the outcome of a given case (the conclusion) and then explaining to the software why this outcome should occur (the justification).
Erudine's Behaviour Manager tool learns system behaviour iteratively. To start with, a domain expert presents a case and informs the software of its conclusion.
The software then asks the expert to justify the conclusion.
If the justification is satisfactory, the software sets the appropriate system behaviour. As the expert presents more cases, the Behaviour Manager builds up a behaviour base.
All new behaviour is validated against the behaviour base to check for clashes, but the software can also use the base to offer suggestions that the expert can accept or reject. Any rejections require the expert to justify their reasoning, further refining the behaviour of the system.
System behaviour is created as the expert or user goes about their usual tasks, continuing to work within their area of knowledge and responding to the queries made by the Behaviour Manager.