Knowledge based researchers have long battled with the issue
of brittleness, which has either directly or indirectly led to the
majority of methodological developments in symbolic based reasoning.
Yet it is an issue that still remains in varying degrees in
most systems today. In knowledge based systems (KBS) brittleness
occurs when a system is asked to inference a situation that is
beyond the knowledge captured in the knowledge base. This brittleness
is particularly problematic in KBSs because the system does
not display an error or crash when it occurs. To the user there is
generally no discernable problem as the system still produces a response.
Therefore, if the user is not sufficiently knowledgeable to
be able to notice the response is flawed then they may treat it as
correct. For instance, a nursing assistant may not challenge a
response and thus not seek an expert’s opinion, resulting in an
incorrect treatment of a patient.