Page 144 -

4 Dynamic Workflow 133
If the conclusion returned was that of a satisfied terminal rule, then the new rule
is added as a local exception to the exception “chain” via a new true branch from
the terminal node
If the conclusion returned was that of a nonterminal, ancestor node (i.e., the con-
dition of the terminal rule was not satisfied), then the new rule is added via a new
false branch from the unsatisfied terminal node
In essence, each added exception rule is a refinement of its parent rule. This
method of defining new rules allows the construction and maintenance of the rule
set by “subdomain” experts (i.e., those who understand and carry out the work they
are responsible for) without regard to any engineering or programming assistance
or skill.
Importantly, each rule node also incorporates a set of case descriptors, called the
“cornerstone case,” which describe the actual case context that was the catalyst for
the creation of the rule. When a new rule is added to the rule set, its conditional
predicate is determined by comparing the descriptors of the current case to those
of the cornerstone case and identifying a subset of differences. Not all differences
will be relevant – it is only necessary to determine the factor or factors that make
it necessary to handle the current case in a different fashion to the cornerstone
case to define a new rule. The identified differences are expressed as attribute-value
pairs, using the usual conditional operators. The current case descriptors become
the cornerstone case for the newly formulated rule; its condition is formed by the
identified attribute-value pairs and represents the context of the case instance that
caused the addition of the rule.
Rather than impose the need for a closed knowledge base that must be completely
constructed a priori, this method allows for the identification of that part of the
universe of discourse that differentiates a particular case as the need arises. Indeed,
the only context of interest is that needed for differentiation, so that rule sets evolve
dynamically, from general to specific, through experience gained as they are applied.
Ripple-Down Rules are well suited to the worklet selection processes, since they:
Provide a method for capturing relevant, localized contextual data
Provide a hierarchical structuring of contextual rules
Do not require the top-down construction of a global knowledge base of the
particular domain prior to implementation
Explicitly provide for the definition of exceptions at a local level
Do not require expert knowledge engineers for its maintenance
Allow a rule set to evolve and grow, thus providing support for a dynamic
learning system
Each worklet is a representation of a particular situated action that relies on the
relevant context of each case instance, derived from case data and other (archival)
sources, to determine whether it is invoked to fulfill a task in preference to another
worklet within the repertoire. When a new rule is added, a worker describes the
2
contextual conditions as a natural part of the work they perform .This level

2
In practice, the worker’s contextual description would be passed to an administrator, who would
add the new rule.