In contrast with the ad hoc approach, authors for Volcano took a very structured approach to their rule building. Using the categories outlined in the previous chapter (conflicting, required, neutral, and irrelevant), one member of the team categorized all of the available points for each claim, just as he would have for using the INDIE 1.0 critiquer.
He then wrote a remediation for each class of mistake (e.g., ``You seem to have irrelevant information here.''). He built rules to fire each of those mistakes for each claim:
He then put predictable mistakes, such as thinking visual information about the volcano is relevant for determining future behavior, and inserted those rules in the appropriate categories:
Other mistakes (such as failing to run important tests) were added to the appropriate places in the hierarchy. Since rules are fired in order, students are encouraged to deal with specific problems with their evidence one problem at a time in order of importance, making sure that more specific critiques are given before general ones, yet also guaranteeing that authors won't miss cases.
This structure loosely resembles the design of the original, knowledge-based critiquer from INDIE 1.0. Unfortunately, Volcano never reached a completely playable state where the critiquer could be tested, so the value of this approach was not verified. However, it does show that, with forethought, authors can produce complete and robust rule-based critiquers.