What this means is that the distinction between the symbol level and the knowledge
level established by Newell (1981, 1993) has to be reflected in both the design process,
and the tools used to represent the model. The dynamic entities that make up the
question and answering entities cannot be bound by the symbol level of the knowledge
system, but will be invoked through either ad hoc or established knowledge
representation mechanisms. We have already explored this emergence mechanism in a
previous paper describing just-below-the-surface systems (Pigott et al., 2004), which
belong in the symbol level, yet are designed to permit the emergence of epiphenomenal
views of the knowledge, typically through frames, databases or spreadsheets. The
same stored values and rules can give rise to different knowledge epiphenomena, and
typically such an epiphenomenon will call on more than one set of values within the
symbol level. Just as the abstraction of the database conceived of by Codd (1969) was a
series of relations epiphenomenal to but inherent in the values, so we must look for a
way of describing knowledge relations, dwelling in the knowledge level while being
epiphenomenal to but inherent in the symbol level[1].