4. Internal and External Conceptions of `context'

4.2. The external approach

Alternatively, we could try to abstract from the concrete manifestation of individual's constructs outwards to the regularities and features these constructs rely on in order for their modelling heuristics to be useful (or even possible). Thus we have talk of context-as-a-resource in NL [11], or the context-we-inhabit in AI [3].

The problem with this approach is that the number of possible outward features that might be useful can be large. In order to focus on the parts that might actually be useful for an intelligence that is attempting to exploit them one has to consider (maybe implicitly) the internal construct of a context anyway. There may be some good grounds for identifying some relevant regularities on a priori grounds (for example, temporal context) but even in these cases it is hard to see how the properties of such contexts could be deduced for actual agents in real examples, without some validation that the presumed a priori grounds were actually used by the agent.

Thus in each case the pragmatics of learning, transferring and applying knowledge creeps in. The only escape from the relevance of these pragmatic roots of context is if one is not considering an actual or applicable rationality and reasoning but only some artificial rationality for use on problems in restricted domains (e.g. a heavily idealised or purely normative rationality).

The Pragmatic Roots of Context - Bruce Edmonds - 31 MAR 99

[Next] [Previous] [Up] [Top] [Contents]

Generated with CERN WebMaker