Epistemic Engineering: Towards a Discipline of Knowledge Representation for Artificial Reasoners in Engineering Organisations

Engineering organisations exercise consequential judgement through architectural doctrine, examples, corrections, review, apprenticeship and cultural practice, only part of which becomes durable in ordinary documentation. Artificial reasoners do not ordinarily participate in the sustained human socialisation through which these fragments acquire local meaning. This essay proposes epistemic engineering as the deliberate elicitation, representation, compression, validation and maintenance of organisational judgement for such collaborators. It introduces agent didactics as the arrangement through which that judgement becomes reconstructable during work and stems as compact, deliberately maintained epistemic representations. Repository exploration and synthetic studies show that withholding local judgement can limit model decisions, that correct representations can restore performance and that a representation of incompatible local practice can cause predictable harm. The tested stem forms did not outperform proposition-matched context; the proposal is therefore disciplinary and representational rather than a claim for a privileged species of model input.