Epistemic Infrastructure

Epistemic infrastructure denotes the largely invisible arrangements — categories, standards, records, review processes, training pipelines — through which claims come to count as knowledge. It is what makes an assertion assessable at all.

AI systems both depend on and transform epistemic infrastructure: they inherit its categories, exploit its records, and reshape the practices that produced them. This creates a reflexive loop classical accounts of evidence do not anticipate.

Treating epistemic infrastructure as a first-class philosophical object reframes debates about bias, generalisation, and trust — moving them from properties of models to properties of the arrangements in which models are trained, deployed, and audited.