CONCEPT
Relational Tacit Knowledge
The species of tacit knowledge that is tacit for contingent reasons — it could in principle be articulated but happens not to have been — and the species that AI systems handle remarkably well by extracting what practitioners know from their textual residue at scale.
Relational tacit knowledge is the first of Collins's three species. It is knowledge that individuals possess but have not made explicit — the recipe grandmother never wrote down, the machining tolerances a factory worker knows from decades of practice but has never documented. The knowledge exists in explicit-capable form inside the practitioner's mind; the barrier to articulation is practical, not principled. Given sufficient motivation and a patient interviewer, it could be extracted and formalized. This is the species that large language models absorb most naturally — training on corpora effectively performs the extraction a patient interviewer would perform, at a scale no human effort could match.
In The You On AI Field Guide
Collins's framework treats the relational species as a genuine achievement ground for AI. When a model is trained on machining forums, cooking blogs, medical case reports, and professional discussion threads, it ingests vast quantities