The Technology Transfer Problem
The technology transfer problem is one of Collins's foundational empirical findings, established through studies of how specific scientific and engineering capabilities have (and have not) moved between laboratories and nations. The paradigmatic case is his study of TEA lasers: American laboratories attempting to replicate British techniques in the 1970s had access to every published paper and specification, but could not make the lasers work. Success came only when British physicists traveled to American labs and worked alongside American researchers, transmitting tacit knowledge through direct social interaction. The published literature was necessary but not sufficient; the social transmission was indispensable.
In the AI Story
The finding has become foundational in science studies and directly relevant to AI evaluation. If technical practices cannot be transmitted through documentation alone to humans with strong prior training, the question of whether they can be transmitted through training data to systems without such prior engagement becomes acute. Collins's framework suggests the answer is structural: the knowledge that resists documentation is collective tacit knowledge maintained in social practice, and no amount of textual ingestion can substitute for the social participation through which it is maintained.
The corollary is sobering for the AI transition: if the human communities that sustain collective tacit knowledge dissolve or are bypassed faster than new social structures can form around AI-directed practice, the knowledge will not persist. Technology transfer requires receiving communities that can absorb the tacit dimension through social contact. The AI transition risks producing a situation in which the receiving 'community' is a statistical model trained on textual residue — a recipient that can absorb explicit knowledge perfectly and tacit knowledge not at all.
Origin
Collins established the framework through studies published in the 1970s and 1980s, most systematically in Changing Order (1985). The TEA laser study remains the canonical example, but Collins extended the framework to numerous other cases of knowledge transmission across institutional boundaries.
Key Ideas
Documentation is insufficient. Published specifications cannot fully transmit technical practice; receiving communities must acquire tacit knowledge through social contact.
Canonical TEA laser case. American labs could not replicate British laser techniques from publications alone; success required British physicists visiting.
Implications for AI. If tacit knowledge resists transmission to trained humans through text, it resists transmission to statistical models trained on text.
Irreversibility. If source communities dissolve before tacit knowledge transfers, the knowledge is lost rather than delayed.
Appears in the Orange Pill Cycle
Further reading
- Harry Collins, Changing Order (SAGE, 1985)
- Harry Collins, 'The TEA Set: Tacit Knowledge and Scientific Networks' (Science Studies, 1974)
- Harry Collins, Tacit and Explicit Knowledge (University of Chicago Press, 2010)