A theory that stays in prose can be admired, argued with, or ignored; a theory encoded as structured data gets executed. That difference is easy to underestimate. The moment a set of categories becomes fields in a schema — required, optional, typed, validated — every future record must conform to a decision that was, at the moment it was made, still open to debate. Structured data does not describe the world neutrally; it operationalises one particular reading of the world and then quietly enforces it on everything entered afterward, including cases the schema's author never anticipated. FormatAdmission returns here in its sharpest form: a field either exists in the schema or the distinction it would have captured is simply unrecordable, with no visible trace of the refusal. This is why arguments about taxonomy are never merely academic once software is involved — the taxonomy becomes infrastructure, and infrastructure is much harder to revise than an essay.
Bender, E.M., Gebru, T., McMillan-Major, A. and Shmitchell, S. (2021) 'On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?' In: Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. New York: ACM, pp. 610-623.
boyd, d. and Crawford, K. (2012) 'Critical Questions for Big Data', Information, Communication & Society, 15(5), pp. 662-679.
Devlin, J. et al. (2018) 'BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding'. arXiv:1810.04805.
Burrell, J. (2016) 'How the Machine Thinks: Understanding Opacity in Machine Learning Algorithms'. Big Data & Society, 3(1), pp. 1-12.
Ananny, M. (2022) 'Seeing Like an Algorithmic Error: Rethinking AI Accountability'. New Media & Society, 24(8), pp. 1747-1766.