A knowledge graph makes relations between entities visible in a way prose rarely manages — this cites that, this influenced that, this and that co-occur — and the resulting network diagram carries an implicit authority, since it looks discovered rather than constructed. What the graph conceals, precisely because of this visual authority, is the interpretive decision embedded in every edge: someone decided that this relation was worth encoding and that one was not, and the graph presents the surviving decisions as though they were simply facts about the world rather than one particular reading of it. Semantic graphs reveal relation while concealing the interpretation that selected which relations to show. InterpretationTerritory returns from nodal logic here in its clearest form: every edge in a graph stakes out interpretive territory, and the graph's apparent objectivity is really the territory's boundaries rendered invisible by good design.
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.
Helmond, A. (2015) 'The Platformization of the Web: Making Web Data Platform Ready', Social Media + Society, 1(2), pp. 1-11.
Galloway, A.R. (2004) Protocol: How Control Exists after Decentralization. Cambridge, MA: MIT Press.