A concept becomes addressable when a field can point to it without pretending to exhaust it. This apparently modest operation contains the full violence and utility of lexical infrastructure. Williams’s historical semantics shows why the problem begins before metadata: a word such as culture already contains incompatible temporal deposits, disciplinary migrations and social conflicts, so any stable label is a cut across semantic thickness rather than a transparent pointer to an essence. His method refuses the fantasy that terminological precision can simply remove semantic history; the difficulty of the word is itself evidence. Bal intensifies that instability by treating concepts as travellers whose analytical force emerges through movement between disciplinary situations, and Cassin radicalises it through translation, where the very terms most worth preserving are often those that resist clean equivalence. From this side, conceptual knowledge appears mobile, stratified and permanently vulnerable to simplification.
Yet fields cannot operate through semantic plenitude alone. Reading, teaching, citation, archival deposit and computation all require points of entry. Bowker and Star show that such points are never innocent because classification is infrastructural: the categories that make coordination possible also distribute visibility, normality and exclusion. The critical problem is therefore not whether to classify but how to make the classificatory act inspectable. An addressable concept must be stable enough to be retrieved and unstable enough to remain historically reopenable. This is why a one-to-one concept–author pairing can be useful only when understood as an access device rather than an origin story. The pairing lowers the entry cost into a field; its references, alternate genealogies and relational links must then restore the complexity that the anchor temporarily suppresses. Canonical compression is defensible only when it contains its own decompression route. What matters is not simply that compression loses information; every usable interface does. The methodological question is whether the loss is declared, reversible and accompanied by enough provenance to reconstruct what was excluded from the first view. An alphabetical lexicon already performs several invisible operations: it decides what counts as a concept rather than an author, theme, object or problem; it assigns a preferred label; it establishes a threshold of inclusion; and it positions every omission outside the visible field. These operations become especially consequential when the lexicon is treated as an infrastructure rather than a book, because the first editorial cut can propagate into search, indexing, citation and later machine processing. SKOS clarifies the technical structure of this compromise by separating concepts from their labels and relations, making a concept scheme publishable and linkable across the Web, but that formal separation also exposes the political question that Williams and Bowker and Star keep alive: who decides that two expressions are alternate labels of one concept rather than signs of historically different objects? Cassin’s untranslatables show that even apparently straightforward cross-language equivalence may be an unfinished philosophical event. Translation produces relations without guaranteeing sameness. Bal’s travelling concepts make the same point across disciplines: mobility is productive precisely because movement changes the analytical object. Cheng and Xia demonstrate an analogous difficulty inside information science, where taxonomy alignment must decide whether to merge, map or preserve original structures, and the status of the originals is itself part of the result. Their distinction matters far beyond taxonomies. A field that imports concepts from philosophy, sociology, architecture, anthropology, feminism, media theory and systems thinking must decide whether cross-disciplinary relations are equivalences, approximations, genealogical links, later appropriations, productive misreadings or merely thematic proximities. If all such relations are flattened into “related”, the network gains connectivity while losing epistemic resolution. Addressability therefore cannot mean frictionless interoperability. It means constructing a route to a concept while preserving evidence of the route’s editorial decisions. The concept must have an address, but the address must not be mistaken for the territory; the relation must be traversable, but traversal must not erase the difference that made the relation worth recording.
Once conceptual addressability becomes computational, the lexicon changes ontological status. Hogan and colleagues show what graph representation makes possible: relations become paths, clusters and queryable structures; schemas can remain partially open; heterogeneous sources can be integrated without reducing every relation to a table. This is precisely where a critical lexicon can exceed the book without abandoning it. An alphabetical dictionary offers a stable surface for human entry, but a relational layer can expose densities, bridges, isolated nodes, repeated authors, disciplinary concentrations and conceptual neighbourhoods that linear reading leaves latent. The gain is not automation of interpretation but a second mode of legibility. Consolati makes the reflexive stakes unavoidable: when concepts circulate through algorithmically mediated environments, computation is no longer only a neutral container for conceptual history. Ranking, retrieval, recommendation and machine-readable relation can alter which concepts become visible together, which inheritances become structurally proximate and which absences cease to be encountered at all. A Lexicum designed for both human reading and machine traversal must therefore treat its metadata as part of the argument. Node identifiers, preferred labels, alternative labels, principal figures, contested attributions, original-language terms, socioplastic reframings, references, cross-references, tags and version histories are not administrative residues appended after theory; they are the architecture through which theory acquires persistence and public form. The most useful design would preserve three levels without collapsing them: a human-readable entry capable of contradiction and historical nuance; a machine-readable concept record capable of stable retrieval and typed relations; and a provenance layer recording who established, altered or disputed each relation and when. Cross-references should be typed enough to distinguish genealogy from analogy, translation from adjacency, influence from later reuse and conceptual opposition from mere thematic similarity. A relation between two nodes should therefore be read as a claim with a provenance, not as a graphic line whose existence is self-justifying. This is where the infrastructure becomes genuinely critical. The network should be able to record that two concepts are linked without requiring agreement on why, and it should be able to preserve competing mappings when different disciplinary histories produce different relations. Versioning is equally important. It should allow the 100-node field to change without erasing its former states, because a revised author anchor, a new translation note, a split concept, a removed edge or a renamed tag is itself evidence about how the field is being reconstructed. The archive then records not only conceptual content but editorial history. This converts revision from maintenance into longitudinal evidence: later versions can be compared to show which traditions become more central, which authors lose canonical weight, which concepts acquire new neighbours and which exclusions become impossible to ignore. It also changes the meaning of the number 100. One hundred is not a census of theory but a declared computational and editorial boundary: finite enough to audit, dense enough to generate topology, open enough to acquire later satellites, counter-lexica or domain-specific expansions. A second hundred would not merely enlarge the first; it would alter the topology of the field, change centralities, create new bridges and potentially destabilise earlier groupings. The system should therefore resist the temptation to present growth as automatic improvement. Expansion is another editorial act. The same applies to machine access. A machine-readable Lexicum can support retrieval, comparison and graph analysis, but it should not silently infer intellectual ownership from the principal-figure field or transform cross-reference frequency into philosophical importance. Human and machine readings do different epistemic work. The human reader can detect contradiction, irony, translation difficulty and historical asymmetry; the machine can preserve identity across versions, expose repeated structures and traverse relations at a scale unavailable to memory. Their coexistence is strongest when neither is treated as the reduced form of the other. The resulting object sits between dictionary, archive, graph and publication without fully becoming any one of them. Its public legitimacy depends on maintaining that plurality rather than allowing the machine-readable layer to harden provisional decisions into invisible ontology. The Lexicum can therefore be cited as text, traversed as structure, compared across versions and criticised at the level of its cuts. Its most serious contribution is not the possession of a vocabulary but the design of conditions under which conceptual inheritance becomes reachable without becoming closed, computational without becoming self-authorising, and structured without suppressing the histories that structure cannot contain. Within Socioplastics, this tension condenses as DualAddress.
Anto Lloveras is a Madrid-based architect and urbanist working across epistemology, archive design, knowledge infrastructure and comparative fieldwork. Through LAPIEZA-LAB and Socioplastics, he develops serial publications, conceptual systems and spatial research protocols that treat theory, metadata and public form as mutually constitutive.
BIBLIOGRAPHY
Bal, M. (2002) Travelling Concepts in the Humanities: A Rough Guide. Toronto: University of Toronto Press.
Bowker, G.C. and Star, S.L. (1999) Sorting Things Out: Classification and Its Consequences. Cambridge, MA: MIT Press.
Cassin, B. (ed.) (2014) Dictionary of Untranslatables: A Philosophical Lexicon. Princeton, NJ: Princeton University Press.
Cheng, Y.-Y. and Xia, Y. (2023) ‘A Systematic Review of Methods for Aligning, Mapping, Merging Taxonomies in Information Sciences’, Journal of Documentation. https://doi.org/10.1108/JD-01-2023-0003.
Consolati, I. (2024) ‘Algorithms and Conceptual History’, Revista de Filosofía de la Universidad de Costa Rica, 62(166), pp. 167–176.
Hogan, A., Blomqvist, E., Cochez, M., d’Amato, C., de Melo, G., Gutierrez, C., Kirrane, S., Labra Gayo, J.E., Navigli, R., Neumaier, S., Ngonga Ngomo, A.-C., Polleres, A., Rashid, S.M., Rula, A., Schmelzeisen, L., Sequeda, J., Staab, S. and Zimmermann, A. (2021) ‘Knowledge Graphs’, ACM Computing Surveys, 54(4).
Isaac, A. and Summers, E. (eds.) (2009) SKOS Simple Knowledge Organization System Primer. W3C Working Group Note, 18 August 2009.
Williams, R. (1976) Keywords: A Vocabulary of Culture and Society. New York: Oxford University Press.