Translation as Cultural Infrastructure in an Age of Machine Mediation
The familiar model of scholarly communication imagines research first and translation later. A text is completed in a dominant language and then, if resources permit, converted for other readers. Performance exposes this misunderstanding. Choices made in libretti, surtitles, and programme notes shape how performances travel, the histories to which they appear to belong, and the audiences who can answer them. Translation is not a transparent channel between already stable meanings. It is an interpretive practice that participates in making performance meaning.
Treating translation as infrastructure changes institutional priorities. Infrastructure is planned, maintained, and governed; it is not requested as a favor at the end of a project. An opera company preparing multilingual surtitles, a theatre festival presenting international work, or a music archive serving global researchers should decide early which languages matter, who will translate, how terminology will be reviewed, and how disagreements will be documented. The budget should include this labor. Otherwise multilingualism becomes ornamental: a programme page may appear in several languages while the performance context remains accessible in only one.
Contemporary machine translation can support navigation, rehearsal comprehension, and communication across languages. Its speed is valuable for performance materials that would otherwise remain inaccessible. The appropriate response is neither uncritical adoption nor blanket rejection. The issue is fitness for purpose.
A rough translation that helps a researcher decide whether to pursue a source is different from an authoritative translation of libretti, songs, oral histories, or performance heritage. In the latter contexts, a plausible sentence can still be seriously wrong. A system may flatten dialect, misread irony, distort musical prosody, or replace an unfamiliar performance concept with a familiar but misleading equivalent. Because fluent output can conceal uncertainty, users may be less likely to recognize an error than they would be when reading visibly imperfect language.
The longstanding idea of the translator's “invisibility” is newly relevant here. Automated interfaces can make mediation disappear precisely when it should be documented. Performing-arts institutions should therefore label machine-generated translation, identify whether a human reviewed it, retain the source libretto, and provide a route for correction. High-stakes performance materials should not be published under the vague claim that they were “AI translated” without naming the responsible human reviewer and explaining the procedure.
Translation is frequently omitted from performing-arts credit systems even when it requires substantial research. A translator may trace allusions, reconcile historical terminology, consult community speakers, and negotiate rhythm, register, or singability in performance. This work can alter interpretation as decisively as dramaturgical editing. Appiah's account of “thick translation” is useful because it emphasizes the contextual work required to place readers in a position to understand difference rather than erase it.
Performing-arts institutions should name translators prominently, provide contributor statements, and permit translator notes when choices materially affect interpretation. Where several people review performance terminology, that collaborative work should also be visible. Credit is not merely symbolic. It creates accountability by allowing readers to know who made interpretive decisions and on what basis.
A credible policy for translation should include at least four elements:
- Purpose: distinguish discovery aids, rehearsal translations, programme communication, and authoritative performance translations.
- Responsibility: identify the translator, reviewer, relevant performance expertise, and process for resolving disputed terms.
- Disclosure: state whether automated tools were used, what they produced, and how humans checked the output.
- Reciprocity: support translation into and among languages other than English rather than treating performance knowledge as destined for a single center.
Reciprocity matters because translation flows are shaped by unequal performance economies. Repertoire produced in widely resourced languages is more likely to be translated, programmed, and taught internationally. Performance in less globally dominant languages is often expected to become legible on terms established elsewhere. A multilingual agenda should therefore support bibliographic discovery, surtitles, teaching materials, and performance routes that allow knowledge to circulate in more than one direction.
Machine mediation makes translation more available, but availability should not be confused with adequacy. The question is not whether a system can produce readable sentences. It is whether performing-arts institutions can create responsible conditions for performances to travel without concealing uncertainty, labor, or difference. Recognizing translation as cultural infrastructure offers a practical answer: plan it early, fund it properly, credit it visibly, review it proportionately, and preserve the possibility of correction.