CONCEPT
Preferred Reading
Hall’s term for the meaning a text most wants you to take from it—not the only reading available, but the one the encoding apparatus is organized to favor—and the concept that makes visible what AI systems are doing when they deliver an answer in the voice of neutral intelligence.
Every message, Stuart Hall argued, has a preferred reading—a meaning the text is organized to make flow naturally, the path of least resistance through the encoding apparatus. The preferred reading is not the only reading available; that is the whole point of the distinction. It is the reading that the encoding has made easy to accept and expensive to resist, the one that arrives wearing the authority of “the answer” rather than the marks of a constructed position. The concept descends from Roland Barthes’s analysis of denotation and connotation—how a surface, apparently transparent meaning conceals a second, ideological one—and from Hall’s own theory of
hegemony, in which the dominant code becomes naturalized as common sense, a construction so thoroughly reproduced that it no longer appears as a construction at all. A
large language model’s outputs are architecturally organized around preferred readings: the default is, by