Whose Interpretation Counts? Reading Generative AI as an Interpretive Technology Across UK and Indian Households

The workshop call frames generative AI as an interpretive technology: a system that produces cultural artefacts and performs acts of interpretation in doing so. We take this seriously and ask the question it implies but rarely poses directly: whose interpretive conventions does a deployed system enact? We re-read two qualitative interview corpora of smart-home AI users, one collected in the United Kingdom (n=29 households) and one in India (n=33 households), against a shared schema. We find systematic differences along three axes of interpretive expectation: how the system should signal its understanding, where interpretive authority is presumed to sit, and what makes an exchange feel complete. These differences are not surface preferences. They reflect distinct hermeneutic stances that current output-conformity benchmarks render invisible. We argue that a positive vision of cultural AI evaluates systems as participants in interpretation rather than producers of outputs, and we sketch four design moves and a concrete evaluation schema that follow. The contribution is a methodological reorientation, from asking what an AI says to asking how an AI reads, and from treating culture as a property of the model to treating it as a property of the deployment.