Programmed Politeness: Gendered Speech Acts in the Response Architecture of Siri and Alexa

Abstract

The current qualitative study is an exploration of Apple's Siri (iOS 16-17) and Amazon's Alexa (3rd and 4th generation) response patterns through Robin Lakoff's gendered language feature’s structure, J. L. Austin's and John Searle's speech act theory and Judith Butler's gender performativity theory. Analyzed with systematic discourse analysis across five interaction categories: task failure, factual information retrieval harassment identity queries, and opinion requests - 150 directly elicited responses per platform, it turned out that female-voiced AI assistants keep producing feminine-coded features: epistemic deference, apologetic face-saving, conflict avoidance, self-effacement, and opinion deferral. Those female-voiced, personified AI systems are the ones who Mostly show deferential features rather than these features being evenly spread across AI interface types, which is evidenced by cross-gender experimental data (Mahmood and Huang; Leisten and Rieser) and a corpus-level harassment study (Cercas Curry and Rieser). To be specific, the paper articulates the concept of ritualized deference markers pre-programmed expressive acts that perform social subordination absence of actual culpability and further maintains that the repetition at the institutional scale is a form of algorithmic gender citation by Butler's definition. Some recent statistics to back up limited cross-sectional spillover onto users' general gender attitudes (Steeds et al.) is recognized; at the same time, the use of longitudinal effects is pinpointed as the key still unanswered empirical question by this paper.