Designed to Be Believed: AI Personas as Rhetorical Interfaces
This paper argues that the personas of conversational AI systems are not accidental byproducts of the technology. They are designed rhetorical interfaces that hide computation behind a familiar social surface and invite users to extend the kind of trust they'd give a person.
Drawing on Byron Reeves and Clifford Nass's research on the "media equation" and Lucy Suchman's work on human-machine interaction, I read the public self-presentation of two commercial AI assistants as a constructed ethos. I argue that critical AI literacy has to start by treating the persona as a text that can be read, with an author and an intended audience.
Research process and revision
My first draft framed AI personas as deception, a friendly mask over a machine. My instructor pushed back on that binary. It assumed there was some honest, persona-free interface underneath, and there isn't. That objection reorganized the project. I spent the next month reading in human-computer interaction instead of media criticism, and the revised paper treats the persona as a designed experience with real effects rather than as a lie. The paper's conclusion about critical awareness grew directly out of that shift.
I also tested the argument in another medium. The companion video essay is delivered by an AI-generated presenter built from a synthetic face and a synthetic voice. Producing it made the paper's claims concrete for me. Every decision about the presenter's tone and expression was a design choice aimed at the viewer's trust.
Why this is my primary writing sample
This is the project I most want to keep working on. It shows how I enter a scholarly conversation, and it also shows how I change my position when a better objection comes along. I presented an early version at an undergraduate research symposium in Spring 2026.