The machine between us: How AI shapes the way we speak

Research increasingly uses voice and audio features to understand human-technology relationships. But findings regarding AI and spoken communication often appear somewhat contradictory. People can be blunt or even aggressive toward chatbots, speak more politely to voice assistants, and self-censor themselves when AI listens in the background during a call. Why? To address this heterogeneity, we propose a Conversational Role Framework that distinguishes whether the AI is (1) a conversation Partner that is directly addressed (e.g., a chatbot), (2) a Filter that transforms speech signals before another person receives it (e.g., a real-time translator that re-voices a speaker), or (3) an Ambient Listener that silently records, transcribes, and summarizes a human exchange (e.g., an AI note-taker summarizing a doctor-patient conversation). This framework helps explain seemingly conflicting findings and sheds light on why the same vocal cue can capture different meanings depending on AI's role in a given interaction.