"My name is... my name is...": A Linguistic Map for Voice Agents — Midam Kim, ServiceNow
Midam Kim, an ML engineer at ServiceNow, demonstrates how a voice bot misheard her name by adding an N, mispronounced it again despite her correction, and later cut her off while she read an account number.
She argues that voice AI failures are not random bugs but follow a structured pattern. Kim presents a grid model with two channels (listening and speaking) and four levels (sounds, words, interaction, mental model). Each type of error—recognition, pronunciation, turn-taking, intent tracking—lands in a cell. Her key insight is that in voice conversation, sounds disappear as they're spoken, but the user's mental model accumulates—and that's what you're actually designing for.
Vibekollen prepared this summary with AI from the original publication. The content belongs to AI Engineer.
