ACM Conversational User Interfaces (CUI)/2025
Conversational Scaffolding: Progressive Intent Discovery in Voice-First AI Interfaces
Designing adaptive audio cues and contextual interruption triggers to guide unstructured stream-of-consciousness dictation into structured schema outputs.
Unstructured voice dictation frequently contains false starts, mid-sentence corrections, and conceptual tangents. We present conversational scaffolding - an ambient interaction model that provides subtle audio and visual cues during continuous speech.
A longitudinal trial with 80 knowledge workers demonstrated that progressive intent scaffolding yielded 54% cleaner database schemas from raw speech notes without interrupting human train of thought.
KEY INSIGHT:
The highest goal of voice UX is not transcribing every word, but understanding the thinker behind them.
VENUE / JOURNALACM Conversational User Interfaces (CUI)
YEAR2025
DOI IDENTIFIER10.1145/3712049.3724108
RESEARCH TOPICS
Voice UXAudio ScaffoldingSpeech-to-IntentMultimodal Systems
