Exploring the Role of Conversational Cues in Guided Task Support with Virtual Assistants
summary
A Wizard of Oz study has participants cook a spice rub for corn while interacting with a simulated voice agent; 10 participants take about 7 minutes on average and produce roughly 19 turns. Transcripts classify utterances as explicit questions or context-dependent implicit cues—about 25% of exchanges—including “okay”/“sounds good,” paraphrases, grounding tokens, confirmations and subvocal repetition. Results recommend capturing those cues (e.g., longer mic windows) and redesigning guided-task dialogue models.
AI-written summary · about
chapters (9)
- 0:00Introduction to research
- 1:52Limitations of current assistants
- 3:17Study goals and methodology
- 6:28Participant demographics
- 7:11Explicit queries vs implicit cues
- 9:26Types of explicit queries
- 9:47Types of implicit cues
- 12:20Design implications and future work
- 13:09Q&A session
related talks
33mStrategies for Conversational Product Experiences
55mDesigning for speech
59mDesigning for speech
10mApp-Based Task Shortcuts for Virtual Assistants
54mConversational UX—Human Speech Patterns Meet UX
30mSystem Prompt Design for Voice Agents
01h 04mBuilding Expertise Into Generated Conversations
13mStepWrite: Adaptive Planning for Speech-Driven Text Generation