Managing Uncertainty in Time Expressions for Virtual Assistants
summary
Study measures how people use uncertain time words and what assistants actually do: a mixed methods pipeline uses 338 survey respondents, interviews with 14 people, and analysis of ~380k emails parsed and extended with SUTime to capture messy phrases. Findings: imprecise temporal expressions are common and long tailed (≈80% occur ≤3 times, 60% unique), people prefer vagueness for low‑priority or dependency tasks, and a built prototype assistant shows opportunities to respect imprecision, delay action, and gather passive signals rather than force precision.
AI-written summary · about
chapters (11)
- 0:00Introduction to talk
- 1:09Problems with time expressions
- 3:38Research questions overview
- 5:00Survey on imprecise time use
- 8:51Email corpus analysis
- 11:41Prototype assistant study
- 15:16Design implications for assistants
- 17:31Q&A on handling imprecision
- 19:19Q&A on user reliance on assistants
- 20:25Q&A on user expectations
- 21:31Q&A on system capabilities
related talks
17mExploring the Role of Conversational Cues in Guided Task Support with Virtual Assistants
06mInvestigating Users’ Preferences and Expectations for Always-Listening Voice Assistants
21mWhy Agent Hype can fall short of reality
29mBuilding an A.I. Assistant that Speaks Human
50mDesigning Experiences For Voice Interactions
25mEngineering voice agents: Latency, quality, and scale
36mHow We Built a Voice Agent with Temporal
49mSolving the Next Billion-People Privacy Problem