Timing the Message: Language-Based Notifications for Time-Critical Assistive Settings
Ya-Chuan Hsu, Jonathan DeCastro, Andrew Silva, Guy Rosman · 2025
Read Timing the Message: Language-Based Notifications for Time-Critical Assistive Settings by Ya-Chuan Hsu, Jonathan DeCastro, Andrew Silva, Guy Rosman (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-07438.txt.
Description
Ya-Chuan Hsu and 3 co-authors study when a spoken assistant should talk in time-critical help settings in a 2025 paper. Alerts and vibrations grab attention fast but leave the person to interpret the scene alone, which can delay action. Spoken instructions carry context, yet speaking takes time, listening takes time, and acting takes still more, so a helpful message can arrive too late.
The team frames the timing-versus-detail trade-off as a sequential decision problem using an augmented-state Markov Decision Process, a model that tracks an enlarged state. Their framework pairs reinforcement learning with an automatically generated taxonomy dataset, a labeled catalogue of message types. In tests with synthetic humans, success rates rose by over 40 percent against methods that ignore timing delays. The lesson applies to assistive driving and other settings where seconds decide outcomes.
Catalog record
| CREATOR | Ya-Chuan Hsu, Jonathan DeCastro, Andrew Silva, Guy Rosman |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-07438