Work Coursework
Speech or tone: takeover alerts in Level 3 automated vehicles
A 2 × 2 mixed-design experiment on whether spoken or tonal takeover requests change how fast drivers take control from an automated vehicle.
When a Level 3 car hands control back, should it say "takeover requested" or beep? Eight participants watched driving video recorded in Atlanta, played Tetris as a distraction, and pressed a key when alerted. Neither alert type nor how much warning they had significantly changed takeover time or user-experience scores.
Why it matters for GNC. Handing control between automation and a human is a supervisory-control problem; the study's main lesson was about making test scenarios critical enough to measure it.
Design
Alert type (speech or a two-tone beep pattern taken from the literature) varied between participants; situation (proactive, 6–7 s warning, or reactive, 2–3 s) varied within each participant. A Python key logger timed the response to each alert, and participants completed the short User Experience Questionnaire (UEQ-S).
Result
Two-way ANOVAs found no significant main effects or interaction for takeover time or UX score. Spoken alerts had lower mean takeover times in both situations (2.90 s and 2.53 s against 3.74 s and 3.26 s), but not significantly with this sample.
Eight participants, only one of whom had driven a Level 3 vehicle. Participants found the clips too undramatic to tell proactive from reactive situations, so the manipulation probably didn’t land. The paper recommends more critical scenarios and a larger, more varied sample.
Documents
- Final paperTian, Muralikrishna, Hovey and Suh, April 2024. 9 pages.