← ProCompNav project page

Interrupt Often vs. Ask Once

Someone is working at their desk while a home robot searches the house for their cabinet. Same robot, same three questions. Only the timing changes. When should it ask?

0:00elapsed (sped up)
  1. 1 · Robot explores
  2. 2 · Questions land
  3. 3 · What people chose

STUDY 2 · QUESTION DELIVERY · 20 PARTICIPANTS

Asked to choose, most people picked asking once.

Preference

14 of 20 preferred asking once
ask onceask often

Preferred asking once: 14 of 20 participants (asking often 6 of 20)

Median response time

Median response time: 5.28 s asking once vs 6.27 s asking often

Why: each interruption pulls attention away from the task; batching questions means fewer interruptions (Bailey & Konstan 2006; Fitz et al. 2019).

So ProCompNav explores and collects candidates first, then asks its few questions back-to-back.

Focus curves and meters in the timelines are illustrative, not measured data. Question times follow the Study 2 setup.