Interrupted tasks must keep their place
Aliases: placekeeping · task resumption · in-car progress persistence
What it is
When driving breaks an in-motion task, coming back must land in the same place, not at the start. Placekeeping is what makes interruption survivable: the selected item, the typed string, the step index are still there. Without it, an interleaved structure is an empty claim—every return is a reorientation, and those locating glances often outlast the work itself. This entry is not about whether a task should be sliced for short glances, and not about whether uninterruptible tasks should exist. It is only about whether state is still there after a break that has already happened.
Why it happens
Interruption clears the working-memory note of “how far I had got.” If the interface is still on the same choice, the same caret, the same unsubmitted field, one short glance can resume, because the display is holding the place the head no longer holds. If it has jumped home, dismissed the keyboard, or dropped a half-typed search, the person has to find the entry again and recall what was already entered—a fresh visual-manual-cognitive stack. When driving interruptions come densely, an unpreserved task is reset over and over, and the occupation is not a sum but an amplification. Session timeouts and “safe” draft-dropping on focus loss look like less distraction. They manufacture the next, longer recovery.
Studying it
Force a driving-priority event mid-flow, then allow a return, and compare a preserve implementation with a wipe. Resumption-time and wrong-step re-entry from interruption-cost research transfer directly to in-vehicle tasks.
Independent variables: interruption length, which step was active, whether the original locus is highlighted on return, whether the draft remains. Dependent variables: duration of the first glance after return, wrong-step re-entry, time from interruption to the next useful action, subjective “where was I.”
Lab interruptions are single and announced; road interruptions are repeated and not. Measuring one recovery undercounts the cost of drafts being dropped under repeated breaks. The moment a system timeout wipes the session is itself a variable and should be reported separately.
Where it stops holding
Steps that spend money, unlock doors, or change automation level may be unsafe to resume as a silent draft; they need an explicit re-confirm rather than continuing a previous authorization. Passenger progress on a secondary display is not the driver’s state; do not “preserve” by syncing a co-driver’s typing into an unfinished driver navigation. Whether progress survives vehicle power-off is a different product decision from surviving a traffic break while driving.
Applying it
- By default, in-motion input, selections, and wizard steps must not clear because the eyes went back to the road, a call came in, or a lead vehicle braked. Return on the same step, with the locus recognizable in one look.
- Timeouts should outlast a typical driving event, or the draft should be stored before timeout. Do not impersonate safety with a short timeout.
- Gate re-authorization (payment, automation level) behind a dedicated confirm; do not throw the whole wizard away.
- Verify by stopping halfway, inserting a lane change or a call, then looking back. If the keyboard is gone, the query is empty, or the root menu is showing, place was not kept.