Self-initiated interruptions have lower resumption cost than externally imposed ones

What it is

Even for the same act of switching away from a primary task and back again, who initiates the switch clearly affects the cost of resuming: deciding on one's own, "I'll go do something else now," produces faster resumption and fewer errors than an externally imposed interruption — a notification, an incoming call, someone else breaking in. This difference isn't a matter of how willing someone felt about it; it's a real cost difference measurable in reaction time and error rate.

Why it happens

Self-initiated interruptions tend to occur at a moment the person themselves judges to be suitable — usually right at a natural pause point in the primary task, where the goal state has already advanced to a relatively complete, easily-marked stage, leaving clearer contextual cues attached to the suspended state. More importantly, there is a brief preparation window before a voluntary switch: right before actually switching away, people tend to perform a short, often unconscious "wrap-up" — fixing the current thought in place for a moment — and that action itself pre-plants cues for the eventual resumption.

An externally imposed interruption has no such preparation window. It can strike at the messiest point mid-execution of the primary task, with no chance to tidy up before suspension, leaving fewer and vaguer cues for resumption. In addition, an external interruption is usually accompanied by an extra orienting response — first having to process the judgment of "what just happened, should I deal with it now" — a step that doesn't exist for a self-initiated interruption, since the person already knows exactly what they're doing.

Studying it

A common approach is a between-groups experimental design: two groups perform the same primary task; one group can freely choose when to pause and handle a secondary task (the self-initiated condition), while the other is forcibly interrupted by the system at fixed or random times (the externally imposed condition). Post-resumption reaction-time cost and error rate are then compared between groups.

Common independent variables: how the interruption is initiated (self vs. external), and the predictability of an external interruption (whether it is preceded by a warning). Common dependent variables: the post-resumption reaction-time increase, error rate, and subjectively reported sense of disruption.

In HCI research, this comparison is commonly used to evaluate timing strategies for notification systems: allowing users to choose when to deal with a notification (a "handle later" option) produces lower resumption cost than the system forcing a pop-up interruption, providing direct evidence for whether users should be given autonomy over interruption timing.

A methodological caution: genuine "self-initiation" is hard to fully simulate in a lab — participants know they are in an experimental setting, and their motive for choosing to switch differs from a real work scenario (such as genuinely forgetting what to do next). The gap measured in the lab may be smaller than what occurs in real settings.

Where it stops holding

• This assumes the user genuinely has a real choice, not a forced interruption merely dressed up as a "suggestion." If declining to handle it carries a downstream cost (the notification persisting, an overdue reminder), it is still, in substance, externally imposed — it does not count as self-initiated.
• The advantage of self-initiation is offset by excessively frequent self-switching. If a user develops a habit of frequently switching away on their own to check other information, even if each switch has a lower individual cost, the accumulated total cost can still exceed that of a few externally imposed interruptions.
• Predictability of an external interruption narrows this gap. If external interruptions always occur at predictable times or come with a clear warning, the user gets a chance to prepare mentally in advance, and the cost moves closer to that of self-initiation.
• This compares the cost difference caused by who initiates the interruption, not the similarity between the interruption's content and the primary task — that is a separate, independent moderating factor.

Applying it

• Give users autonomy over when to handle an interruption: for non-urgent notifications, prefer a form that can be deferred (a badge, a "remind me later" option, batched summaries) rather than forcing a pop-up that interrupts the current action.
• Do not dress up a forced interruption as "handle later." If deferring is accompanied by persistent visual nagging or an escalating reminder intensity, the user is effectively forced to keep it in mind the whole time anyway, which forfeits the protective benefit of true self-initiation.
• For interruptions that genuinely need immediate handling, provide a warning window where possible (a few seconds of audio cue in advance), giving the user a brief moment to wrap up on their own terms — even though the interruption is ultimately externally triggered, this still lowers the cost somewhat.
• How to check: compare users' reaction time and error rate on resuming the primary task under a notification policy that lets them choose the timing versus one where the system forces interruption at a fixed time. A larger gap indicates the current design has handed too much of the interruption decision to the system rather than the user.

Related

• Same group: A5.08.1 Interruption timing should be chosen at subtask boundaries · A5.08.2 The system should preserve the state and position from before the interruption · A5.08.3 A measurable resumption lag exists before performance returns to pre-interruption levels · A5.08.4 Resumption cost has two components — re-executing the primary task and remembering the interrupting task — that must be accounted for separately · A5.08.6 The more information an interruption shares with the primary task, the more reconstruction work resumption requires
• Nearby: A5.14 Interruption timing and break points
• Search terms: self-interruption · voluntary task switching · interruption cost · notification timing