A comparison of human versus virtual interruptions

Abstract

Although a wealth of research has examined the effects of virtual interruptions, human-initiated interruptions are common in many work settings. An experiment compared performance on a primary data-entry task during human-initiated (human) versus computer-initiated (virtual) interruptions. Participants completed blocks of trials that featured either an interruption from a computer or an interruption from a human experimenter. The timing of the onset of the interruptions was also varied across trials. Human interruptions resulted in much shorter interruption lags. No significant differences were observed for the number of correct responses on the primary task for human versus virtual interruptions, but interruptions that occurred later in the task sequence resulted in fewer mistakes. The social aspect of human interruptions may have attenuated interruption lags in that condition, and it is possible that virtual interruptions may permit people greater temporal flexibility in managing their engagement with interruptions.

Abstract

Practitioner Summary: An experiment compared human- and computer-initiated interruptions of a verbal data-entry task. Human-initiated interruptions resulted in much shorter interruption lags. Virtual interruptions may permit people greater temporal flexibility in managing their engagement with interruptions.

Keywords::

• interruptions
• interruption lag
• resumption lag
• human–computer interaction
• multitasking

Acknowledgements

We thank Kristin Anderson for coding the interruption and resumption lag times from the video recordings.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

^1. A reviewer suggested the possibility that the human interruption condition, through demands due to the procedure, may have resulted in a qualitatively different type of interruption (characterised by preemptive as opposed to intentional integration, see Latorella 1999) that required participants to cease the primary task to obtain the interruption task code. We re-examined our data by sub-dividing the interruption lag into two separate epochs for a finer-grained exploratory analysis. The epochs were: (A) the time between the onset of the interruption and the onset of obtaining the information required to engage the secondary task; and (B) the time between the onset of obtaining the information and the onset of the mouse click on the ‘View Codes’ button. These data were examined with a 2 (interruption type) × 2 (epoch A versus epoch B) repeated measures ANOVA, which showed a very large main effect of human versus virtual interruptions, F(1,24) = 78.43, p < 0.001, and a smaller main effect of epoch, F(1,24) = 9.11, p = 0.006. Preemptive integration of the human interruption would predict an interaction such that epoch B would show faster responses as compared to epoch A, but only for the human interruptions. Crucially, this interaction was not significant, F(1,24) = 1.24, p = 0.28, which reflected that the underlying qualitative form of the interruption types was not different. Instead, human interruptions resulted in faster responses for both epochs across the entirety of the interruption lag, and epoch B took longer than epoch A for both types of interruptions.