Stress and productivity patterns of interrupted, synergistic, and antagonistic office activities

We describe a controlled experiment, aiming to study productivity and stress effects of email interruptions and activity interactions in the modern office. The measurement set includes multimodal data for n = 63 knowledge workers who volunteered for this experiment and were randomly assigned into four groups: (G1/G2) Batch email interruptions with/without exogenous stress. (G3/G4) Continual email interruptions with/without exogenous stress. To provide context, the experiment’s email treatments were surrounded by typical office tasks. The captured variables include physiological indicators of stress, measures of report writing quality and keystroke dynamics, as well as psychometric scores and biographic information detailing participants’ profiles. Investigations powered by this dataset are expected to lead to personalized recommendations for handling email interruptions and a deeper understanding of synergistic and antagonistic office activities. Given the centrality of email in the modern office, and the importance of office work to people’s lives and the economy, the present data have a valuable role to play.

The algorithm turns this spatial perspiration pattern into a signal by applying a morphological filter. Elevations in the signal correspond to densification of active perspiration pores, characterizing overarousal bouts.  Figure S5: Sets of perinasal perspiration (PP) signals before and after quality control (QC1). PP signals, much like all cholinergic signals, are of exponential nature [33]. Hence, a log 10 transformation was applied to facilitate visualization.  Figure S6: Sets of EDA signals before and after two levels of quality control. EDA signals, much like all cholinergic signals, are of exponential nature [33]. Hence, a log 10 transformation was applied to facilitate visualization.  Figure S8: Sets of heart rate signals measured on the chest before and after two levels of quality control. There is evidence of measurement disruptions (i.e., zero drops) in certain cases of the original signal sets.  Figure S9: Sets of heart rate signals measured on the wrist before and after two levels of quality control. There is limited evidence of measurement disruptions in the original signal sets.

Report Prompts
Single Task (ST) Report Prompt You have 5 minutes to write about the topic below: The best way for a society to prepare its young people for leadership in government, industry, or other fields is by instilling in them a sense of cooperation, not competition. Do you agree or disagree? Explain your reasoning in detail.

Dual Task (DT) Report Prompt
You have 50 minutes to write. You will be given a topic. You can research the topic on the web. Please do not watch videos on this topic, but rather use written documents on the subject.

version for High Stress Groups
Your performance will be based on the quality of your report, the quality of your responses to email, and the quality of a 5-minute presentation you will give to a group of evaluators at the end.

version for Low Stress Groups
Your performance will be based on both the content of your report and the relevance and sufficiency of your responses to emails.
Paragraph 1: Explain what the Technological Singularity is in your own words.
Paragraph 2: Explain & discuss the view of your first Technological Singularity theorist.
Paragraph 3: Explain & discuss the view of your second Technological Singularity theorist.
Paragraph 4: Explain and discuss your own view of Technological Singularity.
Paragraph 5: Conclude your essay summarizing the views of the theorists and your own.
NOTE: Copying and pasting in the essay window are disabled.

Email 6
Title: I need help scheduling a meeting From: Petre Hinkledge Please find a time slot for a student, administrator, and professor to have an hour long meeting, in a conference room that can seat 3 people. People's schedules are presented on the first tab of the linked spreadsheet, and room availability is in the second. People have two types of availability: completely free (shown in white) and possible commitments (shown in yellow), which can be moved if necessary. You should not schedule a meeting in unavailable time slots (shown in red). Scheduling decisions should minimize the number of possible commitments rescheduled, and prioritize the schedule of the professor, then the administrator, then, lastly, the student. Reply to this email with the meeting start time and room you select.

Email 7
Title: Planing a meeting From: Svetlana Romanoff Please find a time slot for a student, administrator, and professor to have an hour long meeting in a conference room that can seat 3 people. People's schedules are presented on the first tab of the linked spreadsheet, and room availability is in the second.
People have two types of availability: completely free (shown in white) and possible commitments (shown in yellow), which can be moved if necessary. You should not schedule a meeting in unavailable time slots (shown in red). Scheduling decisions should minimize the number of possible commitments rescheduled, and prioritize the schedule of the professor, then the administrator, then, lastly, the student. Reply to this email with the meeting start time and room you select. Email 8 Title: What time will work? From: Jarla Knovak Please find a time slot for a student, administrator, and professor to have an hour long meeting, in a conference room that can seat 3 people. People's schedules are presented on the first tab of the linked spreadsheet, and room availability is in the second.
People have two types of availability: completely free (shown in white) and possible commitments (shown in yellow), which can be moved if necessary. You should not schedule a meeting in unavailable time slots (shown in red). Scheduling decisions should minimize the number of possible commitments rescheduled, and prioritize the schedule of the professor, then the administrator, then, lastly, the student. Reply to this email with the meeting start time and room you select.

Heart Rate Variability (HRV)
Data Records of HRV -Supplementary Data Folder Under the Supplementary Data folder on the OSF repository (Ref. [18]), there is a comma separated value (csv) file that holds the HRV data (15.9 MB). In this file, in addition to the columns holding the participant ID (Column A) and group information (Column B), there are columns holding treatment | task information (Column C | Column D), absolute | relative timing (Column E | Column F), and the RR values (Column G). As it is the case with all other variables in the present paper, the OSF repository holds the quality controlled RR values. The raw variable values and the R code that operates upon them, to implement the quality control and validation processes described herein, reside on GitHub (Zaman, S. & Pavlidis, I. Office-Tasks-2019-Methods. GitHub https: //github.com/UH-CPL/Office-Tasks-2019-Methods).

Quality Control of HRV
In addition to heart rate signals, BioHarness (Zephyr Technology, Annapolis, MD) reports beat to beat time intervals in ms, that is, RR values. These values are indicators of heart rate variability (HRV), a measure that can track sympathetic arousal, much like heart rate. RR values have their own set of noise problems, including ectopic beats, and thus a quality control check is in order.
The left column in Supplemental Fig. S10 shows the superimposed time series of RR values for all participants per treatment. Each RR value is depicted as a vertical line. From this visualization it is clear that there is noise in the form of very large RR values, particularly in the Presentation (PR) treatment, but also in Priming (PM) and in Dual Task (DT). The standard way to reduce such noise in RR data is to remove values that are on the tail end of the distribution (Ref. [30]). Specifically, we filter out RR values that are beyond two standard deviations from the mean, thus retaining ∼ 95% of the values, while eliminating ∼ 5% of extreme values fraught with noise. The cleansing effect of this action is shown in the right column of Supplemental Fig. S10. Please note that filtering is based on means and standard deviations computed per subject per treatment, to account for inter-individual variability.
We re-compute the means per subject per session for the RR data that passed quality control. These updated means, called NN (Ref. [30]), are considered reliable HRV indicators that can be used in analytics. Hence, the next step is to subject NN to experimental validity analysis, much like we did with the other five physiological variables.  Figure S10: Sets of RR data measured on the chest before and after quality control (QC1). To facilitate meaningful cardiac channel comparisons, the original RR sets (left column) match the Chest HR sets that survived quality control (Supplemental Fig. S8). The outliers in the original RR data suggest ectopic beats, which took place mostly during PR and DT. Continual Low Group | CL Figure S11: Experimental validity of the chest NN variable. Antithetical to heart rate boxplots in the main text, NN boxplots that are below the zero line manifest stressful treatments. Hence, NN largely captures the stressful effect of PR, but mostly misses it in ST and DT, a performance nearly on par with the Chest HR channel.

Experimental Validation of HRV
In contradistinction to heart rate, which increases with stress, HRV indicators, such as NN, tend to decrease with stress. Hence, validity for NN manifests when the normalized distribution means of stressful treatments are significantly lower than the zero line (Supplemental Fig. S11 , · · · , 63}, j ∈ {BH, BL, CH, CL}, and k ∈ {ST, Stroop, RV, DT, PR}, respectively. In each group G j , we normalize within-participant the expected NN C values by computing the distributions of paired differences between his/her NN C in T k and T RB : (S1) Equation (S1) produces the boxplots in Supplemental Fig. S11. The NULL hypothesis is that arousal within participants in treatments ST, Stroop, DT, and PR is no different than arousal in resting baseline RB.
The results indicate that NN C mirrors to a large degree the performance of chest heart rate (Chest HR) presented in the main text. Indeed, NN C cap-tures well the stress effect of PR for the High stressor groups (p < 0.01, paired t-tests in BH, CH). For the Low stressor groups BL and CL, NN C exhibits the correct trend but, unlike Chest HR ( Fig. 8a1-d1), fails to reach significance. For all other treatments NN C mirrors the disappointing performance of Chest HR. Specifically, NN C captures the stressful effect of ST only in one group (p < 0.001, paired Wilcoxon signed-rank test in BH), while misses it in the other three groups (p > 0.05, paired t-tests in BL, CH, CL). It altogether misses the stressful effect of Stroop and DT. Based on these results, Chest HR is the preferred cardiac variable for this experiment, as NN slightly trails its performance. Overall, the NN C results reaffirm the conclusion that cardiac variables tend to capture moderate office stressors, bearing an emotional component, such as presentations to an audience. At the same time, cardiac variables tend to perform poorly in mild office stressors of cognitive nature, such as standard report writing.

Quality Controlled 1 Chest HR, Wrist HR, and EDA Signals
This is an exhaustive list of the Chest HR, Wrist HR, and EDA signals that survived quality control 1 (QC 1) checks. Panels where the signals got eliminated in QC 1 feature a cross mark. Chest HR and Wrist HR signals are superimposed on the same graph to facilitate comparison. Display of signals progresses per participant per treatment, with heart rate signal panels appearing on the left and the corresponding EDA signal panels on the right. In heart rate signal panels, dashed lines indicate the respective means. Signals that are found by quality control 2 (QC 2) processes to be invalid, lend a lightened shade of their color to the panel background.