Disrupted Visualization

Disrupted: Losing Focus & Productivity in the Digital Age

by Aditya "Adi" Tyagi

Northwestern University - MS in Analytics '20

Motivation

With the advent of work-from-home during COVID-19 or even much before, have you ever felt that you are losing a battle? Not just any battle, but the battle for your focused attention. On the one side, there are the activities that are part & parcel of your life. These can be working on a homework assignment, preparing slides for a presentation, working out, or even as urbane as cooking/doing laundry. On the other hand, there is the constant barrage of smartphone notifications. Even though, they make take only a few minutes/seconds to address, when you return to your original task, you find yourself thoroughly lost. This is only made worse by the high rate of at which these notifications interrupt your work. I have felt myself losing that battle, and set out to visualize their adverse impact. This led me to building Disrupted.

Methods

As a data scientist, I started out by collecting data. Lacking test subjects, I volunteered myself. As a result, this is an n=1 study of my own personal experience. You may remark, rightly so, that with such a teeny-weeny sample size, any sort of statistical inference to the larger population would be impossible.

That misses the point of the visualization. Disrupted does not profess to represent all smartphone users. Rather, it chronicles one person's battle to get through the week against the myriad digital distractions that compete for our attention. Hopefully, the visualization will convince you to be more mindful when interacting with your phone.

In an ideal world, I would've liked to collect data about the thousands, if not millions, of smartphone users and mine their phone interaction patterns at scale. I'd then try to get a sense of the collective time, we lose as a society from obsessive smartphone interaction.

Sadly, this shall remain a data scientist's fantasy. Privacy regulations mean that most tech firms are unwilling to share individual user's data. This insight shall remain locked up in one of the big Silicon Valley's HQs.

I used my iPhone 6, which I'd purchased in 2014. The Screentime App (provided by Apple) furnished a wealth of useful data about my phone interaction, notifications pushed, app-usage, etc.

I used the holy trifecta of web development - HTML, Javascript, CSS - to create this exhibit. I used d3.js to create the dynamic visualizations. Finally, I was inspired by code and examples written by Jim Wallandingham to build out the scroll functionality for Disrupted.

No HTML Tags were harmed in the making of this visualization!

Data Collection

I collected four types of data over a busy work week in the middle of the academic quarter.

1) Calendar

I fastidiously noted each of my activities on Google Calendar, and then at the end of the day noted all my activities down into a tabular format using Excel. A sample calendar is shown in Figure I (below). Each row represents an activity I was doing during a half-hour interval on a particular day.

2) Interruptions

Every time I would be interrupted by my smartphone I would note the following variables (depicted in Figure II below):

Cause: Cause of the disruption, as specifically as possible. (e.g. friend texting about weekend call)
Time Spent (sec): Mental estimate of the amount of seconds I'd spent in dealing with the task.
Location: I noted the location down where I was interrupted. (e.g. Dining hall)
App: I made a note of the app that pushed the notification.
Time (CST): Local time when I was interrupted
Activity: The activity I was engaged in when I was interrupted.
Important: Whether the interruption was important or not (more below)

While deciding whether or not an interruption was important (a subjective question), I set the following standard:

Does this event require my immediate attention, relative to the activity I am currently engaged in?

Some examples are in order:

If a project group needed my urgent input (and this was blocking project progress), and I was exercizing at that time, then this would be classified as an IMPORTANT interruption.
If the NU President launched a new COVID-19 relief fund, and I was reading my textbook at that time, then this would a NOT IMPORTANT interruption.

3) Pickups & Notifications

This was data about the number of times I picked up the phone as well as the number of notifications pushed by my phone. Both were grouped in hourly intervals. Furthermore, it is important to note that notifications were not "on" for all apps. Rather, only those apps that I usually used had notifications turned on. This was done to capture as much as possible my natural notification patterns (without conscious intervention). Lastly, to ensure that I didn't miss recording any notifications, I intentionally set the notification settings to be "loud" rather than "vibrate". A sample pickup & notification data sheet is shown in Figure III below.

Figure III: Sample Pickups & Notifications Data

4) Time Spent

This was data about time spent on each app grouped on a daily basis. Again, sourced from the Screentime App on my iPhone. This is depicted in Figure IV below. Each row represents a day of the week; each column represents the amount of time (mins) spent on that app for each day of the week.

Analytics

This raises a fundamental question about our very own human existence: autonomy. How much of what we do is truly resulting from our free will versus external prompting? In this case, how much do we really interact with our smartphones because we really want to versus because our smartphone wants to (i.e. by pushing a notification)?

To answer this question, I set out to develop an autonomy score. Simply put, the autonomy score measures how much of your phone interaction (operationalized as ‘No. of pickups’) results from your own free will vs. an external (environmental) push notification (operationalized as No. of push notifications).

To calculate an autonomy score, I trained the simplest possible model (a simple linear regression) to predict my number of pickups throughout the day (dependent variable) based on the number of notifications I received (independent variable). Then, the autonomy score is calculated as:

Autonomy score = 1 – R²_{pickups, notifs}

where R²_{pickups, notifs} is the coefficient of determination of the model.

Intuitively, the higher the R²_{pickups, notifs} , the more your phone picking up behavior is determined by external notifications pushed, and thus the lower your ‘autonomy’. The opposite is true if R²_{pickups, notifs}is low. This implies that you pick up your phone out of your own free will.

Challenges

Some of the challenges I faced in building this visualization were:

Having the discpline to collect data accurately for the entire one-week period.
Understanding the web development techniques used to build such a project (this is my first web dev project of such a scale)
Understanding how to build scroll-based visualizations in d3.js
Tying back some of the conclusions of Disrupted to scientific research in the field of productivity.

Debrief

Perhaps now, you can now fully appreciate the havoc that frequent little notifications can cause. In fact, they often disrupt important activities like studying, or sometimes even harmless pastimes like exercising or reading. Meanwhile, notifications and phone pickups seem to be strongly related. When notifications spike so do, phone pickups. By hovering over individual disruptions, you may even have noticed that an overwhelmingly large number of them are not important. They generally also take on the order of seconds to deal with. They why are they deadly? Explore the science below to find out.

The Science

Psychologists have long spoken of 'attention residue' which is the time our brain takes to fully re-absorb ourselves in a task after an interruption. Psychologists have estimated that this causes about a 40% drop in productivity.

In Deep Work^[4], author Cal Newport talks about how our ability to focus deeply on a task without distractions is dwindling, and why those who possess the ability to do deep work (defined as professional activities conducted under a high degree of focus) will be winners in the new economy. Even more than the economic aspects, there is a happiness component too. According to cancer survivor Winifred Gallagher^[5], people who focus, with fewer distractions are happier and more satisfied.

In fact, multitasking. (i.e. coding, while shopping on Amazon, while writing a slack message) is a myth. In a study^[1]by Rogers & Monsell, it was noted that participants were slower when they had to switch tasks than when they repeated the same task. In another study, conducted in 2001 by Rubinstein et al^[2]found that participants lost significant amounts of time as they switched between multiple tasks and lost even more time as the tasks became increasingly complex.

Rubinstein et al note that there are two stages to the executive control process.

The first stage is known as goal shifting (deciding to do one thing instead of another).
The second is known as role activation (changing from the rules for the previous task to rules for the new task).

Switching between tasks requires both of these steps to take place. Althought they may only add a few tenths of a second, this can start to accumulate when our minds switch tasks repeatedly. Think of immediately answering a slack message while writing code for an important project (and doing this multiple times an hour).

Prolonged 'disruptions' can have impact on the brain. Clifford Nass ^[3]found that people who were considered heavy multitaskers were actually worse at sorting out relevant information from irrelevant details. They also showed greater difficulty when it came to switching from one task to another and were much less mentally organized. To make matters scary, even when these 'multitaskers' were focused on a single task, their brains were less effective and efficient.

Takeaways

After the visualization, what would you do differently? For one, it is clear that frequent notifications can cause more harm than one might suspect. They certainly have an impact on quality of life. Well, a good first step is to resist the urge to check your phone while you are in the middle of an activity. The visualization has hopefully convinced you that it likely isn't important to what you're doing. As a long term solution, you can selectively turn off notifications for non-essential apps. This ensures that you hold the initiative when you interact with your smartphone as opposed to the phone.

Next Steps

I hope that Disrupted has convinced you of the deleterious impact smartphones have on our daily lives. While data scientists can shine the spotlight, they must collaborate with other domain specialists like psychologists, productivity experts, and engineers by:

Making smartphones that are truly 'smart' about managing notifications. For e.g., by automated classification of the importance of a notification
Building context-aware products that are capable of sensing user context (we already have context-aware recommendater systems)
Investigating further the impact of obsessive smart phone usage on one's productivity, and more importantly one's happiness.

References

1. Rogers R, Monsell S. The costs of a predictable switch between simple cognitive tasks Journal of Experimental Psychology: General. 1995;124:207-231.)

2. Rubinstein JS, Meyer DE, Evans, JE. Executive control of cognitive processes in task switching. Journal of Experimental Psychology: Human Perception and Performance. 2001;27(4):763-797.

3. Ophir E, Nass C, Wagner AD. Cognitive control in media multitaskers. Proceedings of the National Academy of Sciences for the United States of America. 2009;106(37):15583-15587.

4. Newport C. Deep Work: Rules for Focused Success in a Distracted World:. Barnes & Nobles 2016

5. 5. Gallagher W Rapt: . Rapt: Attention and the Focused Life: The Penguin Press 2009