Don Knuth, one of the fathers of modern computer science, has this to say on email:
I have been a happy man ever since January 1, 1990, when I no longer had an email address. I’d used email since about 1975, and it seems to me that 15 years of email is plenty for one lifetime.
Email is a wonderful thing for people whose role in life is to be on top of things. But not for me; my role is to be on the bottom of things. What I do takes long hours of studying and uninterruptible concentration. I try to learn certain areas of computer science exhaustively; then I try to digest that knowledge into a form that is accessible to people who don’t have time for such study.
His distinction captures a remarkable fact for those of us who keep our email tabs open all day: email allows anyone to interrupt you, at any time, for any reason.
Imagine describing the analogous interaction design pattern for old-school snail mail to a version of yourself from thirty years ago, before the widespread emergence of email and the web: “Well, instead of having your mail delivered once per day, it will be delivered constantly, at all times, with each new piece of mail. And each time a new piece is received, the postman will call you, and tell you who the mail is from, what it looks like it’s about, and so on. If you want, you can have the postman open the mail, and you can hear its contents.”
Gmail has acknowledged this absurdity with its recent introduction of tabs; emails sorted into the Promotions tab do not trigger the same kinds of notifications as emails sorted into the Primary tab, etc.
But when you step back for a moment, it’s easy to see how the problem is much, much bigger than just email: the notification design pattern has crept its way into the interaction design of almost every major tech product we use.
New text? Buzz buzz.
New comment on a facebook thread? New notification.
New mention on twitter? New notification.
To be fair, not all notifications interrupt attention at all times: whereas the buzz of a text message will probably distract you no matter what you’re doing, notifications in things like facebook and twitter are just a bit softer and more customizable attentionally, and won’t necessarily interrupt you, unless you hear an associated sound, or opt in to a buzz on your phone, etc.
The problem is, whether they happen to interrupt your attention or not, they reinforce a cycle of behavior that benefits the app creator, but damages you, as a user. I call it the “(Interrupt)-Check-Interact-Reward” cycle.
Here’s how it works—if you’re familiar with classical conditioning, this story will be a familiar one to you:
Let’s start with the end of the cycle since it’s the most straightforward. Using facebook, Gmail, twitter are satisfying activities; that’s what keeps us using them. Call it fun, productive, pleasurable, whatever: using the software is a positive experience by our definition, and that’s what keeps us coming back. This is “Interact-Reward.”
Notifications that interrupt your attention take this cycle outside of the world of the app. Instead of opting to visit facebook.com and have a rewarding interaction, facebook decides it’s going to interrupt whatever you’re doing, and notifies you of some activity. Suddenly, the behavior that you go through to get the reward looks slightly different: you are interrupted, you check your notifications, you complete your interaction, and you get your reward. This is Interrupt-Check-Interact-Reward.
Already this starts to make us a look a bit like Pavlov’s dogs. Whatever pleasure we got out of the interaction on the app originally, we learn to associate with the new stimulus of the notification itself. We start to crave the notification for its own sake, instead of the interaction that the app is designed for.
But it gets worse. The “leaky,” soft notification that I referenced above, which is how facebook and twitter operate by default, doesn’t always interrupt you when you have a “notification.” This is where the parenthetical “(Interrupt)” part of the cycle comes in. Associating the notification with the reward of using the software and then withdrawing it some of the time reinforces the desire for the Check part of the cycle.
Does that make sense? Imagine if your phone only notified you of texts some of the time, and you didn’t have full control when or why. You’d get caught in a loop of wanting to check your phone for texts, wondering whether anything had happened since the last time you checked. And if, just often enough when you checked, someone new had texted, this behavior of Checking would be reinforced by the pleasure of seeing the new text, and the Checking behavior itself would come to be what you craved.
This is why facebook stopped emailing copies of notifications to people’s inboxes by default: they learned that emailing people less about what’s happening in the app actually keeps them coming back more. If the Interrupt always happens, why bother Checking? This is like randomly checking your phone for texts even though you haven’t noticed a buzz or notification. You might do it once in a blue moon when you’re bored or an awkward situation calls for it, but since the Interrupt always happens, it’s what you associate with the pleasure of receiving a new text, and you don’t feel a compulsion to constantly Check. Once the Interrupt is selectively withdrawn, the Checking behavior is what becomes associated with the pleasure of the interaction.
Notifications start to look like a great way to keep you glued to an app, and not such a great way to serve your interests as a user.
Indeed, facebook, Gmail, twitter, et al. all suffer from the same fundamental dilemma: they profit from advertising, and the longer a user spends on the site, the more ads they can be served. So on the one hand, in simple terms, it’s their absolute goal to maximize time on site for every user.
On the other hand, time on site is clearly inversely correlated with good design for a whole host of tasks from the user’s perspective. All other things being equal, the longer it takes to compose an email in gmail, to search for an old tweet, to look up a friend on facebook, the worse, for the user. Longer times indicate more difficulty doing whatever it is the user intended to do.
Put another way, to maximize user happiness/productivity/goals, what is the optimal number of times that a user should check gmail, facebook, or twitter, in a day, for example?
Of course, this question has no answer—not all users are alike, not all days are alike for any given user, and so on. But app designers and the companies who pay them have to strike a delicate balance between their short- and long-term interests: in the short-term, their interests are heavily aligned towards maximizing the number of interactions that you have with the app, regardless of whether or not it’s to your detriment. In the long-term, ruthlessly squeezing logins out of you may hurt them if they fear that it will eventually cause so much fatigue in you that you’ll abandon the app, having recognized that it was all just a con game to get you to login again, and again, and again.
But is that a realistic fear for facebook, twitter, Gmail? I am reminded of this brilliant piece by Alexis Madrigal on the “machine zone.” It’s the place that MIT anthropologist Natasha Schüll says that slot players go when they’re at the slots, and, perhaps surprisingly, it’s not about winning money at all: it’s a hypnotic, trance-like state where distinctions between you and the machine, even feelings of time and space, feel like they have fallen away. These are the gamblers’ words, not mine.
Reading Madrigal’s piece, I wonder: are modern web products the most sophisticated addiction machines in the history of mankind? Schüll’s book is called Addiction By Design, and it seems clear to me that notifications in most modern web products are designed to addict, not to help us live better lives.
The test for the appropriateness of the notification design pattern is simple, because the modern mobile phone is essentially just an always-on notification machine. Wherever we see notifications, we should ask: would it be absurd if, instead of receiving these notifications, I received a phone call or text letting me know of the same activity? If the notifications fail this test, they are probably designed to addict, as well as or instead of being designed to be informative.
This is not an entirely fair test, and it is oversimplifying. A phone call is not a text is not an email; there is a time and place for all sorts of information and interruption. But I think it still stands as a useful thought experiment to begin to distinguish between well-designed notifications and those that are designed to create a coerced, unbreakable loop of app usage.
Let’s kill notifications that exist to addict and set each other free from the never-ending arms race of cheap con games to compete for user attention.
We set out to build software to help communities we are passionate about, to solve their problems; what a travesty it would be if we allow ourselves to become the designers of ever-more-sophisticated ad-serving slot machines.