Skip to content

When a feature is invoked more often accidentally than on purpose, it should be considered a bug

Back in 2014 I tweeted this:

I’ve been meaning to revisit that statement for a while now. The link above refers to the following misfeature afflicting Mac users with external displays:

When the mouse cursor touches the bottom of an external display, MacOS assumes you want the Dock to be there, and moves it there from the primary display, often covering up what you were trying to click on.

This happens to me almost every day for several years now – never intentionally. I have looked into it thoroughly enough to know that it cannot be turned off without sacrificing all other multi-monitor features.

Our devices are full of such annoyances, born from designers’ attempts to be helpful. Sometimes they are just irritating, like the above. Sometimes they can be downright destructive.

Naked keyboard shortcuts

Keyboard shortcuts without modifier keys can be fantastic productivity enhancers, as any advanced user of Photoshop or Vim knows. But they are potentially incredibly dangerous, especially in applications with frequent text entry. Photoshop uses naked keyboard shortcuts (to coin a phrase) to primarily select tools or view modes. This may sometimes cause confusion for a novice (like when they accidentally enter Quick Mask mode with ‘Q’), but is rarely destructive.

Screenshot of Postbox Message menuPostbox, my email client, on the other hand, inexplicably uses naked keyboard shortcuts like ‘A’ to Archive and ‘V’ to move messages to other folders. What were they thinking? Email does involve frequent text entry. If you are rapidly touch typing an email, and the message composition window accidentally loses focus (e.g. due to the trackpad), there is no telling the damage that you can do. You may discover (as I have, more than once) that messages you know to have received have disappeared – sometimes only days later, without knowing what happened.

Any application that uses naked keyboard shortcuts should avoid using them for actions, as the application may come to foreground unintentionally. It’s safest to use them to select modes only.

Apple Pay

Here’s another example. Ever since Apple Pay came to iOS, I see this almost every day:

iPhone showing Apple Pay interfaceHow often do I actually use Apple Pay? About once a month, currently. Every time this screen appears unintentionally, I lose a few seconds – often missing the critical moment if my intention was to take a photo. (It is invoked by double-pressing the hopelessly overloaded Home button. A too-long press invokes Siri, which I also do unintentionally about 1 in 3 times.)


After Apple announced yet more gestures in iOS 10 at WWDC last week, @Pinboard quipped

On touchscreen devices, gestures are a powerful and often indispensable part of the UI toolkit. But they are invisible, and easy to invoke accidentally. As I wrote in my recent criticism of the direction Apple’s design is taking, whilst some gestures truly become second nature,

[…] mostly they are an over-hyped disaster area making our devices seem not fully under our control, constantly invoked by accident and causing us to to make mistakes, with no reliable way of discerning what gestures are available or what they’ll do.

Since I use an iPhone where the top-left corner is no longer reachable one-handed, I rely on the right-swipe gesture to go back more and more often. Unfortunately, this frequently has unintended effects, whether it’s because the gesture didn’t start at the edge, or wasn’t perfectly horizontal, or because the app or website developer intended something different with the gesture. And with every new version of OS X on a Macbook, the trackpad and magic mouse usually have more surprises in store for the unwary. I’m sure voice interfaces will yield plenty more examples over time.

Accidental activation – a necessary evil?

In my Apple design critique, I lamented the fact that it is very difficult to make gestures discoverable, as they are inherently invisible – contributing to that sense of “simplicity” which is a near-religion nowadays. You can introduce gestures during on-boarding, and hope users remember them, but more likely they will be quickly forgotten and we all know no-one will use a “help” page.

So you could argue that accidental invocation is the price we may have to pay for discoverability. Even though I rarely use Apple Pay, I am confident I know how to do so – a consequence of its annoying habit of popping up all the time.

With gestures, accidental activation may be critical to their discoverability, and if well implemented need not be irritating. For example, many gestures have a “peek” stage during which it is possible to reverse the action. Unfortunately, today’s touchscreen devices no longer have an obvious, reliable Undo function, one of the many failings highlighted by Tognazzini and Norman.

What are designers to do?

So if you design a helpful, time-saving feature that risks being invoked by accident,

  • consider the value (to the user of the feature)
  • consider the risk (of the user invoking it unintentionally)
  • consider the cost (to the user of doing so)

Is the value of the feature or shortcut worth the cost of it sometimes annoying users? How big is the cost to users? Wasting a second or two? Or possible data loss? Is the user even aware what happened? Is it possible to Undo? What is the risk, or likelihood, of this happening? What proportion of users does it affect, and how frequently? What proportion of usages are unintentional?

Failing any one of these may be enough to consider the feature a bug. Or you could fail two but the positive may still outweigh the negative. It depends on the severity:

Venn diagram with 3 circles intersecting: Low value, High risk, and High user cost

Can a feature be activated unintentionally? Is it worth the risk? The Venn diagram of inadvisability.

So designers should firstly try to ensure unintentional feature activation is as unlikely as possible – preferably impossible. But if it happens, the user should be aware of what happened, and the cost to the user should be low or at least easy to recover from. User testing will be your best hope of spotting problems early. Beta testing and dogfooding, run over a longer time period, are great at finding those problems that may have low frequency but high cost. Application developers may also be able to collect stats on feature usage, and determine automatically if a feature is often invoked but then not used, or immediately undone, which may highlight problems.

Or stick with a simple rule of thumb: if a feature is activated by accident more often than on purpose, it’s not a feature but a bug. Feel free to share more examples!