From ATMs to Siri to the button text in an application user interface, we "talk" to our tech—and our tech talks back. Often this exchange is purely transactional: we input commands; the machine complies. Newer predictive technologies like personal assistant apps have renegotiated this relationship, preferring to relate to us as peers, even friends. Scarlett Johansson's flirtatious operating system in Her brought this idea to a lifelike apex, simulating love, even orgasm—all digitally mediated.
As technology becomes more pervasive and gains access to greater amounts of our personal data, how can we design successful human-machine conversations? Should user interface text approximate the lilt, flow, and syntax of human speech? Or does humanizing UI conversations create a false intimacy that distances even as it attempts to foster familiarity?
The answer, of course, is that it depends. Most of us have encountered voice-automated customer service systems. Some of them, in an effort to make their robot customer reps less droid-like, feature voices that try to approximate human diction. A calm, often female voice pauses, suggests brightly, bridges her prompts with almost-ums. Her attempts at realness further underscore the fact that she is fake, blocking you from an actual human encounter.
A computer that cheerfully calls you by your first name can either delight you or creep you out, depending on the circumstances. Just as robots enter the uncanny valley when they seem too human, a user interface that's too familiar can push people away. The copy needs to strike the right balance.
Consistency within diversity
Until recently, I was a UX writer and content strategist at Google. Specifically, I worked on Google Apps: Gmail, Docs, Drive—all of the productivity tools that help people work and communicate. Writing for a large, entrenched company poses particular challenges because the sheer array of products and experiences offered can make it difficult to achieve consistency in tone and style.
Our audience included people who use Google tools at work. People at work are obviously a very different market from, say, a teenager absentmindedly browsing a consumer app like YouTube for video clips (though that isn't to say that YouTube browsing doesn't happen under the guise of work). But work tools should be as intuitive and (dare I say) delightful as the best consumer apps. They should help you be more productive and creative. You shouldn't have to spend brain power figuring out how to deploy them.
Consider, too, the many modes of work. Some people are desk-bound, whereas others are constantly mobile. Some companies have huge IT departments that can provide support; at others, workers have to learn to use products on their own, often without much technical background. This can be overwhelming. Most workers want to spend time getting stuff done, not learning to use Google Apps. So the experience, and thus the text, must serve the overall usability and ease of the product.
There are ground rules that can guide the writing for all of these interfaces, though they exist in very different contexts and geographies. The YouTube-browsing teen can be in Osaka or Indianapolis. Modifying Chrome settings should be easy and seamless, whether in Farsi, Tagalog, or Italian.
While we can strive for an overarching level of consistency defined by some core principles—be friendly, be helpful, don't use jargon or technical language—each product carries slightly different conventions, expectations, and contexts. Yet they all have to be reconciled within a domain that is recognizably Google, in over 50 languages. Keeping text brief and scannable, and including only the most essential words, will smooth a user's journey.
"OK Google, search for Thai restaurants."
One particular UI conversation that Google may continue to fine-tune is how it initiates a voice search on a smartphone. "OK Google" is the voice prompt Google suggests for striking up interactions on mobile devices. This phrasing suggests that our relationship with both our phone and Google is informal and familiar, even chatty.
If you ever actually "OK Google" your smartphone or wearable device, though, you'll probably find that doing so feels forced and cheesy at best. I'd personally rather just say "Call Trevor" or "Find nearby Thai restaurants" and stick to semantic, if utilitarian, commands.
"OK Google" is awkward because it insists that we're chums with the search behemoth. Google is less a company in this recasting than a helpful friend. Yet this same Google also upholds "Focus on the user" as one of its founding pillars. "User" implies both a certain ascetic distance and an unpleasantly parasitic relationship. How can I simultaneously be buddies with and just a "user" to the same company?
Language reveals social landscapes and highlights power struggles, and can shed light on intimacy or distance. When writing for an interface, the smallest words reveal relationship dynamics and cue motivations, even emotions. The microtext on, say, a button can alter the tenor of the interface conversation. Whether you label a button "Got it" or "Continue" signals more than just information conveyance. "Got it" connotes a certain confidence and informality, and assumes agency on behalf of users. "Got it" asks them to own their comprehension and acceptance of whatever information is presented before moving along, rather than merely assenting to "Continue."
Another common copy example: "enable" versus "turn on." "Enable" feels unnecessarily technical and implies a subtle hierarchy between the enabler and enabled. The softer "turn on," by contrast, could indicate the flow of water from a faucet, or—depending on where the mind goes—a sexy precursor to further action. When aiming for "friendly," where is the line between cloying and mechanical?
Brevity with soul
Being conscious of words doesn't mean that one needs to make UI language purely functional. Balancing well-placed, clever copy with short, concise text can add delight and magic to an experience. Note Chrome's "Aw, snap" for a page-load error, or the sly personality that suffuses the airfare purchase flow on Virgin America's site:
Virgin's brand voice is flirtatious, fun, and irreverent. Their approach resurrects an earlier age when air travel promised a thrilling luxury rather than a cramped seat, broken pretzels, and the purgatory of airport security. They don't take themselves too seriously and their approach injects brassy humor into a task as lackluster as flight booking.
That tone comes through in small ways, such as this playful modal dialog about additional upgrade charges. The button is labeled not with the expected “Okay,” but with "I understand, let's do this.”
When a user enters her name while booking a flight, the form field greets her with a sly bit of text: "Hey there.” Subtle winks like this can humanize an interface without being intrusive, yet aren't so colloquial that they're alienating.
Text should inform readers and help them along—and then it should get out of the way. A well-written UI recedes into the background, imbuing—but never overpowering—the user experience. There's poetry in writing for the web, but it isn't the luxuriant run-ons of Whitman. Rather, it's the economy of poet Masaoka Shiki's haiku—so spare it's almost missed.
Friendly but functional
The popularity of "digital detoxes" hints at a growing frustration with our reliance on tech interactions. The future may well contain more unobtrusive and silently helpful technologies, rather than intimate human-machine relationships à la Her.
As such, UX writers and designers might consider how we can keep conversations friendly but functional. We can provide signposts without the baggage of a relationship. Sue Factor, a former colleague and Google's first dedicated UX writer, taught me that short text is often the best text. Although I earn a living writing for the web, I don't flatter myself that anyone opens an app to carefully read and savor my language. We've all got better things to do. As Shiki writes: