I Remember Clippy: Who do we want our computers to be?

Back in my freelancing days, I wrote a piece about the history of Clippy, the Microsoft Word Office Assistant, for an online literary magazine called The Y. That magazine is now defunct, and this article disappeared with it. I felt like it needed to see the light of day — not because it is particularly well-written, but because the story of Clippy needs to be shared.

I’ve been using the same word processor for the past 12 years. While others move on to the zingier applications of OpenOffice or Google Docs, I fall behind. Stagnated, at age 21.

Why? you might ask. Why put yourself through the agony of downloading doc to docx converters and application plug-ins? I default to the following argument: Number one, I don’t like Cambria. Number two, paragraph breaks are awful. I want my paragraphs spaced out in single lines, rather than having them lurch forward into two-line monstrosities. And I think it’s dumb that you have to add x to the file extensions of your Word files, like some godforsaken treasure map with no particular destination.

But the real reason why I can’t let go of my word processor is a little simpler: I really like my Office Assistant, Clippy. Remember Clippy? Groucho eyebrows, puppy dog eyes, paper clip wire perpetually bent into a smile.

Yes, I still have Clippy.

As I turn on my laptop and open my word processor, Clippy will zoom into view, his wire artfully bent into a bicycle shape, eyes forming spokes on the wheels. What would you like to do? he asks. Once, at a lecture, I forgot to mute the sound on my computer. When Clippy bounced onto the scene, clanging and clanking and making a racket, a few people turned in their chairs, wondering at the source of the commotion.

Clippy’s antics don’t resemble anything that exists in the up-to-the-minute technological world. He is too noisy, too medieval. I am singled out for the overfriendliness of my Office Assistant.

In December 1993, designer Kevan Atteberry was asked to interview for a position on the Microsoft Social Interface Team. Atteberry, a convivial children’s book illustrator who liked to wear his hair long and donned Hawaiian shirts to work, recalled in an interview the “mesmerizing” environment of the Microsoft offices, an imposing stronghold guarded by large-necked security officers and winding pathways, with loud music and toys strewn across the rooms.

The project Atteberry was invited to work on was a suite of applications called Microsoft Bob.

Back in the 1990s, computers were new and scary. Microsoft Bob was designed for the novice computer user — an application so intuitive and user-friendly that somebody completely unfamiliar with the computer would be able to be up and running immediately. “Bob is so helpful, you don’t even need a manual. All you need is an 8 megabyte computer,” proclaimed an early advertisement in Wired Magazine.

Bob’s central interface took place in a colorful virtual room environment that resembled a personal home. The home could be customized with number of styles: a garage, a castle, a spaceship, or a log cabin. Inside the home, there were iconic elements that, once clicked, would navigate the computer user to the application she wanted to use. If you wanted to write a letter, you could click on the piece of paper lying on the desk. This would open Bob Letter Writing Application. If you clicked on the calendar, this would open Bob Calendar. The other applications followed a similar pattern: Bob Checkbook Writing Application that helped balance checks, Bob Household Manager Application, which contained templates to enter information about car maintenance and medical records, Bob E-Mail (sic), Bob Address Book, Financial Guide, Great Greetings, and GeoSafari, a child’s geography quiz game.

All these programs were to be facilitated by interactive characters that would reside on the lower right side of the screen. The animated characters would communicate with the user via cartoon speech bubbles to help the user plumb the depths of the big, bad computer. They would answer the novice computer user’s questions in an approachable, “comical” way without the need for them to read an intimidating user’s manual.

Microsoft had run focus groups using an animated waterfowl as a computer helper a few years earlier. In an interview with The Wall Street Journal, Bob program manager Karen Fries remembered one man’s response to the character: “This guy was very emotional about it — he grabbed my arm … He said, ‘Save all the money on the manuals and just give me this duck to always be there and tell me what to do.’ ”

This is where Atteberry would come in. He was going to design these characters.

“They called them ‘interactive character technology,’ ” Atteberry said to me in an interview. “To me, it was a lot simpler. I’m gonna be making cartoons.”

Atteberry (who used a Mac) was put to work right away, directed to produce characters that appeared intelligent and interactive, but not so intelligent that they would appear unpleasant. Every Bob character had a gimmick. The dog, Rocky, rewarded the user by wagging his tail and rolling over. Chaos, the cat, fell asleep if he sensed the user was being inactive. And Java, the dragon, was extremely excitable — a symptom of drinking too much coffee.

The team had to use a proprietary vector animation software that Microsoft had created for the sole purpose of designing these characters. “It was a little clunky, and it was in constant development, and it had a tendency to convert the vectors to not so attractive bit mouse in the animation process,” Atteberry said. “But it was fun to see our characters take on a life and a little bit of personality.”

Microsoft had also hired two Stanford University professors, Clifford Nass and Byron Reeves, to examine the patterning of human behavior with computers. For three years, Nass and Reeves worked on the project, producing several social theories applicable to human interactions with technology.

The bulk of Nass and Reeves’s research found that humans often give computers human attributes, also called anthropomorphization. “We respond to our computers like we would to other people — screaming at them when ‘they’ make a mistake, praising and talking sweetly to them when they do what they’re supposed to do,” they wrote in a 1995 article published in USA Weekend.

The main takeaway, according to Nass and Reeves, is that we’d enjoy working on a computer more if we had a friend or tutor by our side. If people already naturally humanized their technological gadgets, why not make the transition easier by creating actual sentient characters to guide our computer tracks?

In that spirit, Microsoft Bob was to be the “novel blend of social psychology and computer science.” “Virtually all interfaces have a personality,” Nass said in a 2010 issue of New Scientist. “Personality creeps in everywhere because our brains are wired to quickly look at people and predict behavior.”

From the sketch stage all the way to the animated stage, the characters were put in front of volunteers, who would judge the characters on a number of set criteria: likeability, trustworthiness, perceived intelligence, and friendliness. The best-evaluated characters were incorporated into the final version of Microsoft Bob 1.1.

Some characters were rejected due to their physical size — a cow was passed over because he could not fit into the house. Others were discarded because they were not family-friendly enough. Martina, the talking olive, was said to have a “negative alcohol connotation,” so she had to go, too.

Among the chosen menagerie of 12 final characters were Blythe, the efficient bumblebee; Hopper, the little blue bunny; Ruby, the aggressive parrot (and Bill Gates’s favorite character); and Digger, the friendly green earthworm. Clippy, the perky office paperclip, had yet to be created.

Nearly 2600 years ago, the Greek philosopher Xenophanes coined the term “anthropomorphism” — anthropos after “human being,” morphe after “shape” — after observing that people tended to worship gods that resembled themselves. Ethiopians bowed to dark-skinned deities, while the Greeks depicted their gods as red-haired and blue-eyed. “But if cattle and horses and lions had hands or could paint with their hands and create works such as men do, horses like horses and cattle like cattle also would depict the gods’ shapes and make their bodies of such a sort as the form they themselves have,” he criticized in Die Fragmente der Vorsokratiker.

With Xenophanes’ ideas in mind, primatologists have observed a strange behavior in chimpanzees called the rain dance. Performed toward the beginning of the rainy season, the chimpanzees will sway in slow motion when lightning and thunder appear, displaying signs of aggression to the hovering sky.

“Those primates may well be ‘chimpomorphising’ the storm,” writes Douglas Fox in a 2010 issue of New Scientist, “shaking their sticks at the hairy-knuckled Zeus who is hurling lightning bolts from the great treetop in the sky.”

Three researchers from the University of Chicago found in 2007 that we are more likely to anthropomorphize when we feel lonely. They presented subjects with descriptions of consumer gadgets such as Clocky, an alarm clock that wheeled itself away when the user pressed the snooze button, and asked people to rate the gadgets based on human qualities like “emotion,” “free will,” and “consciousness.” Those who saw the most human-like traits in Clocky and others were the ones who were demonstrated to be lonelier in personality surveys.

The researchers hypothesized that a person feeling socially isolated would think in anthropomorphic terms as a means to recover from their pain. fMRI scans supported this finding, showing that thinking about unpredictable gadgets like Clocky led to increased activation of the ventromedial prefrontal cortex, an area of the brain known to be associated with social functioning.

The process of anthropomorphizing as a means to counteract loneliness might extend to a certain digital paper clip.

“As a freelance writer I spend the bulk of my days alone — with just my paper clip for company,” writes Catherine Jarvie in a 2003 issue of The Independent. “To you, Microsoft’s Clippit is an animated office assistant available on most word-processing packages. To me, he is a supportive colleague and friend.”

“Clippit really comes into his own during the long pauses of downtime: the gaps between emails and phone calls. That’s when his toe-tapping routines, his impish smile, cocked eyebrow, and cheeky wink are most appreciated.”

Though anthropomorphism shows no direct benefit to the human race, it’s possible that it could serve as a protective sealant against the ravages of social isolation. It’s been shown prolonged periods of isolation can shorten lifespan of humans and fruit flies, and those afflicted with chronic loneliness have a greater risk of obesity, memory loss, and depression. So anthropomorphism could allow us “to sort of tread water while searching for that real sense of connection,” said Adam Waytz, one of the researchers in the study.

Bob died on its feet, selling a meager 13 million copies in its first year, but the technology for the assistants remained. Bob’s Social Interface Team was adopted by the Office family, and the interactive character technologies got considerably more sophisticated from there. Clippy, a second generation Bob Assistant, was also incorporated into the move.

At Office, Clippy’s animation was taken over by artist John Michaud. Michaud decided to model Clippy’s movements after gee-whiz boy hero Tom Terrific, a shape-shifting 1960s cartoon character. “I figured that a mute paper clip character would most naturally communicate by rebending himself into shapes and symbols, like Tom Terrific,” Michaud said in an interview.

Clippy was imbued with several new motions corresponding to Office commands — and whether he was bending into a bicycle, zooming off in an airplane, or fashioning himself into a pair of glasses, he never lost his characteristic good cheer.

Microsoft coined two phrases for Office 97: Intellisense and Discoverability.

Intellisense was developed to provide help based on an understanding of the context of the user’s actions (AutoCorrect being one of the more recognizable Intellisense features), while Discoverability was built once Microsoft realized that a majority of users only took advantage of 35–40% of the program features embedded within the program. Office 97, with the help of its resident expert Clippy, sought to incorporate the key themes of Intellisense and Discoverability into its interface.

Michaud spoke fondly of his experience at Microsoft. “We were all excited to be working on something brand new that would be seen by millions of people and might make using software easier for people who were intimidated by computers, similar to the goal of Microsoft Bob.”

Yet it’s questionable as to whether this goal was reached, at least in the case of Clippy. A 1997 Guide to Microsoft Office for Dummies contains exactly 2 pages of instructional content devoted to decoding the charms of the animated paper clip. One page is dedicated to how to make your Office Assistant go away. The other reads like a paean to the loveableness of the animated dog, Rocky — hardly fodder for such lofty aims as predictive intelligence or easy discoverability.

No matter. By 1997, Clippy was everywhere. Users around the world popped in their Office 97 CD-ROMs to find a paper clip blinking innocently at them, eager to assist in any way. Some were pleased. Others thought it was a joke. Then, the backlash came. Poems were penned, t-shirts were created, slash fiction written. One of the Internet’s first viral videos — well before the age of YouTube — depicted a bearded man ripping out the eyes of a life-sized paper clip, shouting “Enough, you fucking paper clip!” A widely circulated cartoon portrayed Clippy grinning shiftily at a Word document as he cheerfully stated, “Looks like you’re committing suicide! … First, tell us how you plan to kill yourself.”

In response, Microsoft simplified the process of hiding Clippy in Office 2000, reducing his function to a right click and disable button. In Office XP, the Assistant was taken off the default functionality of the Office programs. And by Office 2007, Clippy had been retired for good, replaced by a search tool and an office ribbon. To celebrate Clippy’s final burial, Microsoft ordered a public execution of the Office Assistant, where frustrated users could fire staples at the offending paper clip.

Frustrated by the bureaucracy surrounding the company’s corporate environment, Atteberry departed Microsoft shortly after the final Office Assistants were selected. He returned to the world of children’s illustration, where he won several Newbery awards. He does not tell his clients he designed Clippy.

Michaud, too, is hesitant to show too much enthusiasm from being involved the Office Assistant project.

“Now I only confess that I animated 2-D Clippy (not 3-D Clippy) if someone asks if I’ve animated anything they might know,” he said. “And I brace myself for their reaction.”

So, what was the problem?

It wasn’t so much that users didn’t find Clippy human, or at least human-like. In a study performed at the Free University of Berlin, 105 test subjects of various ages and disciplines were asked whether they were familiar with the office paper clip. Many responded with the comment, “Oh yeah, I hate that guy!” Emphasis on the word “guy.”

In 2003, a Stanford computer science student named Luke Swartz defended his master’s dissertation on the ’90s Office Assistant, in search of the reason why users found Clippy so repellent. (Professor Clifford Nass, Microsoft’s perpetrator of computer-based anthropomorphism and the symbolic originator of Microsoft Bob, acted as his adviser.)

Swartz first ran a qualitative study on the paper clip and concluded that many users found Clippy too cute, too annoying, or too patronizing. Clippy obstructed the user’s visual field, his wide grin blocking the text that the user typed on the screen. And he had an unsettling way of providing proactive help. “I don’t want it to think you need help … I want to ask for it,” one subject stated.

More interestingly, the use of Clippy was found to negatively impact work efficiency. “Results showed again and again that the distraction was huge,” said Ben Shneiderman, a professor at the University of Maryland in a 2010 article in New Scientist. “The users’ performance was worse in that it took them longer to complete the task.”

Swartz (now a product manager at Google) then quantitatively examined the pairing of an interactive character’s personality to its perceived function. Further, an interactive character that guarantees easy functionality ought to execute that promised functionality. If Clippy said he was going to provide assistance, he should’ve been able to provide legitimate assistance, rather than oscillating in an infinite “I see you are writing a letter. Would you like some help?” loop no matter how many times the user had previously rejected his offer. In short, nobody cared how cute or loveable Kevan Atteberry or John Michaud made the animated paper clip if he simply didn’t do what he was advertised to do. Clippy’s friendliness was forevermore misperceived as asininity.

A year after the release of Office 97, Nass was commissioned by Microsoft a second time to optimize its failed interactive character technology. Was there a way to make Clippy more likeable?,they asked.

“The most powerful strategy I found was to create a scapegoat,” Nass wrote in the book “The Man Who Lied to His Laptop.” His redesign turned Clippy mean. After Clippy answered a question, it would proceed to ask, “Was that helpful?” If the user selected “no,” Clippy would say, “That gets me really angry! Let’s tell Microsoft how bad their help system is.” An option to send an e-mail to the manager of Microsoft Support would pop up with the subject, “Your help system needs work!” After the user had spent a few minutes typing a complaint letter, Clippy would say, “C’mon! You can be tougher than that. Let ’em have it!”

Nass and his team showed the improved character to 25 randomly chosen computer users, and people were ecstatic with the change. Choruses of “Clippy is awesome!” or “He’s so supportive!” or “I wish all software was like this!” echoed through the halls. “Virtually all of the users lauded Clippy 2.0 as a marvelous innovation,” Nass wrote.

And so, the original perpetrator of the Clippy problem had discovered its resounding solution. “Without any fundamental change in the software, the right social strategy rescued Clippy from the list of Most Hated Software of All Time,” Nass wrote. “Creating a scapegoat bonded Clippy and the user against a common enemy.” With one fell swoop, Nass had eliminated the creation of hyper-advanced artificial intelligence technology as a necessity for likeability. It didn’t really matter how complicated the mechanics were, so long as the animated character acted like a complete jerk.

Nass’s proposal was not adopted by Microsoft.

Microsoft Bob and the early iterations of Office, so eager to please and be liked, ingratiated themselves to lonely people by offering temporary companionship. But what’s the defining feature of active, moving personhood? Anger. The folks at Microsoft were trying so hard to make the characters likeable that they failed to make them truly human.

In a 1997 issue of Network World, Mark Gibbs writes the following about Clippy: “I wonder … what Assistant really is. I would guess that today’s Assistant is probably the first incarnation of a real Assistant — a process that will learn your habits and preferences and monitor your work environment.”

Such a comment, published 15 years ago, proves surprisingly prescient.

Clippy’s evolutionary progeny is most fittingly the iPhone agent Siri, or perhaps the measured female voice of your trusty automobile GPS. But comparing Clippy — visible, eager to help, anticipating every move but getting it wrong the majority of the time — to its contemporary counterparts reveals a conspicuous difference. While Clippy will go through the motions of helping the computer user but will never actually deliver such help, Siri and her ilk readjust, recalibrate to our needs. But not without a cost: they make us wait. “Recalculating,” the GPS says (angrily?) every time you make a wrong turn. If we make too many navigational demands in a short space of time, she eventually shuts down.

Waiting isn’t necessarily a quality limited to our interactive character technology. For all the diversity our technological gadgets promise, don’t they all make us wait, to a certain extent? Media-sharing sites like YouTube and Pandora have begun to package stymieing advertisements into their interfaces. Gchat puts a time stamp on text clusters whenever the conversation starts to lag. It’s an experience not unlike being put on hold by Customer Service. We might get angry, but we don’t hang up because we’ve convinced ourselves we can’t live without their assistance.

There are some who don’t agree with the notion of anthropomorphism at all. Professor Ben Shneiderman of the University of Maryland stated in a paper that the best kind of computer technology is the kind that is flawless and completely devoid of personality. We prefer neutral machines, he said, because they put us in the center of the action, firmly in the driver’s seat. We are happiest when we are the masters of our own destinies.

It’s a fair assessment, but is that what we really want, absolute control? Pushkin once wrote a fairy tale about a man who catches a magic fish and is granted one wish. He asks for a new washtub, exuberantly goes home to his wife, and gets scolded for not asking for something greater. The man revisits the fish and asks for a new cottage the next day, and the fish complies. The wife is still not satisfied. The man returns the day after and asks for a marbled mansion, and the day after a Tsaritsa’s castle, and the day after a king’s palace. But as the wife approaches the zenith of dominance, she becomes more and more unhappy. The greatest motivator and source of happiness is envisioning that which we cannot have. so what is it we really want from our computers, total compliance or total human-hood? The two are mutually exclusive.

In the Winter 2012 issue of n+1, David Auerbach penned an article titled “The Stupidity of Computers.” In the article, Auerbach cites a primitive chatbot named MGonz erected in 1989, which managed to not only convince another user at Drake University that it was human, but also “browbeat this confession out of him:”

mgonz: when was the last time you had sex
 someone at drake: yesterday.
 mgonz: ok honestly when was the last time you got laid
 someone at drake: i told you it was yesterday.
 mgonz: you are obviously an asshole
 mgonz: ok honestly when was the last time you got laid
 someone at drake: ok ok it was over 24 hours ago. for you it must have been 20 years

“MGonz had no understanding of what the other user was saying,” Auerbach wrote, “but succeeded … through ‘profanity, relentless aggression, prurient queries about the user, and implying he was a liar when he made responses to these.’ ”

Perhaps the true appeal of machines is the way they address our need to be accepted by a higher power. When a computer responds to our demands with anger, we don’t leave. We stagger back, desiring more. We are the desperate, overager Clippies of the relationship.

“Do I need to point out that … our relationship was entirely one-sided? Let me point it out anyway,” Jonathan Franzen said of his beloved BlackBerry in a commencement speech at Kenyon College. “Let me further point out how ubiquitously the word ‘sexy’ is used to describe late-model gadgets.”

Sexy intimates that which we cannot have. Sexy is all about fantasy. The computers have come to resemble us because they’ve imbued into their systems the crucial characteristic of humanhood: our nastiness.

Alternatively (a scarier thought), maybe we’ve imbued into our human systems a crucial characteristic of computers. In a New York Times Magazine Riff, Carina Chocano notices that a dinner party argument surrounding Burt Reynolds and iconic mustaches of the ’70s is quickly resolved following a quick swipe of the Google search function. And recently, an entire generation of texters at the Indiana University–Purdue University Fort Wayne were diagnosed with phantom vibration syndrome, reporting cell phone vibrations against their legs when, in fact, no messages had been received. We’re all cyborgs now, Chocano intones. What was initially conceived as a secondary brain supplementary to our own nervous systems is quickly congealing into a mutant superbrain — part flesh, part metal — before our eyes.

What does it all mean? Well, for starters, we’ve never been more perfectly isolated. Even as we chat or Skype, surrounded by friends from our past, friends we’ve never met, or people who are not our friends but would like to be, we do it alone, in repose, laptop perched on our bedroom mattresses. Clippy annoyed us so much that he forced us to get up from our chairs and talk to real people. But what happens your technology begins to represent a more perfect version of yourself? Then what’s the point of going outside? Everything you need is perfectly contained in an electronic gadget. We’ve moved beyond the clumsily formed love letter, turning to to-do lists for memory, SpellCheck for error, the Tweet for news. Our only companion, the glassy-eyed Clippy, has dissipated into the primordial ether of the “meme.”

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.