article

Only Human

Posted: 03/1999

Only Human
Moving from Technocentric to Usercentric Product Development
By Ken Branson

In 1854, the news that
Maine and Texas could communicate by telegraph left Henry David Thoreau underwhelmed.
"No one seems to have inquired," the sage of Walden Pond noted crankily,
"whether Maine and Texas have anything to communicate.

Thoreau spoke, unknowingly, for every marketing manager asked to find a market for a
new technical product, every public relations person asked to explain it over the phone to
reporters and every frustrated customer surrounded by parts and manuals and instructions,
listening to hotline-hold Muzak, and driven to the ragged end of his very last wit.

"I think a lot of companies ignore the user when they’re doing the design,"
says Linda Roberts, acting department head of user interface technology in the Bell
Laboratories unit of Lucent Technologies Inc., Murray Hill, N.J. "And they don’t take
the time to include the user when they test [the products]. A lot of times, the user is
eliminated until the product is sold."

Eugene Galanter, Gelhorn Professor of Psychology at Columbia University, New York, and
a man with five decades of human-factors work behind him, says that "technology is
moving so fast that nobody is thinking about whether anybody can use this stuff or
not." Galanter, who has spent the past four decades studying how human beings relate
to technology, says that the pace of change is much faster than it was even 10 years ago.

The pace of technological change and the pressure of the market point to a difficult
time ahead for those of us who have to log on and off, plug things in, turn things on and
get things fixed. On the other hand, the producers of products, systems and services are
aware of all this and are trying to deal with it.

PHONE+ spoke to usability specialists (sometimes called human-factors specialists) at
several companies, and found them war-weary but undaunted as they fight to make products
useful. At bottom, their case seems so simple and straightforward: First, ask what the
customers need; then design the product; then build it to the design; then test it with
some real customers and refine it; then launch it. It’s just common sense, isn’t
it?

"You might say that," says Ed Israelski, director of human factors at Hoffman
Estates, Ill.-based Ameritech Corp., "but you’d be surprised. Developers will think
they know what the user wants, and guess wrong. It happens at all companies, even those
with stellar reputations. They’ll have products coming out of some maverick development
organization where people think they know better."

Thomas Landauer would not be surprised in the least. Now a professor of psychology at
the University of Colorado in Boulder, Landauer worked for many years at the former Bell
Laboratories and at Bellcore, where he tried to make sure people at telecommunications
companies could use the support systems his employers built for them. He is the author of
"The Trouble With Computers" (MIT Press, 1996).

"We’re loading everything up with too many features, done too many different
ways," he says. "Every voice-mail system, every answering machine … works in a
different way, and you never know how a particular one will work. We just went to 10-digit
dialing (in Colorado) and nobody’s voice mail works any more … I know I couldn’t manage
to follow the instructions that were read to me 100 miles an hour over the phone.

"Well, there are wonderful things being done, wonderful features on them when they
work. But putting them all together makes for complexity and they get harder to
manage," Landauer adds.

In Landauer’s view, the pressure to get a product designed, developed, built and out to
the marketplace accounts for a large part of the problem. Whatever the product is–a new
cross-connect, a new computer game, a new feature on a software system–it has to be new,
it has to be different and it has to be first. The question of whether anyone can use it
can wait for later.

"Developers will think they know what the user wants, and guess wrong. It
happens at all companies, even those with stellar reputations. They’ll have products
coming out of some maverick development organization where people think they know
better."

— Ed Israelski, director of human factors,
Ameritech Corp.

Or, in the words that clang in Israelski’s ears: "Those are nice suggestions;
we’ll put them in the second release." Leaving aside the question of whether there
will be a second release, Israelski says the engineers who speak these words are
shortsighted. They hope to save some development money and beat the market by getting the
product out as fast as possible. If only they would take a deep breath, he says, before
they get started.

"These products can be developed without these problems," he says. "It
costs more money, but not a lot more. The rule of thumb is an extra 8 percent (of
development costs)."

And Landauer, to whom the "second-release jingle" has been sung more than
once, says time should not be an issue, either. "It doesn’t really take much
time," he says. "You just have to do it. Every study that I know of that looks
at the actual time in getting things out the door says it doesn’t add much to the
development cycle."

Producing products that perform useful work and are easy to use–be they computers,
switches, software systems or zippers–is not a dark art, Israelski says. He and his 12
human-factors colleagues at Ameritech have developed a simple five-step process that he
believes will help get the job done right every time.

Step-by-Step Approach

First, Israelski says, comes customer modeling. This means "understanding who your
users are, getting background of their experience," he says. "If you have a
complicated product, this model can be complicated. If you’re trying to model a hotel
chain, it would be different than if you’re trying to model an airline reservation
system." It always helps to talk to actual customers, of course, to observe them, to
ask them what they want and listen to their answers, he says.

Once the customer model is built, Israelski says, it’s time to create "user
scenarios." These will help developers understand what the customers are supposed to
do with the product, what tasks they are supposed to complete. "If you have your
customer models, you’ve done focus groups, you’ve hopefully talked to users," he
says. If you haven’t, however, then your user scenario may be somewhat flawed, he says.

Then comes the technical part: usability objectives. Despite the rather awkward name,
usability objectives are measurable expectations for a product’s usability. How many
complaints should you expect to receive from customers in the first month? How fast do you
expect a user to type on this keyboard? How much more can you charge for a product whose
main differentiation is usability?

Finally, there are two steps that Israelski says shouldn’t be thought of in sequence:
iterative design and usability testing. These should go on all the time, he says.

"Psychologists don’t have hard and steadfast rules for human behavior, so we have
to create prototypes, do early usability testing, see how long it takes people to learn to
use the product or do the task," he says. "If the objective is three minutes to
log on, and it takes most people 20 minutes, well, you know you have a problem."

Israelski’s counterpart at Bell-South Corp., A.C. (Chet) McQuaide Jr., says his company
follows a similar method: user-interface requirements are written at the beginning of a
product’s or system’s development cycle and usability studies happen early and often;
products are refined in the light of those studies; and limited trials precede general
deployment.

"It’s important to have human-factors people present from day one, so they
understand the time-to-market pressures that apply to a service, and also to have them
accepted as part of the design team instead of as a ‘quality control agent’ that looks at
the results after it’s done, when it’s too late to make changes," McQuaide says.

McQuaide and Israelski both point to recent product successes at BellSouth and
Ameritech as evidence that their companies, if not perfect in such matters, are at least
on the side of the usability angels.

Ameritech recently launched a product called Privacy Manager, available to those of its
customers who already have the Caller ID with Name service. Privacy Manager intercepts
calls that show up as "out of area," "private," "blocked" or
any one of the other guises that frequently disguise telemarketers selling term insurance
or time shares in the Dry Tortugas. A stored voice asks the caller to identify himself or
herself. If the caller declines to do so, the call is disconnected. The phone never rings.
If the caller does provide identification, the customer is told who is calling, and has
the choice of picking the phone up or not, or playing a prerecorded message saying
telemarketing calls are not welcome and asking that the customer’s name be removed from
the telemarketer’s database.

Early usability testing, Israelski says, showed that 70 percent of the calls screened
this way never get past the first question. The product, available first in Detroit and
Chicago, has been popular, and if customers have found it hard to use, nobody has told
Israelski.

In BellSouth’s case, the products are flexible call forwarding and call waiting deluxe.
Flexible call forwarding, McQuaide says, allows customers, using an interactive voice
response (IVR) with an administrative menu, to tell the network where they want calls sent
and for how long. Call waiting deluxe allows a customer to see a caller’s name and number
while online.

"A lot of work went into these products, with very complex usability
testing," McQuaide says. And so far, so good.

This last product, which sits at the juncture of telephony and the Internet, is the
sort of product to raise opportunities and danger signals for Charles Kreitzberg, CEO of
Cognetics Inc., a software design and consulting company in Princeton Junction, N.J.

Speaking of telecommunications systems and products in general, Kreitzberg says,
"The people who build those systems tend to be very technocentric rather than
usercentric. And that’s something phone companies are going to have to think about. Phones
are no longer about little machines with dials on them; they’re about vast, complicated
networks, they’re about transferring information. It’s an integrated system, and you have
to look at users’ needs and how to satisfy them."

In any case, the main objective for developers and designers might be summed up in
another exhortation from Thoreau: "Simplify, simplify!" And while Israelski and
his colleagues work at it from one end, some people on the receiving end of some high-tech
products have decided not to wait.

Kreitzberg, for example, has issued a call for everyone who is ticked off at
hard-to-use software to join the Logical User-Centered Interaction Design (LUCID)
computing movement. He asks people to contact his website (www.cognetics.com) with tales of woe and ideas for
making things better.

Like Landauer, Kreitzberg believes it isn’t difficult to make software easy to use, but
software producers often go wrong at the beginning of their projects.

"Usually, they ask a programmer to ask the users what they want," Kreitzberg
says. "They come back without a clear idea, get very enmeshed in the technology, and
start building the software. Then they find halfway through it that people are saying,
‘that’s not what I wanted.’"

Kreitzberg says the web has changed everything for the better, by giving many more
people a reason to use computers. Still, he believes "that in a lot of cases, there
are still engineers who drive the product design." For telecommunications and other
technology-driven companies, this is a holdover attitude from the days when the computer
was new.

"Back in the old days of mainframe systems, nobody cared about usability," he
says. "You hired people who sat behind terminals and everybody else used interoffice
mail. But the web transformed computers into a mass medium, and when things become mass
media, they have to be usable."

Then there are the folks at Isys Information Architects Inc. who build software systems
and interfaces from offices in Greensboro, N.C. They host the Interface Hall of Fame and
its much longer, varied and more interesting website, the Interface Hall of Shame.

The Hall of Fame is a fairly straightforward list of interfaces Isys thinks are worthy
of emulation; the Hall of Shame contains several different categories Isys thinks are
"worthy of extinction." Microsoft Corp. appears in both halls, but is more
prominently mentioned in the Hall of Shame. In the category "Interface
Stupidity," for example, Isys reports this error message, taken from Microsoft’s
Windows 95 operating system by a user trying to delete files from a full hard drive:
"X Error Deleting File. Cannot delete 016. There is not enough disk space. Delete one
or more files to free disk space, and then try again."

"This has got to be the No. 1 stupidest error message I have ever seen,"
Isys’ correspondent writes.

Microsoft long has been criticized for messages such as this–a fact acknowledged at
last year’s spring Comdex show by no less a personage than founder and CEO Bill Gates, who
conceded to a plenary session that some of those messages were "cryptic" and
would be improved in future Microsoft products.

But Microsoft has a usability engineering group, and is working hard to make things
better, says Jonathan Usher, Microsoft’s telecom industry marketing manager.

Usher says Microsoft is backing an initiative called Telecom Dashboard, which, when
it’s done, will help telecommunications companies sell enhanced telecom services through
their customer’s computers. This is part of an overall effort by Microsoft to convince
telecommunications companies that "NT is a great platform for intelligent network
(IN) applications such as toll free and virtual private network (VPN)," Usher says.

Usher says he and his colleagues want to help telecom companies in their drive to push
graphical user interfaces (GUIs) out to their customers, who then will use these GUIs to
order and cancel new services, and to order and provision their own lines.

("Whoopee!," Landauer says. "Just like pumping your own gas.")

So, are the advocates of usability making headway? It depends on whom you ask, and on
how they measure the data.

Kreitzberg takes hope from the web, and from the conviction that the world is full of
people who, like him, are mad as hell and aren’t going to take it any more. Lucent’s
Roberts takes heart from the career prospects of usability engineers. "Anybody
graduating now has 20 job offers," she says. "Fifteen years ago, there were five
people for every job offer."

On the other hand, there is Professor Galanter of Columbia. Galanter has taught future
architects, computer scientists and engineers how to make products usable for 50 years.
Asked if his current crop of students gave him hope for a usable future, Galanter sighs
deeply. "Well, no," he says. "Not really."

My Virtual Assistant Isn’t Much to Look at, but It’s Got a Great
Personality

General Magic, Sunnyvale, Calif., recently teamed with two social psychology experts to
develop the "personality" for the company’s intelligent voice user interface
(VUI), called magicTalk, the patent-pending platform for the forthcoming Portico virtual
assistant service.

Byron Reeves, Ph.D., and Clifford Nass, Ph.D., are recognized experts in social
psychology research from Stanford University and co-authors of "The Media
Equation," a seminal text on how people interact with computers and new media.

The personality developed by Reeves, Nass and General Magic is a key element of the
Portico service and is designed to make the user’s experience natural, easy and efficient.
Users will be able to interact with their virtual assistants in much the same way they
would interact with a human being. General Magic is expected to offer several different
personalities by the end of next year.

In addition to working with Reeves and Nass, General Magic also is collaborating with
Maz Kessler and Robby Kilgore, founders of FunArts, to develop, script and produce the
magicTalk interface with the human-like personality. Grammy-award winners Kessler and
Kilgore have 15 years of experience at the forefront of art and technology. In addition to
writing, producing and performing for major recording artists such as The Rolling Stones,
Paul McCartney and Madonna, Kessler and Kilgore have designed and produced agent
interfaces for leading technology companies, including Microsoft.

Working with Reeves and Nass as well as FunArts, General Magic has defined and
implemented a number of patent-pending characteristics necessary for a virtual assistant
to achieve a human-like personality. These characteristics include:

  • Personality type that is well-defined and evident in the assistant’s vocabulary, tone of
    voice and sentence construction. To achieve this, the virtual assistant’s remarks are
    professionally scripted, performed and recorded, but are not menu-driven.
  • Variety in the way a virtual assistant expresses or recognizes a thought. When speaking,
    the virtual assistant will randomly vary its speech responses ("sure" in
    addition to "yes") and use temporal words ("good morning" in addition
    to "hello"). When listening, the virtual assistant will recognize various ways
    of expressing similar ideas ("Read my mail," "Do I have messages?" and
    "Is there mail for me?").
  • Politeness while speaking by using "please" or "thank you," proper
    names and using a manner and tone that is pleasant to the user. The assistant also should
    recognize polite phrases ("Please read my mail," in addition to "Read my
    mail").
  • Helpfulness by delivering shorter prompts as the user spends more time on the system;
    help messages that provide more detail with increased difficulty and coaching upon
    request.
  • A sense of humor that is interjected, at times, into the conversation to keep the
    interaction fresh.
  • A selection of virtual assistant "types" that a subscriber can select or
    "hire." These candidates may differ in gender, personality type or both. By the
    end of 1999, Portico is expected to offer a selection of different personalities and both
    genders.

"The voice user interface needs to be confident, capable and intelligent to the
user," Reeves says. "People need to feel comfortable with the experience. If the
interface expresses personality, people will interact in a human-like manner–which is the
most natural interaction of all. This level of attention to the development of a
personality in a voice interface is long-awaited and a key to the user’s acceptance."

Working with Kessler and Kilgore of FunArts, General Magic took these desired traits
and produced a VUI with human-like personality. FunArts scripted statements, auditioned
actors and recorded the voice for magicTalk.

"The voice reveals certain things that can be hidden in other forms of
communication. For instance, gender, competence, intelligence, likability are all
instantly revealed in the voice of the character,” says Maz Kessler, co-founder of
FunArts. "In addition, for the magicTalk interface, the personality had to be dynamic
enough to interact in one-on-one conversations with users who are nonpredictable.”

magicTalk currently includes more than 5,000 responses and helpful hints such as
"I have your address book open. How may I help you?" or "Where would you
like your calls forwarded?" Each response was professionally scripted, rehearsed and
recorded to invoke personality traits identified by Reeves and Nass.

Ken Branson is business and finance editor for PHONE+ Magazine.


Leave a comment

Your email address will not be published. Required fields are marked *

The ID is: 67994