The visible problems of the invisible computer:
A skeptical look at information appliances
Andrew Odlyzko
AT&T Labs - Research
amo@research.att.com
Revised version, September 7, 1999
Abstract:
The future is said to belong to information appliances, specialized
and easy to use devices that will have the car tell the coffee pot to
brew a cup of coffee just in time for our arrival home. These gadgets
are supposed to eliminate the complexity and resulting frustrations of
the PC. The thesis of this essay is that while information appliances
will proliferate, they will not lessen the perception of an
exasperating electronic environment. The interaction of the coffee
pot, the car, the smart fridge, and the networked camera will create a
new layer of complexity. In the rush towards the digital era, we will
continue to live right on the edge of intolerable frustration.
The paradox of information appliances is that while they are presented
as products for a mature market, their main effect will be to unleash
a tidal wave of innovation. When technology changes rapidly, greater
ease of use serves to attract more users and developers, creating new
frustrations. The most we can do is ameliorate the spread of the
information appliance products and services. To do this, it appears
necessary to recognize that flexibility and ease of use are in an
unavoidable conflict, and that the optimal balance between those two
factors differs among users. Therefore systems should be designed to
have degrees of flexibility that can be customized for different
people. It will also be essential to provide for remote
administration of home computing and networking.
1. Introduction
The PC is dead. Technology gurus assure us that the PC is passe, and
we are entering a new era of computing, often referred to as
ubiquitous or pervasive computing. It is to be dominated by
information appliances, specialized and easy to use devices that avoid
the complexity of the PC. An enthusiastic cover story in Newsweek
heralded the wonders of the coming new age: "Your alarm clock might
ring later than usual if it logs on to find out that you don't have to
get the kids ready for school -- snow day!" [Levy]. The most
prominent proponent of the post-PC movement is Don Norman, whose
influential book, "The Invisible Computer," presents detailed
criticisms of the PC and a vision of an information appliance future.
Even Bill Gates accepts most of the gospel of the invisible computer.
He argues that the PC will continue to play a central role, but that
it "will also work in tandem with other cool devices," and that we
will be able to share our data across different machines in a seamless
fashion [Gates].
Are we entering a new era, and is it going to fulfill all the
extravagant promises that are made for it? There are reasons to
temper our enthusiasm and be skeptical. We can distinguish three main
predictions by the advocates of the invisible computer:
(a) There will be a proliferation of information appliances.
(b) Information appliances will eliminate the frustrations of the PC.
(c) The dominance of computing by the PC and Microsoft will end.
My prediction is "yes" on (a), a decided "no" on (b), and a "maybe" on
(c). Information appliances will be popular, since they will provide
many novel services that the PC is ill-suited for, and will do so in
user-friendly ways. However, they will introduce their own
complexity, and the level of frustration with technology will not
decrease. This is a result of the conflict between usability and
flexibility. The human-centered engineering advocated by Norman is
feasible, but only when technologies and markets are mature. The
information appliance market will be anything but mature for a long
time to come. The emphasis in information processing has been, and is
likely to continue to be, on development of novel applications.
Further, the PC, in spite of its shortcomings, may indeed manage to
play a key role in the new era. Arguments supporting these
predictions are presented in later sections.
If the level of frustration is not going to decrease, is there any
point in developing new technologies, and in paying any attention to
ease of use? There certainly is. We will still be frustrated, but at
a higher level of functionality, and there will be more of us willing
to be frustrated. Just consider the Mosaic browser. It lowered the
complexity of accessing the World Wide Web below some magic threshold
and drew millions of people onto the Internet. These millions of new
users then created new content, which drew in millions of additional
users. That is how the Internet graduated from a research toy to a
revolutionary tool that is transforming mainstream society. However,
we now get frustrated by unreliable servers, network congestion, and
stale URLs, problems that we did not worry about just a brief
half-dozen years ago.
The main point of this essay is not to debunk information appliances,
but to temper the overenthusiastic promises that are being made for
them. In Section 7 some suggestions are also made for a smoother
introduction of information appliances. It appears essential to
develop systems that allow for setting the flexibility at different
levels for different users. It will also be necessary to provide for
remote administration of home networks by experts, leading to the rise
of a new outsourcing business.
2. Information appliances will proliferate
Information appliances are not an overhyped fad such as interactive
TV, push technology, or "buddy lists." There is substance to their
promises, and the Palm Pilot is just one early example of the devoted
following they can acquire. As another example, in "Finland, people
are using their Nokias to pay bills, access bus schedules on the
mobile-phone display and punch in payment codes for car washes or
juke-box tunes" [Levy]. We will want all these services, as well as a
variety of others that have not been thought of yet.
Fuller and more eloquent expositions of what information appliances
will bring us can be found in [Levy, Norman]. I will mention briefly
just two key points. One is that the information appliance is the
natural outcome in the evolution of information processing. That is
why they were foreseen a long time ago, with the late Mark Weiser the
most influential early pioneer. Digital computers started out as
expensive mainframes accessible to a few. The next step was the PC
that individuals could own. Yet even the PC was for a long time an
expensive instrument, and there was seldom more than one per house.
Thus it was essential to have as much functionality in the PC as
possible. Technology is making feasible small and inexpensive devices
that are smart. This helps push the intelligence closer to the
people, the ultimate customers.
The second point is that there is already a proliferation of primitive
information appliances. The average middle-class American household
already has around 40 microprocessors, in cell phones, microwave
ovens, self-focusing cameras, and the like. Furthermore, many of
these microprocessors are astonishingly powerful. For example, the
latest game consoles have more processing power than the
supercomputers of a decade ago. However, this power is hidden from
the users, who see only a simple interface designed to provide just
the basic functionality the device is designed for. Thus the
"invisible computers" are already with us in large numbers. What is
still lacking is the pervasive communication system that will link
them together.
3. The visible problems of the invisible computer
The central paradox of information appliances is that they are aimed
for a mature market with a mature technology, but their wide spread
will ignite an explosion of innovation that will destroy any stability
that might exist. Don Norman argues in his book "The Invisible
Computer" that the PC was aimed at the "early adopters" (in the
terminology popularized by Geoffrey Moore [Moore], see also [Rogers]).
The next step is to design information appliances for the mass market.
He advocates a "user-centered, human-centered, humane technology of
appliances where the technology of the computer disappears behind the
scenes into task-specific devices that maintain all the power without
the difficulties" (p. viii of the Preface in [Norman]).
Norman's vision is certainly an appealing one. His book cites the
instructive story of radio. It started out as a complicated device
that required much practice from users to obtain even a noisy signal.
The user instructions for an early radio reproduced in [Norman]
illustrate beautifully how far we have come. Whereas the first radio
users had to have the patience of Job, today we can select any radio
from among hundreds of models, take it home, plug it into the electric
outlet, push some buttons, and listen to our favorite music station.
There is great technology inside the radio (technology that keeps
improving from one generation to another), but we do not have to know
anything about it.
Don Norman would like computers to evolve the way radio receivers did.
The problem is that with radio, we know well what we want, since the
basic services we desire (such as music, talk shows, and news reports)
are well understood and stable. That is simply not what we will see
with information appliances, not for a long time. We cannot know how
people will want to use information appliances. Note that even the
Palm Pilot, beloved of millions of users, and frequently cited as the
ideal outcome of the human-centered engineering advocated by Norman,
is not stable. Not only is there a succession of new models from its
manufacturer, but there are myriads of accessories offered by outside
suppliers for wireless communication, control of other devices, and so
on.
Even the user-friendly radio that makes such an effective case for
Norman's proposals is not likely to remain stable and user-friendly.
It is likely to be swept up in the whirlwind of change that
information appliances will unleash, since we will want our radios to
communicate with our other gadgets.
Thus even from a high level systems view, there are reasons to be
skeptical about the ability of information appliances to fulfill all
their promises. Next we look at what specifically is likely to go
wrong.
4. The inevitable frustrations of information appliances
Careful design that is focused on human factors, and incorporates
powerful processors and software, can provide information appliances
that are a delight to use. The Palm Pilot and game consoles prove
this. However, that does not mean that we will be delighted with the
new electronic environment full of such gadgets, even if (and this is
a big if) each is excellent by itself. Information appliances are not
supposed to be standalone devices. In Don Norman's definition (p. 53
of [Norman]), "[a] distinguished feature of information appliances is
the ability to share information among themselves." Information
appliances are meant to be "cooperating devices," a felicitous term
coined by Bob Frankston. We will want our car to tell our house
control system to warm up the family room in time for our arrival, and
"the refrigerator [to] know it was low on milk and eggs and place an
order with the local supermarket" [Lewis]. Once all the radios,
refrigerators, dishwashers, clocks, coffee pots, and other devices in
our houses are replaced by new models that are information appliances,
the current 40 isolated microprocessors per household may grow to
perhaps 400 communicating devices. Will they all interoperate
smoothly? They certainly do not do so now. Consider just the
difficulty of setting up home networks, even for simple connections of
PCs [Lewis]. Similar problems arise in setting up cable modem and
ADSL connections. Once the number of devices to be connected
increases, and wireless communication expands, the difficulties will
increase. No single problem will be insurmountable. However, the
range of problems to be solved will be growing rapidly with increasing
complexity of the system.
Don Norman recognizes the difficulty this poses (see Chapter 3 of
[Norman]) but forecasts that a solution can be achieved through
"world-wide agreement on the appropriate infrastructure that will
allow appliances to share their information with appropriate other
appliances." Bill Gates promises to fulfill that vision, so that
"when you buy a new device, you'll know it will function with your
existing equipment" [Gates]. Yet will Microsoft deliver, given that
it now creates software that does not allow for easy transfer of
information from one Microsoft software package on a PC to another
copy of the same package on a different PC [Alsop]?
It helps to compare information appliances to programs on a PC. Each
application might be delightful to use, but it is the interaction of
these applications with each other, and with the operating system,
that creates most of the complexity and frustration (cf. [Alsop]).
The PC is used widely in spite of its shortcomings because most people
rely on just a few applications, and in an application, they usually
depend on only a small subset of its features. They thus learn to
live with the complexities of the PC by avoiding them. However, those
complexities are there. Einstein said that "everything should be made
as simple as possible, but no simpler." Unfortunately we are asking
our computers, whether standard PCs or the information appliances of
the future, to do complex things. Even if a spreadsheet and a word
processor work fine, asking for the ability to bring in a graph from
the spreadsheet into the word processor creates a new level of
complexity, with more opportunities for bugs.
In the information appliance environment, complexities similar to
those of the PCs will also be present, and in many ways will be
magnified. After all, on a PC everything is in a single box, and the
standard procedure for dealing with problems is to reboot the PC. Will
we have to go around the house rebooting the potentially hundreds of
information appliances that we might own? Even if we could do it, it
might not solve the problem if the difficulty is in interaction with
our neighbor's system, or that of our in-laws on the other side of the
continent. A small taste of the problems that are likely to plague us
is given in [Levy]:
... Bill Joy [of Sun Microsystems, a vocal critic of the PC and an
advocate of information appliances] offers to print out a paper that
illustrates a salient point. He reaches for his laptop, which
is equipped with the sort of wireless high-speed Internet connection
that, one day, may be a routine adornment in all our cameras,
palmtops, game machines, medical sensors and, yes, dishwashers.
According to the theory, these will all be linked together, ot
course, in an infrastructure that will virtually eliminate crashes
and glitches. He keyboards the command to print the document in
the adjoining room. And nothing happens. "You know what?" he
finally says. "I think this did get printed--on the printer back
in my house across town."
The proponents of information appliances promise that technologies
such as Bluetooth and Jini will solve the problem. Yet one should be
skeptical of whether these promises will be realized. The problem is
not necessarily that the technologies are inadequate to achieve the
promised goals. Rather, it seems likely that, just as in the past,
the computing and communications industry will not concentrate on
those goals. Consider again the PC. Graphical user interfaces,
object-oriented programming, and Java are just three of the
technologies that were supposed to revolutionize computing and make
life simpler. Remarkably, these three did succeed (even if not to the
full extent promised by their early proponents), and our computing
would be much more primitive without them. Still, their main effect
has been to create more complicated systems, not to simplify old ones.
Building complicated systems that work is hard. Building ones that
work and are user-friendly is much harder. Further, it is necessary
to balance the demand for user-friendliness with the demand for more
features. Although most users complain that they want simpler
versions of applications such as Microsoft Office, their "responses
support Microsoft's contention that while few people use more than a
tiny percentage of the programs' features, everyone wants a different
10%" [Wildstrom]. The history of the past two decades shows that when
the choice was between new features and ease of use, new features have
won. The victory of the PC over the Mac is just one example of this.
As Edward Tenner [Tenner] pointed out,
Microsoft has triumphed because it has given us what
we asked for: constant novelty coupled with acceptable
stability, rather than the other way around. ...
People talk simplicity but buy features and pay the
consequences. Complex features multiply hidden costs
and erode both efficiency and simplicity.
In the evolution towards the information appliance era, we can expect
similar outcomes, not because they are preordained by technology or
dictated by Microsoft, but because that is what people are willing to
pay for. The premium will continue to be on being first to market
with the latest innovation, not on ease of use.
5. The dominance of the PC and of Microsoft
Centralized Web servers are already usurping much of the PC's role.
Will information appliances deliver the final blow, and lead to the
Post-PC era in which the PC is marginalized, as Microsoft's
competitors predict? Or will they lead to the PC-plus era, in which
the PC plays a central role, as Microsoft hopes [Gates]? It appears
impossible to predict because of uncertainties in both technology and
industrial politics.
The complexity of managing the interaction of all the invisible
computers could be tamed most easily, at least initially, by using a
powerful central processor, a role that the PC can naturally aspire to
play. That would also simplify the integration of existing PC
software with the new information appliances. On the other hand, the
strength of the PC in legacy applications is also a weakness, in that
the PC is not well suited for the new distributed environment. This
creates an opening for potential rivals such as the Aperios and Epoc
operating systems.
If Microsoft concentrated exclusively on the PC, one could easily
foresee a future in which information appliances would play the role
of a disruptive technology [Christensen]. They would develop in the
shadow of the dominant PC, serving new markets, until those markets
would dwarf the basic PC industry. At that stage information
appliances would relegate the PC to a secondary role, just as the PC
did to the mainframe. However, Microsoft is vigorously pursuing the
information appliance market [Gates, Lewis], and may become a rare
case of an established player that is nimble enough to change
directions. Whether it will preserve its dominant role depends not
just on technology, but also on political alliances. There are many
other powerful players that are reluctant to concede the leadership
role to Microsoft. Thus the eventual outcome will be less a matter of
technology than of politics, and is much harder to predict.
The emerging competition between Microsoft and its rivals for
dominance of the information appliance market is a bad omen for the
ease of use that we are promised. Just as with PCs, victory will go
to the camp that gains the allegiance of developers, who will be
creating all the enticing new devices and services that will attract
customers. Hence the premium will be on making the developers' task
easy, not on users' convenience. That was a major factor behind the
evolution of the frustrating PC [Odlyzko].
6. The unavoidable tradeoff between flexibility and ease of use
The PC is extremely flexible. In Bill Gates' words [Gates],
[s]itting at your PC, you can do your taxes, surf the Web,
write letters, e-mail friends, play games, plan a business,
buy a car, do your homework ... in fact, do whatever you
want.
A network manager found 350 different software packages on the 1,000
PCs in his company [Jaffe]. Even more remarkable than the variety of
applications that run on a PC is that the PC was not designed with
them in mind. The basic architecture of today's PC is not much
different from that of the early Atari machines marketed to hobbyists.
Yet spreadsheets and desktop publishing, the two "killer apps" that
propelled the PC to its current dominant status, could run on it.
More recently, the rise of the Internet in public consciousness can be
dated to the mass distribution of the Mosaic browser. It penetrated
as widely and as rapidly as it did because it could be easily
installed on millions of PCs that had been acquired for other
purposes. The idea of a universal information processing engine,
which is what the PC embodies, is extremely powerful.
Unfortunately, as Don Norman says (p. 181 of [Norman])
Computers are general-purpose devices, designed to do everything.
As a result, they can't be optimized for any individual task.
That is one difficulty with the PC. Another, related, problem is that
in the design of the PC, many choices were made intentionally to make
it as flexible and as user friendly as possible. (Yes, paradoxically,
it was the desire for ease of use that led to many of the problems the
PC is derided for.) Users have complete control over their machines,
and can even modify the operating system at will, just by clicking on
an email attachment. This model makes any real security impossible.
Further, it makes it hard even for experienced computer experts to fix
problems (cf. [Alsop]). Thus long-range ease of use has been given
up in favor of short-term convenience, in enabling users to modify
their machines on the spur of the moment. This is great if you care
about rapid diffusion of the next Mosaic, but it leads to frustration
when things go wrong, as they often do. To provide stability,
security, or transparency requires limiting users' flexibility.
A tradeoff between flexibility and ease of use is unavoidable.
However, there is no single tradeoff that is optimal for everyone.
Don Norman argues that the PC was aimed at the "early adopters," and
that its lack of success in penetrating about half of the households
in the U.S. is a sign of its poor design. Popular perception of the
PC is certainly one of "infuriating complexity that makes us want to
toss our beloved PCs out the window about, oh, once an hour" [Levy].
The success of the iMac is another sign that consumers do value
simplicity. Norman argues that information appliances can and should
be designed for the mass market. Proper design of simple interfaces,
appropriate when a restricted set of tasks is to be enabled, does make
this possible.
The problem, as was explained earlier, is that we should not be
thinking just of individual information appliances. Those can be made
to appear simple through careful design, and in particular by limiting
their functionality. We have to be concerned with the whole system,
which is likely to be complex. Further, there is no single tradeoff
of flexibility versus ease of use that is optimal for everyone. There
is not even a single tradeoff that is likely to be optimal for any
individual for long. A person learning a new system can usually
handle progressively more features. Thus we cannot hope to design
information appliances to a single standard. Norman cites the example
of the evolution of radio receivers as models of how computers should
change. However, there is a substantial difference between radios and
computers. We need a much greater variety of computers than of
radios. Further, in the networked environment, the full range of
information appliances with varying capabilities will have to
interoperate.
To appreciate the wide range of computing that we have, and are likely
to have in the future, consider open source software. It is often
touted as a proof that a viable competitor to Microsoft's Windows can
arise. Yet it seems that that the main lesson to be drawn from the
success of Linux and Apache is different. These systems are built by
experts to be used by experts. There are many people (although a tiny
fraction of the whole population) who know what regular expressions
are, and can use text commands to execute programs much faster than a
graphical user interface would let them. They also tend to be in
charge of important resources such as Web servers, and appreciate (and
use effectively) the flexibility that access to source code provides.
Apache and Linux are ideal for them. They are not satisfied with the
black-box software from commercial vendors. On the other hand, it is
doubtful whether those among them who contribute to the code, as
opposed to just using it, have the interest in creating the easy to
use but much less flexible interface that would appeal to a wider
market. That is the province of Microsoft, Apple, and other software
companies. (There might be a business opportunity for companies to
put simple interfaces on top of Linux for the mass market, though.)
These expert users do not account for a large fraction of desktop
computers, but do control a large share of computing budgets. They
form a substantial market for computers where flexibility is dominant,
even at the cost of ease of use.
At the other extreme, about half of the households in the U.S. still
do not have any computer, and often this is because of the perceived
difficulty of using current PCs. Further, there are millions of VCRs
whose clocks flash 12:00. The owners of these VCRs are ignoring the
ability to program video-taping on their devices. This is the
standard response of consumers to features that do not provide enough
value compared to the hassle of using them. What it means is that
information appliances will have to be extremely attractive and easy
to use to gain wide acceptance. Further, the full range of users,
from the computer experts using open source systems, to the totally
non-technical folks, will have to operate within the same
communications infrastructure.
7. Customizable flexibility and computing and communications
outsourcing
Flexibility does conflict with ease of use and the optimal balance
varies widely among users. Further, flexibility is valued even when
it is not used. There are millions of VCRs with 12:00 flashing in
their clock display because their owners use them exclusively for
playing prerecorded tapes, and have not felt the need to set them up
for programmed recording. However, play-only units, although less
expensive, have had disappointing sales. Being able to record at a
moment's notice has significant value. Similarly, there is value in
being able to install the next Mosaic on an existing device without
hardware modifications.
The problem is how to balance flexibility and ease of use in a way
that can be customized for people with different needs. Furthermore,
the right balance is likely to vary for different people in the same
household. It seems that the only way to solve this problem is
through the logical evolution of the approach that is already followed
in corporations as well as universities. Almost all such institutions
have groups of experts that provide computing and networking
assistance. These groups often specify what types of equipment and
software they will support. Exceptions can be made for specialized
needs, but then users are often told that they have to be responsible
for the operation of the special systems. Most users live within the
limits imposed by the support group.
The home information appliance environment is likely to be more
complicated than the office environment today. Also, many users will
be less knowledgeable about electronics than the typical office
worker. Therefore it will be essential to outsource the setup and
maintenance of home computing and electronics to experts. It will not
be economically feasible for them to visit in person every time
something goes wrong, or a new device is to be added to the system.
Therefore all devices will have to be designed for remote
administration. (Most of it will be automated, and it will be
facilitated by, and may essentially require, broadband access to the
home.) Perhaps even more important, all these new information
appliances will have to be designed for customizable flexibility, so
that only the administrators will have full control of them. Users
will be given varying degrees of control, depending on their skills
and trustworthiness. The operating system will need to be rigidly
isolated from the applications, and the applications will have to be
tested for compatibility by the administrators before they are
installed. This will reduce users' freedom to modify their systems.
However, it should bring in some sanity to the potentially chaotic
scene and make possible deep penetration of information appliances
into society. If Aunt Millie wants to give a new toy to your son Bill
for Christmas, she may first have to check with your system manager
whether that toy will interoperate with all the other information
appliances in the house. Most users are likely to accept such
restrictions to simplify their lives.
8. Conclusions
We were frustrated with computers a decade ago, we are frustrated with
them now, and will continue to be frustrated in the future. As long
as technology offers enticing new products and services, we will
continue to live on the edge of intolerable frustration. However, by
providing for customizable flexibility and developing outsourcing
services for computing and networking support, we can smooth the
transition to the information appliance era of computing.
Acknowledgements: I thank Rob Calderbank, Dan Duchamp, Bob Frankston,
Nick Frigo, Paul Henry, Bob Kurshan, Hilarie Orman, David P. Reed,
Quentin Stafford-Fraser, Hal Varian, and Bill Woodcock for their
comments on an earlier version of this essay.
References:
[Alsop] S. Alsop, Have I told you lately that I hate Windows?,
Fortune, July 5, 1999, pp. 177-178. Available at
.
[Christensen] C. M. Christensen, "The Innovator's Dilemma: When New
Technologies Cause Great Firms to Fail," Harvard Business School
Press, 1997.
[Gates] B. Gates, Why the PC will not die, Newsweek, May 31, 1999,
p. 64. Available at
.
[Jaffe] B. Jaffe, So much software, so little time, PC Week, May 17,
1999, p. 84. Available at
.
[Lewis] P. H. Lewis, The once and future network: Wiring the home has
come a long way but has a way to go, pp. G1 and G7, and State of the
art: Carrying Windows around, in your pocket, pp. G1 and G3 in New
York Times, July 15, 1999. Available at
and at
[Levy] S. Levy with J. Sandberg, J. Tanaka, and B. Stone, The new
digital galaxy, Newsweek, May 31, 1999, pp. 57-62. Available at
.
[Moore] G. A. Moore, "Crossing the Chasm: Marketing and Selling
Mainstream Customers," HarperCollins, 1991.
[Norman] D. A. Norman, "The Invisible Computer: Why Good Products Can
Fail, the Personal Computer is so Complex, and Information Appliances
are the Solution," MIT Press, 1998.
[Odlyzko] A. M. Odlyzko, Smart and stupid networks: Why the Internet
is like Microsoft, ACM netWorker, Dec. 1998, pp. 38-46. Available at
.
[Rogers] E. M. Rogers, "Diffusion of Innovations," 3rd ed., The Free
Press, 1983. (First edition, 1962.)
[Tenner] E. Tenner, The Microsoft we deserve, New York Times, Op-Ed
page, March 8, 1998.
[Wildstrom] S. H. Wildstrom, Technology & You column in Business Week:
Office lite: Less would be more, p. 28 of June 14, 1999 issue, and
Give us a simpler "Office," p. 20 of July 5, 1999 issue, both
available at .