Upload
heinz
View
212
Download
0
Embed Size (px)
Citation preview
Abstract
ON THE PSYCHOLOGY Of PROTOTYPING
Anker Helms Jörgensen
Oatacentralen
Retortvej 8
OK - 2500 Copenhagen Valby Oenmark
This paper relates the concepts underlying prototyping, evaluation and iteration, to properties of human cognition. Prototyping lends itself to humans' acquisition of real-life experience as opposed to dealing with FORHALIZED descriptions. Based hereon
the paper first argues that prototyping in the short term will weed inappropriate gesigns in that users' say has more weight when based on prototypes than on written
specifications. Secondly, as long as user interface design is suffering from lack of
appropriate theories and methods, designs will be based on individual designer's experience and intuition. These must therefore be cultivated in trying to achieve more usable interactive systems. Prototypes provide an efficient means to this end in that
the designers are confronted with the consequences of their own decisions.
Introduct;on
Prototyping! Vet another catch-phrase flooding the dp-world1 This paper advocates that the answer is No. Along with the propagation of interactive computer systems the
need for usable systems has become even greater. This paper views prototypes as one of the potential means for achieving this end.
In other engineering fields prototypes in different forms have been applied success
fully for a long time. They are indispensable vehicles for obtaining irreplaceable
HANDS-ON experience with a system, be it a chair or a ship. In contrast the data
processing world has always suffered from an extremely low valuing of empirical evidence. We think we know enough during design and don't really care about how the
systems interact afterwards with their audience of users. We're satisfied as long as
the internal architecture of the systems is neatly structured and consistent.
This attitude has prevafled even though unusable systems has been the order of the day for several decades. I'll denote this approach to systems engineering as the TRADITIONAL approach. It's based on written specifications FROZEN at an early stage
and doesn't fnclude feedback from one stage back to prev;ous ones. A basic premise
for this approach is that the future application of the system and fts environment is
R. Budde et al. (eds.), Approaches to Prototyping© Springer-Verlag Berlin Heidelberg 1984
279
viewed as predictable. Contrasting this view, the concept of prototyping is founded
in another attitude to systems engineering. The underlying view is that systems oper
ate in so complicated and dynamic contexts that it's impossible to predict their
operation satisfactorily. The concept is therefore based on EVALUATION of decisions
and feeding the results and experience into previous development stages, i.e.,
ITERATING these stages.
Research evidence of design processes is slowly emerging. A recent impirical study of
design processes [MALHOTRA] provided a model of design processes where a designer
helps a client solve a design problem. The model has a st rong cyclic structure and
thus provides support to the iterative approach. Moreover, the body of evidence ap
pearing in the literature concerning the power of this approach applied to practical
systems development is growing steadily. In the development of Xerox Star and Apple
Lisa work stations, two of the new breed of EASY-TO-USE systems, iteration was one of
the fundamental principles of the development process [SMITH, MORGAN]. As another ex
ample, [CLARK] reports on a design experience using prototypes for usable product
design. In the design of HELP display frames for a business graphics system for
non-dp users, empirical evaluations were made early on by testing potential users'
understanding of the HELP frames. This testing yielded potential benefits {or the
overall usability of the system. This is particularly critical for systems intended
for non-dp users and for systems where the training is based on self-study tutorial
manuals.
It is important to stress that there are no inherent differences between the TRADI
TIONAL approach and the iterative approach. In principle a TRADITIONAL system could
be evaluated and subsequently redesigned. That however, hardly ever happens - for
many reasons. The cost of substantial modifications is considerable and users don't
like to have their habits pushed around too of ten. Thus the difference between the
two approaches is rather based on practical circumstances. Going to the limit one
could say that these practical constraints have been used as an excuse by system
designers for not having to look critically at their products. Iteration implies dis
carding one's own work!
The Inherent Nature of Interact;ve Systems
The basic view expressed in this paper is that describing, designing and using inter
active systems are not easy tasks. This section looks at some inherent characteris
tics of interactive systems that cause these difficulties.
The traditional approach to systems development has its origin in batch systems where
communication between users and systems is based on forms, i.e., written words on
paper. In that context written specifications easily capture the major communication
aspects.
280
Not so in interactive systems. These have a LIVE dialogue component as well that
written words on paper are completely unable to capture. The saying "An image is
worth more than a thousand words" is well known. It has an analogy here: "A hands-on
session with an interactive system is worth more than a thousand images". This is
illustrated in Figure 1 showing two screen displays from a paper by [TEITELMAN]. He
presents a highly interactive system by guiding the reader through a detailed task
session illustrated by 42 figures, each accompanied by a verbal description of the
task step at hand. But even af ter meticulous reading one is left with a meagre im
pression of the facilities and appearance of the system. A short HANDS-ON session
would be far more beneficial.
Fig.1: Two screen displays (taken from [TEITELMAN]).
Describing interactive systems is difficult. That's one of the reasons why designing
interactive systems also is a difficult task. Other reasons are lack of theory,
methods and tools. Hence design is largely based on individual designer's intuition
and experience. Even if designers do their best they may inadvertently overlook im
portant issues in task context or user predispositions. Prototyping provides a means
for bringing such issues closer to the development process.
Computer-based systems are extremely powerful compared to manually based systems in
terms of speed and storage capacity. But they appear completely different to their
users. In manually based systems both operations and objects are more or less
directly comprehensible, even to an outsider. Typical operations are filing of a
sheet of paper, exchange of currency notes and phoning, while objects can be piles of
reports, filing cabinets and security officers.
281
Not so in computer-based systems. The operations and the objects are hidden to the user. Virtually everything takes place behind a screen, perhaps even in some remote
computer. The communication channel bet ween the system and the user is very narrow. It is typically restricted to a small vocabulary, of ten with odd syntax rules and incomprehensible messages. Moreover the communication takes place on the terms of the system. The user is not allowed to introduce a new style of communication. In some systems the user is allowed to choose between menu dialogue and command dialogue,
perhaps even choose bet ween verbose and terse messages!
An example of the width of the communication channel can be drawn from microcomputers with floppy discs. Of ten the progress of the system can be detected by the CLICS of the disc drives. Lo and behold, one can suddenly HEAR the system. For experienced users this information is of ten far more useful than cryptic error messages. In
stantly one reacts to a wrong MELODY: "Oh yes, I forgot to switch on the printer!".
In addition to the hidden objects and operations the computer-based systems also impose great demands upon the cognitive skills of the user, ranging from very concrete
to very abstract. A single character, for example caused by an inadvertent hit of the
space bar, may cause complete failure of an exchange of information. Objects become
abstract and appear in different manifestations at different times. Take a memo in a document processor as an example. Throughout it exists as electronically recorded information, perhaps even in two versions: a WORKING version and a DISC version. The memo sometimes manifests itself on the screen, sometimes on a,sheet of paper. In some systems the memo appears on the screen af ter it has been filed! Lo and behold, it may even disappear from the FILING CABINET - if it hasn't been PERMANENTED. Moreover, the memo can be f;led MANUALLY when residing on a floppy disco It can also be mailed
electronically or manually as a floppy disco It is hardly surprising that non-dp users have difficulty in establishing a CORRECT mental model of the system.
Finally, interactive systems - as do other systems - handle formalized information. Comprehension of the underlying structures of the data or procedures is of ten crucial
for efficient use of the system. In manually based systems the range of sources of information that can facilitate understanding of the formalization principle is far
wider than in computer systems. For example, in exploring urban bus services one
knows that there are more services in the rush hours, that industrial areas aren't serviced during weekends, that probably fewer services exist if parallel transpor
tation services are available, etc. In other words a range of COMMON SENSE sources
can be drawn upon.
In contrast, in computer systems the formalization is far more extreme (e.g., in use of codes) and in addition there are less sources of information to seek recourse tOe The sources available, e.g., a manual or a dp-professional, are of ten equally dif
ficult to grasp as the computer system itself.
282
As an example, a secretary couldn't make a text processing system print in two
columns. The text was printed in one column in the standard margin setting. The
reason was that the margin-setting command was overrfdden by a pitch-setting command issued later! Not only is this contrary to all common sense concerning manuscript preparation, but also the manual didn't state this peculiarity in a manner comprehen
sible to the secretary.
In summary, interactive systems possess some inherent characteristics that the traditional approach to systems design fails to take into account appropriately.
Human Cognitjon
Cognition is the psychological term for humans' thought processes. These include learning, planning, language processing, higher levels of perception, problem solving, memory, etc. This section discusses some fundamental aspects of human cognition pertinent to the two approaches to systems development.
The first aspect is experiencing as opposed to being taught factual knowledge and
skflls. By far the most of the totality of humans' skills and knowledge has been
acquired byexperience. Think of a child playing with toy bricks. The child learns about spatial relations (the bricks can be put on top of each other>, shapes (one
brick is oblong, another cubic), gravity (when the pile of bricks fall over), weight (two bricks of the same shape made of different materials weigh differently), etc. We are also taught about these issues. Imagine however, how little sense it would make to learn about gravity in physics if one hadn't had the experience of throwing balls?
The second aspect is the ability of humans to perceive and operate at different levels of abstraction. We conceptualize all knowledge and experience into levels of abstract ion. This facilitates understanding of general relationships between concrete
matters.
However, we also work from higher levels towards more concrete levels. This facil
itates specific in-depth understanding of general relationships. This point is utilized in using examples in teaching. In school we are taught the skill of subtrac
tion. Could you imagine learning subtraction without using examples?
The third aspect is the familiarity with those fields of life we have experienced as
opposed to those we're only knowledgeable about in terms of descriptions. The first example illustrates the gap between what can be experienced but not described fulLy. The skill of bicycle riding is a matter of experience. It takes a lot of trying but sUddenly you've got the knack. But how? You don't know. Even experienced cyclists are
unable to give a full account of the skill of bicycle riding. It is beyond our in
tellectual understanding, and thus description.
283
The second example illustrates knowledge that can be described perfectly well but
hardly ever is. Anyone living in a house with astaircase will KNOW the number of
stairs. That is, if asked, an occupant will most likely say"I don't know!" but
nevertheless inevitably stumble if an extra stair is added on top of the staircase.
Humans operate perfectly well without having to make knowledge explicit even in areas
where it can be expressed in formal terms.
Another similar example is the recognition of objects without being able to tell the
name of the object. Any mushroom enthusiast is familiar with the excitement of
finding a beautiful edible mushroom. You know exactly what species it is, but its
name has escaped your memory. lts got the right characteristics, you know that there
are no similar poisonous LOOK-A-LIKES, you know the best way of preparing it and even
recall its taste and crisp texture on the tongue. But sharing the excitement with
another mushroom enthusiast is hampered incredibly if you haven't yet managed to
recall its name.
The next example goes the other way, from written specifications to real-life. It
shows how little of reality the written specifications capture. ear brochures come
with captivating photos and detailed specifications of speed, acceleration, type of
suspension, etc. But this information provides but a meagre impression of the real
car. It can be put this way: Who would buy a new car without having driven it? Per
haps some. But hardly anyone would buy it without having seen it!
The context of a description is very important. You may re ad some material but
nevertheless lack the feel for the information because the author has failed to
establish the right frame of reference. The information remains as fragments rat her
than as a coherent body. Introductory textbooks in psychology of ten use a description
of the process of washing clothes as an example. The clothes are first taken out of
the container, sorted into piles according to the type of fabric, put into the
washing machine, etc. The description however, is kept in general terms: The material
is' first taken out from where it resides, sorted into groups according to type of
material, etc. The point is that you do UNDERSTAND the general description but not in
the right context. The description makes little sense and is therefore difficult to
remember. In the textbooks the context of washing clothes is revealed af ter the gen
eral story, and you utter "Oh yes, of course, it's washing", One little hint is
enough, and you come to grips with the whole story.
The last two examples show humans' reactions when given the choice between a FORMAL
and INFORMAL description. Suppose you as a tourist ask a citizen in the street for
the way to the station. You use a map of the city serving as your frame of reference.
In most cases the citizen will point down the street and inform you as follow: "Turn
left at the light down there, continue until you see an old church on your right hand
side "- even if you try to make the citizen also use the map as a reference.
Of ten the citizen will think hard in trying to envisage the surroundings of the old
2M
church. The point is that information based on experience is drawn upon rather than
the formal representation on the map. You however, have no such experience and is
referred to formal aids.
The final example deals with the priorities that are given by humans to FORMAL versus
INFORMAL evidence. Very of ten the latter is preferred at the expense of the former.
Think for example of planting black currant bushes. These come in tiny exemplars from
the nursery and everybody knows that they grow to a considerable size. Perhaps the
market gardener even provides instruction as to spacing them several meters apart.
Nevertheless this factual knowledge is all too of ten disregarded in favour of im
mediate impressions. The bushes are spaced at too short intervals causing them to
suffer from lack of light and space when they grow bigger - because "the bushes look
so tiny".
In summary, examples abound of the shortcomings of written descriptions of real-world
contexts. So many aspects can't be captured by this medium. This is not to say that
descriptions are useless, only to pinpoint some of their limitations.
Advantages of Prototypjng
The previous sections have prepared the grounds for a discussion of the advantages of
the prototype approach to systems development. This is presented here. The advantages
fall into two categories: short-term and long-term advantages. The short-term
advantages are those of actual systems development, i.e., the provision of a means to
obtaining a more usable system. The long-term advantages is the potentialof the
prototype concept to influence the system designers' view of systems development and
application. The short-term advantages are discussed first.
Systems development is also a LEARNING process. One of the most striking dis
advantages of the TRAOITIONAL approach to systems development is its failure to in
clude the learning taking place during the development process.
No matter how experienced and skilled the designers are they inevitable learn during
the development. Specificly they learn the fundamental concepts, the dynamics and the
social context of the task environment. This learning - which takes place throughout
the development - is witnessed by the phrase of ten uttered upon complet ion of a task:
"Only now I know how I should have done it"
Presumably every human being recognizes this from one daily-life context or another.
Why should systems development differ in that regard?
285
The users invol ·d learn about their own work. ~. ing to state what you're doing
of ten causes new lnsights and perhaps even a recogni on of other ways of doing
things. The users also learn about data processing, both specificly in formalizing
their own task area and generally in taking part of a systems development effort. All
this experience gained during the development process is not included adequately in
traditional systems design, partly because it relies on a FROZEN specification early
on in the process.
The prototype approach,
learning in that it
on the other hand, provides a vehicle
lends itself to its iterative nature.
for including such
Also the users are
presented with the system or parts thereof in a form and context much closer to the
final application context. This provides a far more solid ground for perceiving the
system and thus establishing a view of the system as a counterpart to the views of
the system designers.
In other words, facilitation of COMMUNICATION, is also a key concept in the prototype
approach. This goes both ways, from designers who have a means for presenting their
ideas and from users who se grounds for establishing an opinion are substantially in
creased.
But even within the TRADITIONAL dp-context itself the prototypes offer advantages,
e.g., in reviewing. In systems design the appearance of an implemented system of ten
causes surprise on the part of its designers. Hence the totality of design aids is
increased. These points are illustrated by quotes from interviews with designers who
had designed the user interfaces of interactive systems intended for non-dp users
[JORGENSEN1.
In the design of a document processor the designers had trouble in designing a so
cal led scale-line. (This is a line on the screen visualizing the margin, tabulators,
etc.):
"We worked like crazy to get the scale line visualization done on paper and we put
out at least half a dozen versions that were totally unsatisfactorily to the
reviewers. One day by putting the symbols up on the screen we got everybody to
agree."
The designer himself expresses the cause:
"The difference was the ability to be able to see the results as to trying to con
ceptualise the result was just like night and day. 50 obviously a working model, a
means by which you can empirically play with your concepts is useful in design."
This leads us to the long-term benefit of prototypes. In using prototypes the
designers are confronted with the consequences of their decisions on the terms of the
users, be it the visualization of a detail in a document processor or be it watching
the users' uncertainty about the phrasing of the menu options. This is of paramount
286
importance· in trying to achieve usable systems. As long as the lack of theory and
methods persists in user interface design, we are left with individual designer's EX
PERJENCE and JNTUJTJON.
One of the means to obtaining more usable systems is to cultivate these domains. And
one of the most efficient means is to see the consequences of one's own decisions.
Decisions may be made appropriately in one context, e.g. in designing the internal
architecture of the system. But invariably the decisions will influence other aspects
of the system, in particular the user interface. This is illustrated by a quote from
the design of a hierarchically structured menu system:
"Initially we thought it's very important to have a nice structured hierarchy of
menus. Af ter we did some user testing on it we realized that it's a good starting
point, but I think that a successful set of menus is one that doesn't stick to a
hierarchy. The user wants to go from one path in the tree to another, so why not
let him?"
Another designer comments the same problem:
"In a hierarchical structure it seems unnatural to jump from one branch to another
the hierarchical structure pervades our thinking ••• "
It is striking that the concept NATURAL is brought up here. What is natural to a sys
tems designer is not necessarily natural to a user. [HAMMOND] present a range of ex
amples of user difficulties that can be tied back to the designers' perception of
what's natural.
In summary, prototypes have the potential to ameliorate the usability of systems,
particularly the user interfaces. They also provide a means for changing the system
designers' attitudes to systems design away from the general emphasis on the internal
structure of the system towards a more general recognition of the psychological and
social contexts of system application.
pjtfalls of prototypjna
The reader may have got the impression that prototypes is THE ANSWER to developing
more usable interactive systems. No, they are not. No such answers exist. But they
are likely to overcome some of the difficulties and constraints of the present ap
proaches. However, concern can also be raised regarding a number of pitfalls of
prototypes in addition to the promises.
Firstly, a large part -of the prototyping efforts will inevitably be carried out in a
more or less artificial context and not in the eventual application context. In
artificial contexts the USERS are far less likely to notice shortcom1ngs, to envisage
needed functions and shortcuts, to envisage the requirements of experienced users as
287
opposed to those of early users, etc. This is a constraint on the range of feedback
received. Real-life contexts, on the other hand, are very complicated and dynamic.
Of ten events interact in an unforeseen manner and cause new problems.
Secondly, along the same lines, lack of function during the tests may severely
restrict the value of the test. For example, screen image generators are excellent
means for dealing with screen layout. They fail, however, to capture the psychologi
cal complexity of the dialogue communications, e.g., the interplay between surface
appearance and task domain.
Thirdly, there's an obvious danger that prototypes take over the role played
previously by the written specifications. The fundamental idea of prototypes is to
iterate the design, not to FREEZE it. This leads to the fourth point, namely that
iteration implies discarding one's own work! The excuses used previously such as "It
can't be done" and "It will be too costly" don't work anymore. A fundamental change
of attitudes is needed. If users are to be involved in testing the systems they must
also be taken seriously.
Finally, the prototype approach implies involvement of users in the development
process, directly along with the designers or indirectly by testing the designs
proposed by the designers. This matches the current trend towards user participation.
It is, however, tempting to ask:
"Are users good designers?"
An explicit answer shall not be given here. That would be misleading. Instead two
design processes involving users are reported. The first concerns development of a
system for analyzing temporal data from psychological experiments [WRIGHTJ.
Two systems were developed, carrying out the same task. The first system was
developed from the user's conception of the task domain. The second was developed on
a concept derived by careful analysis carried out by the system developers
themselves. It turned out that the users preferred the second system! It was based on
far more clear-cut concepts and therefore easier to use.
The second experience concerns the development of a newsprint formatting system. One
of its designers narrates:
"The customer's view of the system was that there were 10000 little things to do
and each one of them should be worked out individually. We said we could never im
plement anything like that, we had to bring some consistency to the whole effort.
So we just ignored the customers requirements."
288
This is not exactly in line with current philosophies of user participation. However,
the designer continues:
"We took hundreds of newspapers and said: "How would we construct a system that
would produce this end result?""
The customers weren't obviously satisfied:
"They vehemently opposed what we were doing throughout the life of the thing. The
only reason it wasn't cancel led was that in the end it was given free to the
customer."
Nevertheless, the users eventually changed their minds:
"Their operators would start out, and it was very slow going because it looked so
foreign to them, just didn't fit their idea. But once they began to get a model of
the system, a view of the system, they had a tremendously accelerated learning
curve to a point where they could do virtually anything they were told the system
could do. The consistencies and similarities were such that they could just sit
down and use it!"
The lesson learned from these two stories is that it seems that a conceptually con
sistent view of the system is critical to its usability and that users don't
necessarily have such views, even in their own domain.
Conclusion
The concept of prototyping is likely to turn out to be a means for obtaining more
usable interactive systems in that it lends itself to fundamental properties of human
cognition. In the short term inappropriate designs caused by individual designer's
idiosyncrasies will be weeded. The say of the users will be based on far more solid
foundation with prototypes than with written specifications and can therefore be
advanced with more weight. In the long term prototyping provides a means for changing
designers' attitudes towards a broader view of systems development including the con
text of application with its complicated social and psychological interplays.
References
CLARK, I. A.:
Software Simulation as a Tool for Usable Product Desjgn.
IBM Systems Journal. Vol.20, 272-293 (1981).
HAMMOND, N., J. MORTON, A. MACLEAN, AND P. BARNARD:
Fragments and Signposts. User's Models of the System.
289
Proc. of the 10th International Symposium on Human Factors in Telecommunications.
Helsinki 6-10th June 1983.
JÖRGENSEN, A. H., N. HAMMOND, A. MACLEAN, P. BARNARD, AND J. LONG:
Design Practice and Interface Usability: Eyidence from Interviews wjth Designers.
(Report 83/10).
Copenhagen: University of Copenhagen, Institute of Datalogy.
MALHOTRA, A., J. C. THOMAS, J. M. CARROL, ANC L. A. MILLER:
Cognitjve Processes in Design.
Int. Journalof Man-Machine Studies. Vol.12, 119-140 (1980).
MORGAN, C., G. WILLIAMS, AND P. LEMMONS:
An Interview wjth Wayne Rosjng, Bruce Daniels and Larry Tesler. A behind the
Scenes Look at the Deyelopment of Apple's Ljsa.
BYTE. 90~114 (February 1983).
SMITH, D. C., R. KIMBALL, B. VERPLANK, ANC E. HARSLEM:
oesigning the Star User Interface.
BYTE. 242-282 (April 982).
TEITELMAN, W.:
A Display Oriented Programmer's Assistant.
Int. Journalof Man-Machine Studies. Vol.11, 157-187 (1979).
WRIGHT, P., AND G. BASON:
Detour Routes to ysabjljty: A Comparjson of Alternatjye Approaches to Multjpyrpose
Software Desjgn.
Int. Journalof Man-Machine Studies. Vol.18, 391-4DO (1983).