Back to the Outline.

The Computer User Interface
Robert Shear

Introduction:

Perhaps no human machine interface is more influential than the human computer interface, as we transition into the digital, information age, how we deploy and exploit computers will determine success or failure. As computers become more powerful, the demands on users escalate, creating problems of complexity and fatigue. An effective interface can go a long way to alleviating these problems. Humans are very highly evolved for shapes and colors which implies that the most effective interface would be graphical. With this essay we intend to explore the evolution of the Graphical User Interface (GUI) using a human factors framework. We will use the milestones of GUI design, DOS, Windows, and Bob, as a guide and then hypothesize on future directions.

DOS:

The first widely used minicomputer interface was DOS. DOS was instrumental in allowing home and business users to program and implement computers. This simple command line structure worked adequately for simple tasks but soon became cumbersome as tasks became more complex. Anyone who has used DOS remembers the "bad command or filename" prompt; usually the result of a misplaced space or a slash instead of a backslash. This quickly became annoying and it was painfully obvious that a better interface was needed if humans were to grow along with the computer. From a human factors point of view DOS is a disaster, DOS becomes incredibly complex very easily and provides no opportunity for experimental learning. It is all rote learning, memorize the correct command and syntax or forever wallow in error messages!

Xerox Star:

The first significant advance over DOS was the work carried out at Xerox's Palo Alto Research Center (PARC) which produced the Star system. The Star system was a true revolution in the design of computer systems, one in which the computer was designed around the user rather than requiring the user to adapt to an existing system. Jonathan Seybold of PARC put it this way, "Most system design efforts start with hardware specifications, follow this with a set of functional specifications for the software, then try to figure out a logical user interface and command structure. The Star project started the other way around: the paramount concern was to define a conceptual model of how the user would relate to the system. Hardware and Software followed from this."(1) Xerox took this philosophy seriously, devoting about thirty work years to the design of the Star interface. PARC researchers came up with a list of human factor issues which the Star system would address:

Need for a familiar conceptual model
Seeing and pointing versus remembering and typing
What you see is what you get (WYSIWYG)
Universal commands
Consistency
Simplicity
Modeless interaction

Developing a familiar conceptual model was the primary objective that had to be met. PARC researchers defined a conceptual model as follows, "A user's conceptual model is a set of concepts a person gradually acquires to explain the behavior of a system, whether it be a computer system, a physical system, or a hypothetical system. It is a model developed in the mind of the user that enables that person to understand and interact with the system."(1) An important idea to notice here is that the conceptual model is built gradually over time, so in a rapidly changing environment it is more effective to mold the technology to the model rather than mold the model to the technology. PARC researchers based their system on an office analogy, pretty much what we see today with Windows, "We decided to create electronic counterparts to the physical objects in an office: paper, folders, file cabinets, mailboxes and so on - an electronic metaphor for the office."(1) The use of metaphor allows the computer to be assimilated into the workhabits of the user much easier and with less training.

The remainder of the design objectives are concepts which allow the user easier access to the computer through the framework of the conceptual model. "Pointing versus typing" and "What you see is what you get" are two very interrelated concepts. PARC scientists developed the mouse to facilitate the former. By pointing and dragging the physical office analogy becomes effective. The latter is important in uncluttering short term memory, "A well designed system makes everything relevant to a task visible on the screen. It doesn't hide things under CODE+key combinations or force you to remember conventions. That burdens your memory."(1) The interface in a sense becomes a virtual short-term memory making thinking easier and more productive. Also, what you see is what you get allows the user to edit documents on the screen and then print an exact reproduction. Before Star, fonts were not discernible on screen, with it's bitmap display, Star allowed users to really work on a virtual document. From a human factors perspective this is vitally important for maintaining the office space analogy.

Building on top of the aforementioned attributes, Star employed four design guidelines which would allow users to intuit applications and learn by doing, these were: universal commands, consistency, simplicity, and modeless interaction. These four guidelines are vitally important to the final success of the Star operating environment, this is where the rubber meets the road for the user. By maintaining these objectives the Star system was the first to let users interact in an instructive manner with a machine. Universal commands and consistency allowed skills learned for one application to be transferable to other applications, thereby reducing the time needed to learn and become proficient at new applications. Simplicity is the cornerstone of all human factors designs. Especially with the computer, complexity kills productivity. This is what happened to DOS. As the complexity of applications increases our ability to assimilate information remains relatively constant: we need the computer to take some of the burden from us. The most effective way to insulate the user from complexity is to make the interface as simple as possible. Modeless interaction is akin to WYSIWYG design. The need for DOS based machines to have mode keys to squeeze more functions out of the keyboard essentially "hid" the information and relied on the user to remember which mode was operating. This increased burden on short-term memory and reduced user effectiveness. Employing modeless design in the Star system relieved the user of this burden and simplified the execution of tasks.

The work done by Xerox PARC on the Star system laid the foundation for the way we interact with computers today. Incremental advancements have been made but the underlying design philosophy has remained the same. MS Windows and MacOS are really just modified Star interfaces.

Microsoft Bob:

The next real attempt to apply human factors to computer interface design was Microsoft "Bob". Bob was a short-lived operating system which was an altered version of Windows. Instead of using a desktop and office space analogies Bob went the next logical step and used a work room analogy. For example, to bring up a calendar application the user would click on an image of a calendar hanging on the wall of the office. To check the time, the user would click on the image of a clock on the wall and so forth. The designers of Bob relied on a different conceptual model than the designers of Star. In our opinion the model was valid but it violated some key human factor tenants, primarily the desire for simplicity. The Bob interface was quite cluttered with unnecessary images of windows, plants, wood paneling, etc., things that are in a real office but have no purpose in a conceptual model. Additionally there was not congruence with the intended user and the price point. The users who would benefit most from Bob were computer novices whom had not adopted the standard Windows model. The problem was that Bob was complex and resource hungry. The incompatibility of first-time users and top of the line system requirements helped to kill Bob. The lessons to learn from Bob are: 1) Simplicity is king and, 2) Human factor design is not enough for acceptance.

Microsoft Active Desktop and Netscape Constellation:

What we are seeing today is a continuation of a trend exploited by Microsoft in MS Office and Windows 95, bundling and cross application commonality. As MS Office allows users to set configuration preferences between applications and the ability to port work between applications and users, Active Desktop and Constellation will allow us to integrate our desktop and network needs. We will soon be able to interact with one interface for all of our computer requirements. From a human factors perspective this allows for convenience and lessened aggravation. Another significant human factor consideration is the capability of Constellation to store and retrieve user preferences at any workstation. This means that my personal configuration settings will travel with me no matter what machine I log on to, allowing for more familiar and intuitive interaction.

Conclusion:

Human computer interaction will continue to be a hot topic for decades to come. Until we can directly transmit thoughts to machines, the way we exchange information with the computer will be a primary determinant of user effectiveness. As processor power and bandwidth continue to grow the limitations of the human user will become more and more apparent and products will be differentiated on utility not power. The PARC scientists realized this when working on Star, "In the 1980's the most important factors affecting how prevalent computer usage becomes will be reduced cost, increased functionality, improved availability and servicing, and, perhaps most important of all, progress in user-interface design. The first three alone are necessary, but not sufficient for widespread use. Reduced cost will allow people to buy computers, but improved user interfaces will allow people to use computers."(1) This passage has become even stronger with time. The technology is in place, the challenge now is to adapt it to the user and extract is full potential.

References:

See HF Group Bibliography.

Back to the Outline.