Human Computer Interaction
ERCIM News No.46, July 2001 [contents]

Specifying User’s Knowledge: A Cognitive Ergonomic View on IT Design

by Gerrit van der Veer

There are strikingly different views on what should be the focal points in designing technology for human users: graphical design, interaction design, engineering of state-of-the-art requirements, or a cognitive psychological approach. At the Vrije Universiteit Amsterdam scientists developed an eclectic approach towards the design of interactive systems named DUTCH (Designing for Users and Tasks, from Concepts to Handles).

The core of this approach is to define all relevant knowledge a user should possess to use the new system. Such a full specification of all aspects of the technology to be (re)designed, from the point of view of the prospective user, is called UVM (user’s virtual machine).

In DUTCH we consider interactive systems where various types of users may have different roles, each of which needs different knowledge to use the system. Specifying a complete interactive system will amount to specifying various more or less overlapping UVMs, one for each role, based on the user’s viewpoint. Adaptability is clearly separated from specifying the knowledge needed to perform role specific tasks. Whereas for the user the system appears monolithic, the designer has to find out what types of knowledge will have to be specified. In the first place the system should speak a language that is understandable for the user in relation to his intentions and tasks. In interactive systems, however, much more complex knowledge is needed. Users need to be able to instantiate a mental model of what is going on in interaction, based on knowledge available in human memory and information provided by the system at the relevant moment. This knowledge and information must be specified in an early phase of the design. We should systematically take care of the different types of knowledge that will be engineered at a later stage (probably by very different specialists who may never meet each other, like software engineers, hardware designers, graphical designers, workflow managers, and systems architects).

What is going on behind the Screen
Users hardly care about ‘what’s inside’. Only if certain internal processes take more than a few seconds, or are unpredictable for the user, the system should make the user aware of exactly this, nothing else. On the other hand, certain things going on behind the screen are needed to understand the interaction and perform the intended tasks successfully (eg, “is this object still on my clipboard?”; “is the original file untouched after I performed a ‘save as’ operation?”).

What is going on behind this Computer
Current interactive systems are seldom stand-alone. Even if we connect the machine to the outside world solely by exchanging diskettes, users need to know the reason for certain unexpected events, and take measures for safe, reliable, and efficient interaction with external agents. In the case of more sophisticated connectivity, users need to understand aspects of local and world wide connections and safety procedures (fire walls, signatures, authentication). In certain types of tasks, the actual location of certain processes should be known and understood (“is my pin code transported through the network to the bank, or is it validated locally at the terminal?”).

What Organizational Structure is behind my Computer
In the envisioned interactive systems, a user needs to understand that other users are players in the same game, influencing the total process, for example by managing their workflow process or monitoring their actions, or by having different rights (“who may inspect or change or delete the e-mail I have just sent?”).

What Task Domain is available through my Computer
If a fashion designer uses an interactive system, several tasks could be delegated to the computer with considerable added value, for example to search multimedia databases of video clips from cat walk shows, images of the people and the location the clothes are intended to serve, interviews with potential clients, and works of art that are characteristic of the cultural ambiance of the intended buyers. So far, possibly because the right media are not yet available, other parts of the task domain are not yet supported successfully, for example assessing the texture quality of fabric, or impressionistic sketching of the process of moving clothes at a stage before the design of shape has developed to enough detail for a mannequin to be actually dressed.

What are the Process and Time Aspects of delegating Tasks to my Computer
IT time scales are completely different from those of human information processing. Because of this, systems often react to user requests either amazingly fast or annoyingly slow and unpredictable. A more fundamental problem occurs if the computer is a front-end to a process where time characteristics are of a non-human scale. Here either the feedback is slow (eg, monitoring chemical plant processes) but the user is supposed to anticipate the outcome of previous actions in order to optimize the continuation of the process, or the process is fast in performing irreversible tasks (eg, delegating certain flight operations to the auto pilot) and, consequently, the user is expected to assess the expected consequences in all details that affect safety and reliability. Actually, both cases are requesting the same type of insight and knowledge from the user. Various recorded cases where pilots ended up ‘fighting’ the auto pilot are illustrations of designs where this type of knowledge was never considered. The same can be found in recorded disasters with nuclear power plants where potential consequences of task delegation were never modelled, and failure indications were completely inadequate and even disturbed human problem solving processes.

Specifying the above-mentioned types of user knowledge requires a modelling language that is precise and unambiguous enough to be understood by engineers without deep insight in cognitive science. At the same time this language should enforce complete specification of what matters to the user and what is relevant from the point of view of human information processing. We developed NUAN (new user action notation, adapted from Hix & Hartson) into precisely the modelling language and supporting tool that is needed for this.

Specifying the details of technology, from the point of view of DUTCH design, is equivalent to specifying all relevant knowledge needed for each type of user. Designing technology is thus translated into building prescriptive knowledge models, intended for the implementing engineer. Exactly this knowledge should also be available to be instantiated into a mental model whenever interaction requires it. Of course, in prescribing this knowledge, the cognitive ergonomic insights in human information processing are of paramount importance. The system should be understandable, usable, and require the lowest level of effort on the part of the user.


Please contact:
Gerrit van der Veer — Vrije Universiteit Amsterdam, The Netherlands
Tel: +31 20 444 7764