A user interface can be described as anything that enables us to perform a (usually practical) task. In that sense, almost everything can be considered a user interface: a coffee pot is a user interface that enables us to make coffee, a door is a user interface that enables us to enter and exit a room, ... even a sidewalk is a user interface, enabling us to walk from one point to another.
The complexity of a user interface generally increases with the complexity of the task it is facilitating. It is therefore understandable that the user interfaces for computer systems—which offer us the ability to perform a multitude of complex operations, from managing an MP3 collection to running scientific simulations—can be very complex. This makes the job of the user interface designer very difficult. How does one design an interface to all this functionality, while at the same time keeping it usable?
Adding to this difficulty is the fact that computers provide a visual environment, and there is an increasingly visual nature to computer interfaces. Specifying all the interactions mediated by these GUIs—a necessity for any comprehensive design document—is a near-to-impossible feat when compared to doing the same for text-only command-line interfaces (due to both inadequate theories, and the creativity and complexity inherent in graphical media). But wait ... it gets worse.
Consider the future trends in computing [4]:
Eventually, we may get to the point of autonomic computing, when computers are self-repairing, autonomous, transparent, and ubiquitious. As these trends become part of our here and now, currently available user interfaces will no longer suffice. If user interface designers suggest that GUIs have made their lives tough, how will they react when virtual or augmented reality interfaces replace the standard keyboard, mouse, and monitor?
User interface design (UID) for computer systems is not a simple business, but neither can it be swept under the rug. Even if a system meets all its performance and functional requirements, a poorly designed user interface can result in its being completely useless. It is therefore important to consider what makes a user interface "good", and what issues are involved in designing such an interface.
The ultimate goal of UID is transparency. This occurs when the interface becomes "invisible", and the user can focus on performing the task at hand. For example, to many computer users, the mouse is a transparent interface to the pointer on the screen: they can think in terms of moving the pointer , rather than moving the mouse to move the pointer.
[6] provides five "goals" of UID, which can instead be thought of as "gauges" of transparency. These are:
An interface that performs better as measured by these gauges can be said to be "more transparent". (Especially noteworthy among the gauges is the rate of error by users. Transparency breaks down when the interface breaks down, and this very commonly occurs when faced with unexpected user behavior—an additional incentive for user interface designers to gracefully handle all error conditions.)
According to [4] there are four levels to UID issues:
Though often ignored in actual UID, there is now widespread recognition of the significance of social issues (the first three issues are well studied and understood, and will not be discussed further here). Social issues (i.e., user values) are more difficult to incorporate in design (by definition, they require user involvement), but are as (if not more) important. Data from "software economics" reveals that about 50% of the cost of error correction for large projects comes from errors in requirements, and most serious problems with requirements are grounded in social issues [4].
The most obvious way of incorporating social issues in a user interface is iterative design. This involves designing in stages, with user testing and evaluation following each stage. Unfortunately, this is time consuming, and therefore expensive.
A better solution derives from the field of ethnomethodology, which among other things, deals with discovering the values of a social group. What if user interface designers could use the techniques of ethnomethodology to learn the values of their users? They would still require a way to incorporate those values in their design. The answer ... semiotics.
Semiotics is the study of signs. Here, signs are not just the simple, physical things commonly called "signs" (these will be henceforth referred to as "tokens"). Instead, a sign is anything that mediates meaning; this can include words, images, sounds, and even gestures. Signs can be complex combinations of lower level signs; for example, a book is a complex sign that is a combination of paragraphs, which in turn are combinations of sentences, and so on. The term "semiotics" was coined by the nineteenth-century American logician Charles Sanders Peirce (pronounced "purse"), who also introduced several of its basic ideas. Most importantly, a sign's meaning is dependent on context: the same sign can mean something else to somebody else. Peirce also described a system of classifying signs by how they convey meaning. The other significant contributor to semiotics was the Swiss linguist Ferdinand de Saussure, whose key insight was that signs come in systems, and should therefore be studied as such.
A user interface is a complex, structured sign. The semiotics of Peirce and Saussure ("classical" semiotics from this point on) provides a theory for deducing how a user community might interpret such a sign, but by itself, is insufficient for UID. The "D" in UID stands for design, which refers to the process of creating a configurations of signs that will enable a user to perform some task. We therefore need a way to not only analyze signs, but to synthesize them. Another drawback of classical semiotics lies in its somewhat Platonistic beliefs, suggesting that signs and their meanings can exist without people (i.e., without the social element). To correctly utilize semiotics for UID, we require a synthetic, socially-situated semiotics.
Algebraic semiotics is a new field, drawing on the most important concepts of classical semiotics, and offering solutions for its failings. It is a precise formalism of sign systems, providing a way to analyze signs, as well as to synthesize them. Once a task has been investigated with respect to the user community, algebraic semiotics simply requires that you determine the source system, and map to a design space. Using priorities and orderings with this mapping ensures that the user values are retained. This formalized process allows the prediction of measurements for the gauges of transparency mentioned earlier, and consequently allows determining the optimal choice from competing designs.
As a form of disclaimer, a thorough understanding of classical and algebraic semiotics does not guarantee that you will design a perfect interface. Oftentimes, there is no "perfect" interface. As the linguist Edwin Sapir liked to say, "All systems leak." Basically, no formalism is flawless, but there is still use in writing it down. Some model is better than no model at all, and in this case you are provided a way of incorporating user values into the design of an interface, which is otherwise a very difficult task. It's no utopia, but it is a step forward.
The remaining pages of this site discuss classical and algebraic semiotics in more detail. The examples given are primarily related to user interfaces, particularly those of personal computers and their software. This should not be construed to mean that classical, or even algebraic, semiotics, is in any way limited to such applications. Quite to the contrary, semiotic theory emerged from philosophy and linguistics, and has been most broadly applied to the arts and social sciences. The explicit use of semiotics in UID is a comparitively recent development and is not yet widespread.