TiChi Introduction
Cassie Thomas
October 28 2002

As the domain of Human Computer Interaction (HCI) has evolved, research has led to the development of many explanatory models, prescriptive guidelines, and predictive laws, all of which could be called theories. These theories have borrowed heavily from other disciplines, most notably psychology, graphic design, and information science. Theory construction accelerates progress in all disciplines as it helps to describe phenomena, explain processes, predict outcomes, support recommendations.

A primary goal of HCI theories is to aid practitioners to design interactive systems that are easy to use, rapid in performance of standards, and aesthetically pleasing to the user.

A second important role of theories is to provide a clear definition of concepts so that researchers can coordinate their efforts and instructors can explain concepts to students.

A third role is to help researchers to guide their investigations, which eventually facilitates the development of better systems. Research projects that are guided by theory are more likely to have repeatable (replicable) and generalizable results because theories guide researchers to plan more careful experiments and derive more potent results.

Two important qualities of a theory are "verifiability" and "falsifiability". A theory is verifiable if the theory makes predictions that can be confirmed through experiments. If a theory's predictions do not match results, then it can be updated or discarded in favor of a more accurate theory. A theory is falsifiable if it is possible that the outcome of an experiment would show that the theory is incorrect [2]. For example, Freud's theory of id, ego, and superego is not falsifiable: since any possible human behavior could be explained by this theory, it is not possible to disprove. Generally, the scientific community rejects theories that are not falsifiable. However, in an applied field such as HCI, theories that are not falsifiable may still find applications.

Theories of HCI tend to fall into the following five categories: descriptive, explanatory, predictive, prescriptive, or generative [2]. These categories are not mutually exclusive, and so some theories fall into more than one category. These theories allow practitioners and researchers to [1]:

  • describe objects and actions in a consistent and clear manner to enable cooperation,

  • explain processes to support education and training,

  • predict performance in normal and novel situations so as to increase the chances of success,

  • prescribe guidelines, recommend best practices, and caution about dangers, and

  • generate novel ideas to improve research and practice.
  • A prescriptive theory gives directions or rules as to how something should work or be carried out. For example, a model may suggest how a menu should be laid out.

    A descriptive theory seeks to support the process of thought, by providing consistent and appropriate terminology.

    An explanatory theory can explain the world. It support training people and building conceptual models.

    Predictive theories, most commonly used in HCI allow researchers to predict user experiences/interactions with performance metrics and error counts in use of a system. If designers propose a new system, predictive theory allows them to predict what the outcome might/could/should be.

    Generative theories catalog existing user interfaces and therefore could guide designers or researchers in inventing novel applications. Much as the periodic table of chemical elements guided scientists to look for unknown chemicals, a generative theory of HCI might suggest novel applications such as email, instant messaging, online communities, or digital photography. A generative theory of HCI might be based on human needs, such as the Activities and Relationship Table [5]. The 4 x 4 Activities and Relationship Table can be used to catalog existing applications and possible suggest what might be helpful and well received by users.

    Depending on the theory chosen to use for a knowledge background in HCI, will depend on the kind of methodologies that are appropriate to use on subjects. Two important research methods are qualitative and quantitative experiments. A qualitative experiment seeks to produce a description, typically without the use of numbers. It can be subjective in that the results are based on the experimenters' point of view. Several methods are used in a qualitative experiments; introspection, surveys & questionnaires, and direct observation [4]. These methods mostly involved observation of the user from the researcher.

    Quantitative experiments follow the reductionist scientific method more closely. Researchers who employ quantitative experiments usually develop a hypothesis about a situation and then conduct a controlled experiment to measure statistical results . They will alter a small number of independent variables and measure a small number of dependent variables tied to the hypotheses they are seeking to prove. In HCI, typical dependent variables are user performance on a system for specific tasks, for example error counts and completion time. It may compare a set of users and their performance using a set of systems.


    [1] Bederson B. & Shneiderman B. 2002. The Craft of Information Visualization: Reading and Reflections.

    [2] Popper, K. The Logic Of Scientific Discovery, 2nd ed. Harper Torchbook, New York, NY, 1968

    [3] Shneiderman B. Designing the User Interface: Strategies for Effective Human-Computer-Interaction, 3rd. ed. Addison-Wesley, Reading, MA. 1998

    [4] Wilson, E.B. Introduction to Scientific Research, McGraw-Hill, New York, NY. 1952

    [5] Shneiderman, 2002. http://mitpress.mit.edu/main/feature/leonardoslaptop/pdf/chapter5.pdf. October 28, 2002