The U P A voice
April 2004 Contents

Featured UPA Links:

Job Bank

Member Benefits:
Discounts and Special Offers:

Event Discounts
Order your usability books now from the link from this site. Part of all sales help fund UPA.

Difficulties in Modeling GNU/Linux User Behaviors

Matt Queen

Washington University

Creating models of user behavior has been helpful in predicting basic outcomes of computer usability testing involving human subjects. However, models and methods have been based on a narrow view of computer use; namely, they are not compatible with behaviors resulting from using the Linux operating system. How different could Linux be from other operating systems?! This article provides a few points of comparison.


multiple modes in a GNU/Linux OSUsability has been a top concern for desktop environment development groups[1] and plenty has been written about usability of the Linux operating system (OS) and the GNU package of tools meant to provide Unix-like capabilities[2, 3, 4]. However, existing models used to characterize computer users do not directly apply to GNU/Linux users. GNU/Linux is fundamentally different than other OS's in terms of users' cognitive behaviors. OS's like Windows and Mac make installation and system administration tasks fairly autonomous. GNU/Linux users are forced to consider installation locations and are often responsible for validating and clarifying dependant software libraries. Additionally, typical GNU/Linux users are required to use both the graphical desktop environment and the command prompt.

Current models of user behavior are not equipped to predict responses to these complex multi-modal cognitive tasks. Card et al. (1983)[5] created a basic behavioral architecture Human Processor Model (HPM). Kieras and Meyer (1997)[6] developed Executive Process-Interaction Control (EPIC) by adapting HPM to include programmable rule structures and constraints that simulate results found in the human subject testing literature. For a review see Hornof and Kieras [1999][7]. Additionally, attempts have been made to use cognitive behavior models like EPIC to describe perceptual tasks such as planning eye movements while navigating hierarchical computer menus[8]. However, generalized biological models of vision have not yet been integrated into computational behavioral models for Human-Computer Interaction (HCI) testing. Task scripting has also not grown to encompass complex modal shifts like those required of GNU/Linux users.

Shifts in Modality

Users switch modes while learning and using an OS. Modes can be characterized as goals for OS use. Passive goals, such as chat, play, and visual customization, are not hindered by deadlines, are less structured, and have a large exploratory component. Active goals, such as word processing, programming, and presentation design must adhere to deadlines, are heavily structured, and contain a relatively small exploratory component. The exploratory component is where users build their knowledge structure of the OS. The exploration component of user modes fluctuates not only with goals but also with experience. Less exploration is necessary when the majority of the OS has been learned. In fact, users migrate from passive use of the OS to active use of the OS.

Task Complexity

Aside from modalities, users also perform tasks of varied complexity. Halford et al. [1998][9] have defined three levels of complexity for cognitive tasks based on the number of relational processes that must be performed simultaneously. 0, 1, and 2 dimensional relationships denote the number of simultaneous processes necessary to complete the task. Use of the GNU/Linux OS commonly requires tasks that entail all three of these levels of relational complexity. 0-level tasks, such as using menus, require no relational integration. 1-level tasks like searching/ browsing (i.e. scan till you find X), using edit-command applications like Vi and Emacs (i.e. change mode to issue command or edit text), and copying and pasting text require integration of only one dimension. 2-level tasks are necessary where information must be integrated across applications. For example, computer mediated communication applications like email and chat are often used as a mode of collaboration while completing group projects involving word processing, graphic manipulation, and event planning. An additional example of 2-level task requirement is collaborative software development.

Further Work

In order to account for GNU/Linux user task complexities and modalities an attempt should be made to combine the passive versus active tasks framework and Halford and colleagues' relational task complexity framework.


  1. Benson, C. 2001. Evaluations of GNOME Usability: Expanding the Appeal of GNOME, GUADEC, Copenhagen , March.
  2. Nichols, D. and Twidale, M. 2002. “Usability and Open Source Software”,, October.
  3. Kamiyo N., Yasuhiro Y., Yoshiyuki N., Kouichi K., and Yunwen Y. 2002. "Evolution Patterns of Open-Source Software Systems and Communities," Proceedings of the Workshop on Principles of Software Evolution, International Conference on Software Engineering, pp. 76-85.
  4. Watkins, R. 2003. What is Open Source Software and is it Usable? The UPA Voice , Vol. 5, No.1.
  5. Card, S. K., Moran, T. P., & Newell, A. 1983. The Psychology of Human-Computer Interaction. Hillsdale , NJ: Lawrence Erlbaum Associates.
  6. Kieras, D. E., & Meyer, D. E. 1997. An overview of the EPIC architecture for cognition and performance with application to human-computer interaction. Human-Computer Interaction , 12(4), 391-438.
  7. Hornof, A. J. and Kieras, D. E. 1999. CognitiveModeling Demonstates How People Use Anticipated Location Knowledge of Menu Items. Computer-Human Interaction , 99, 15-20.
  8. Nilsen, E., Evans, J. 1999. Exploring the Divide Between Two Unified Theories of Cognition: Modeling Visual Attention in Menu Selection. Computer-Human Interaction , 99, 15-20.
  9. Halford, G. S., Wilson, W. H., and Phillips, S. 1998. Processing capacity defined by relational complexity: Implications for comparative, developmental, and cognitive psychology. Behavioral Brain Science, 21, 803–831.
  Usability Professionals' Association
140 N. Bloomingdale Road
Bloomingdale, IL 60108-1017
Tel: +1.630.980.4997
Fax: +1.630.351.8490
Contact the Voice