Mental Models - A Theory Critique

By Andrew Kurtz

This paper is a critique of the human-computer interaction theory of mental models. The origin and motivation for the development of the mental models theory will be discussed along with a analysis of how useful the theory is in informing the design and evaluation of user experiences.

 

Summary

A mental model is a conceptual representation within a person's mind that is used to help the person understand the world and to help the person interact with the world. For example, my mental model of how a telephone works might be that I pick up the phone to initiate a connection, dial the number I want to call, hear the phone on the other end ringing, and then the other person answers. That model is not very detailed and actually is inaccurate (the ringing I hear is not the other phone), but this model is sufficient for me to be able to operate the device.

The mental models people create of computer systems are normally inaccurate (Norman, 1983). Having an inaccurate model of how a system works may cause problems while interacting with the system. By studying how people create mental models of interactive systems and by designing interactive systems that help the user create a more accurate mental model of the system, usability will improve.

 

Motivation

Researchers have been trying to understand how the human mind works for centuries. Various theories have been introduced to help researchers model human thought and behavior. The theory of mental models comes from ideas introduced by Johnson-Laird (1983) that involves a type of mental representation called mental models, which are structural analogies of the world. The idea is that humans use these mental models to understand the world and how to interact with it. By studying the mental models people use to explain the world, researchers can better understand how people perceive the world and what directs how they interact with the world. If we, as human-computer interface designers, can create interfaces and interaction methods that help people create a more accurate mental model of the system being interacted with, the assumption is that people will have more success interacting with the system.

 

History and Background

Some of this history was adapted from Sasse (1997).

The term mental models came from the work of Johnson-Laird (1983) and refers to an internalized, mental representation of a device or idea. Johnson-Laird credits Craik (1943) with the initial idea of mental representations.

Norman (1983) was one of the first attempts to create a terminology for a human-computer interaction theory of mental models where he introduced the idea of there being different models of a system. The mental model is envisioned by the user, the conceptual model is an accurate model of the system created by the designer, and the target system is what the user interacts with.

Later, Norman (1996) modified this terminology to be a user's model and a design model, which are both conceptual models, and a system image which is implementation of the system. This is the terminology I will use in this paper. The designer creates a design model that is communicated through the system image and a user develops user model through interaction with the system image.

The idea is that if the designer creates the design model right, and communicate the model successfully through the system image, users interacting with the system will develop an appropriate user model, which will allow them to interact with the system successfully (Sasse, 1997).

Over the years there have been different perspectives of mental models introduced by various researchers. One of the main complaints about mental models is this existence of different definitions. Recently, Payne (2003) discusses some issues he finds in the past theories of mental models and proposes a new interpretation. One of the issues that Payne has with current mental model theories is that people tend not to create the same mental model of a system. He discusses a study where people are asked questions about how an automatic teller machine (ATM) works. People report very different understandings about how the ATM does what it does. Payne implies that this is a weakness of mental models, but I would say this is precisely why studying mental models in valuable. The screens and interaction of the ATM machine lead the users to create invalid mental models of the system. Comparing the user's mental model with the system image may lead to redesign suggestions that will improve the user's understanding of the system.

 

Explanation

The theory of mental models is based on creating mental representations of things in the world. Those models may then be used to help train a user on a system or to help explain a user's interaction with a system.

Mental models were first introduced as an internalized, mental representation of something in the world. Johnson-Laird started this idea and applied it to things such as the spatial arrangement of objects (Johnson-Laird, et. al., 1998). Others, including Norman and Payne, adapted the idea for use in human-computer interaction.

Mental models may be (1) An image, (2) A script, (3) A set of related mental models, (4) A controlled vocabulary, or (5) A set of assumptions (McDaniel, 2003). In many cases a mental model may contain aspects of one or more of these types of models. A user may have an image of the look of an interface, a script of the process to be followed when completing a task, knowledge of the vocabulary the system uses, and assumptions about the behavior of the system.

Two types of mental models have been identified, structural and functional (Preece, et. al., 1994). A structural model is used to describe the internal workings of a device, which is then used to make predictions about the operation of the device. A common example of this is a home furnace thermostat. If the user wants the temperature to reach the desired temperature faster they will typically turn the thermostat higher than the desired temperature assuming that will cause the temperature to rise faster. That is an incorrect assumption and will create unrealistic expectations about the furnace.

A functional model describes how a device works and can be used directly to interact with the device. An example of this would be my phone model described earlier in this paper. The model was inaccurate, but it was accurate enough to cause me to properly use the phone.

The accuracy of all of a person's mental model will affect how the user interacts with the system. The more accurate the model, the more successful the interaction.

Methodology

Two methods of creating and analyzing mental models come to the fore font based on the literature. When the model is being used for training, it is created by the researcher based on the system image. Presumably this model is very accurate and provides the correct "view" of the system. That model is used during the training to help the users understand the function of the system and when such a model is used, it appears to improve a user's performance.

When a mental model is being used to understand a user's behavior, the model may be interpreted from a verbal protocol analysis of a user's interaction with the system. The model that results from this analysis may not accurately reflect the mental model the user has since it is extracted from the verbal protocol or through questioning the user. In addition, it may be difficult for the user to explain their mental model of the system.

 

Case Studies

One of the uses of mental models is to aid training users in the use of a system. Borgman (1986) performed a study where notice users were trained in the use of an online catalog. The control group was given a set of procedures for retrieving literature from the catalog. The experimental group was given the procedures and had the system explained through an analogy with the card catalog. Completion times and number of tasks completed for simple tasks were not different between the group, but for complicated tasks the group that was trained on the model performed significantly better. This may imply that having a more accurate model of the system will help users complete complicated tasks.

Mental models may also be used to explain the behavior of users. Gray (1990) observed and questioned novice users of a hypertext system to determine the user's mental model of the system. The models were inferred from the user's behavior and verbal protocols. In addition, drawings were used to show information was connected. Most user's models were of a sequential nature, like they would search in a book. Over time the user's models improved to reflect the hierarchical nature of the system. This shows that user's adapt their model over time based on interaction with the system. In addition, Gray discovered that many of the problems came from misunderstanding of the meaning of the terms used in the system. If the user's would have a more accurate vocabulary model, they may have had more success with the interface.

 

Pros and Cons

Mental models are applicable in many situations, but the diversity of definitions and the lack of a coherent methodology may cause confusion and may result in contradictory results. It appears that most researchers develop their own methodology based on verbal protocol analysis. This type of analysis may contain significant bias introduced by the experimenters interpretation. Since it is difficult to determine the user's mental model, different experimenters observing the same interaction may derive different mental models.

Even though it is assumed that people use mental models and designers are encouraged to design interfaces that help users create accurate modes, there is little research about how to design to help the user create the best mental model (Preece, et. al., 1994).

From this evaluation, it appears that the biggest gain from using mental models can be achieved by using them in training. Providing the user with a more accurate initial mental model will start the user performing at a level that normally would come from the user's mental model adapting over time.

 

Links

Eliciting and Describing Users’ Models of Computer Systems
Martina Angela Sasse, Ph.D. Thesis, Computer Science, University of Birmingham
This Ph.D. Thesis provides an overview of mental models and examples of a number of empirical studies using mental models.
What's Your Idea of a Mental Model?
Scott McDaniel. Boxes and Arrows. February 10, 2003
This is a short article that provides an simple introduction to mental models.
Mental models: a gentle guide for outsiders
P.N. Johnson-Laird, Vittorio Girotto, and Paolo Legrenzi
This paper talks about mental models with examples modeling the spatial arrangement of objects.
 
CPSC 444: User Interface Design - Lecture Notes (in PowerPoint)
Brian Fisher. Media and Graphics Interdisciplinary Centre, The University of British Columbia.
These lecture notes describe the use of mental models in conjunction with Norman's execution-evaluation model for evaluating systems.

 

Comments

Through this evaluation I have concluded that my view of the applications of mental models is different from most of the literature I could find. The experiments discussed involve the use of mental models to facilitate learning by the user or understanding by the experimenter. There was little discussion about the use of mental models during the design process.

My view is that looking at the execution-evaluation cycle (Norman, 1990) of a system in conjunction with the user's mental model will help designers understand where user may have problems interacting with the system. I did find this opinion implied in a set of lecture slides for a user interface design course (Fisher, 2003). Fisher discusses how mental models provide an explanations and predictions about the interface and how these relate to Norman's execution-evaluation cycle. Mental models explain what the user is seeing and what the system is doing along with predicting what the user can do next, what the system will do given certain user actions, and what the user will see as a result of those actions. These explanations and predictions relate to errors that may occur in the interaction cycle. Not recognizing the system state interferes with the perception stage, not selecting the appropriate action is a problem in converting an intention into an action, and unexpected results from the system will interfere with the evaluation interpretation stage.

Analyzing the interaction design of a system using the execution-evaluation cycle based on the design model can help refine the design before user testing. Performing user testing and extracting the user model based on verbal protocol analysis and questions then looking at the interaction cycle based on the user's model will help find where the user might have problems moving through the cycle. When used in that way, mental models may be very useful in the design process.

 

References

Borgman, C. L. (1986).
The User's Mental Model of an Information Retrieval System: an Experiment on a Prototype Online Catalog. Journal of Man-Machine Studies, 24, 47-64.
 
Craik, K. (1943).
The Nature of Explanation. Cambridge: Cambridge University Press.
 
Fisher, B. (2003).
CPSC 444: user interface design - lecture notes, from http://www.cs.ubc.ca/~cs444/lectures/444-4_mentalModels.pdf
 
Gray, S. H. (1990).
Using Protocol Analyses and Drawings to Study Mental Model Construction during Hypertext Navigation. International Journal of Human-Computer Interaction, 2, 359-377.
  
Johnson-Laird, P. N. (1983).
mental models. Cambridge: Cambridge University Press.
 
Johnson-Laird, P. N., Girotto, V., & Legrenzi, P. (1998).
Mental models: a gentle guide for outsiders, from http://www.si.umich.edu/ICOS/gentleintro.html
 
McDaniel, S. (2003, February 10, 2003).
What's Your Idea of a Mental Model? Boxes and Arrows.
  
Norman, D. A. (1983).
Some Observations on mental models. In D. Gentner & A. Stevens (Eds.), mental models. Hillsdale, N.J.: L. Erlbaum Associates.
  
Norman, D. A. (1990).
The Psychology of Everyday Actions. In The Design of Everyday Things. New York: Doubleday/Currency.
  
Norman, D. A. (1986).
Cognitive Engineering. In N. D. A. & D. S. W. (Eds.), User-Centered System Design: New perspectives in human-computer interaction. Hillsdale, NJ: L. Erlbaum Associates
  
Payne, S. J. (2003).
Users' mental models: The Very Ideas. In J. M. Carroll (Ed.), HCI Models, Theories and Frameworks: Toward a multidisciplinary science. San Francisco: Morgan Kaufman.
Preece, J., Rogers, Y., Sharp, H., Benyon, D., Holland, S., & Carey, T. (1994).
Human-Computer Interaction. Harlow, England: Addison-Wesley.
  
Sasse, M. A. (1997).
Eliciting and Describing Users’ Models of Computer Systems. Ph.D. Thesis, Computer Science, University of Birmingham.