Cognitive style and its implications for navigation strategies Aude Dufresne Dept. of Communication Sciences University of Montreal C.P. 6128, Succursale A Montreal, Quebec, CANADA H3C 3J7 +1 514-343-7371,
[email protected]
Sylvie Turcotte HCI Group Computer Research Institute of Montreal 1801, McGill College, Suite 800 Montreal, Quebec, CANADA H3A 2N4 +1 514-840-1263,
[email protected]
ABSTRACT In this paper, we investigate the effect of cognitive styles (field dependence)for navigation strategies. The application teaches the use of the database features in Microsoft EXCELä. Two versions of the interface were developed offering free vs constrained hypermedia access to the information. A statistical analysis of the trace revealed important differences of exploration behaviors based on user characteristics. The paper provides extensive studies review and empirical validation for the need to consider cognitive styles to design more adapted user interface. It also highlights the importance of combining a log of user's actions and satisfaction measures, to better understand the user reaction to hypermedia. KeywordsÊ: Field dependence, cognitive styles, navigation strategies, information processing, computer-based instruction.
1. INTRODUCTION The field of computer technologies has exploded since we started to focus on human factors, in order to build more user-friendly interfaces. This widespread use of computers provides now new goals for designers. As more people use computers, the demand for quality interfaces increases. It is necessary to provide for usersÕ needs, expectations and preferences for different interaction styles and modalities. Now interface designers pay more and more attention to usersÕ characteristics and the way they interact with computer systems. There is a lot of effort being put in the area of task analysis, user profile, participatory and user-centered design. There are many excellent methods for learning more about users, and the general picture of interfaces has improved. Despite this, some applications are still too complex for many users and do not always fit their needs. As Gentner & Nielsen point out : "human interface is stuck" [10]. One way to develop interfaces a step further could be to study how individuals differ in their use of computers and to adapt according to these differences. Improvement of interfaces have always drawn from the studies of human errors, they may profit from solving problems of specific population. We are not yet able to effectively predict how users are going to use the software, especially in "open shop" environments like hypermedia that present a various and free access. Hypermedia has the potential of being adapted to user's strategies and cognitive style, but are often confusing and inefficient for fostering learning. As Gentner & Nielsen state, "direct manipulation" and "user control" are not well adapted to every situation [10]. As the complexity of applications grows, it is important to avoid the double learning problem (construct both a mental model of the system structure and the knowledge to be assimilated). The interface should be designed to adapt more to the usersÕ abilities and learning styles. Human cognition is one of the most complex areas of study. Much of the work concerned with human aspects of information technology systems is based on generic models for human information processing and little attention is paid to individual differences [23]. A possible avenue resides in system adaptation to cognitive styles, since they influence the use of computers and may explain some usability problems and behavior characteristics observed from users. We think it is important to observe learning strategies and their relation to cognitive styles of users. In this study, we have experimented with different levels of control of the system in order to facilitate interaction for different types of users. 2. COGNITIVE STYLES The study of cognitive styles as an individual difference is a prominent and extensively researched area in the human sciences. Research began in the early 1950Õs and has engaged research interest over the years. Cognitive styles are high-level heuristics that organize and control behavior across a wide variety of situations [19]. They are one of the most stable user characteristics overtime. They are consistent across a variety of situations, as opposed to user knowledge or experience that are more specific and evolving. Many researchers have shown the importance of cognitive styles in the area of HCI research and their implications in interface design [27]. Cognitive styles are important in determining the most effective
interface style for a particular category of user, especially in the formative stages of an interaction [8] [20]. The literature shows more than nineteen cognitive styles such as holistic & serialist strategy, convergent & divergent ways of thinking and field dependent & independent. Field dependence is the cognitive style associated with the most substantive research in the past thirty years. For a review of these styles, see Ausburn [2]. 3. FIELD DEPENDENCE Derived from the work of H. A. Witkin et al. on visual-spatial abilities, field dependence is related to the ability to abstract items in a specific context. It affects the way a person structures and processes information [29]. Field dependence is related to the tendency to be distracted by the perceptual field during information processing. The study of field dependence seems important in HCI, first, because it is associated with a valid measure instrument [8] [23], but also because it was shown to be related to human information processing and information restructuring [29] [28]. Though Chen et al. [4] found differently, it is generally thought that cognitive styles may explain why subjects have problems in the consultation of hypermedia documents. Cognitive Restructuring skills are central to the construct of field dependence. This could be defined as the level of capacity to organize items in a field which lack of inherent structure. Frank & Noble [9] mention that cognitive restructuring involves three operationsÊ: 1- Break up an organized field into its basic elements, 2- Provide structure for an ambiguous stimulus complex, 3- Provide a different organization to a field than that which is suggested from the inherent structure of the stimulus complex. This seems appropriate for all these acts since they involve making changes in the field, or "going beyond the information given", rather than following the field "as is" [30]. There are two modes of processing information: Field Independent (FI) people tend to have good analytical and cognitive restructuring skills. They will actively reorganize information according to contextual demands and impose structure when necessary according to their experience. They are likely to form a mental model of the situation before proceeding with their task. FI people seem to follow more easily a restructuring approach and use internal referents in other situations. Field Dependent (FD) people tend to adopt a passive approach in learning and problem solving. They prefer to be guided and to rely on external referents. Perception is dominated by the prevailing field. When internal referents are less available, FD people are more likely to respond to the dominant properties of the field as given. Lesser use of restructuring may handicap FD people in unstructured situations. FD people may need more explicit instructions in problem solving strategies or more exact definitions of performance outcome than FI, who may even perform better when allowed to develop their own strategies [30]. However, the restructuring process occurs only when the field lacks organization. When the material to be learned is presented in an already organized form, so that structuring is not particularly called for, FD and FI people are not likely to differ in their behavior and learning [31]. There is no apparent difference between FI and FD people if the external organization is clear and memory demands are minimal. Field dependence can be identified by the use of the "Embedded Figures Test" [29] which ask people to identify a specific embedded figure in a more complex one. 4. STUDIES OF FIELD DEPENDENCEÕS IMPACT ON HCIÊ: The effects of field dependence on HCI have been investigated in a few experimental studies. 4.1 Field dependence and its effects on learning UNIX Coventry [5] found differences in the knowledge and use of UNIX commands, according to field dependence. After making sure that users knew basic UNIX commands, the experimenter observed help requests. Subjects were free to ask the tutor to explain anything at any time. Results revealed that FD subjects asked for more help than FI subjects. FD subjects are thought to adopt a trial-and-error approach to learning rather than doing pre-planning. They do not actively seek extra information as a result and prefer a system that provides guidance and is well-determined. They learn each command and its syntax independently of other commands. They tend to have difficulty structuring the information that they assimilate. FD subjects were not aware of as many UNIX commands as FI subjects. FD subjects had not explored the system to the same extent than FI subjects, but the commands that they did know were used more accurately than the commands issued by FI subjects. In the experiment, FI subjects asked the tutor to confirm command names and the accuracy of procedures, but only after execution. FI subjects have a more flexible approach and develop a model of the system at an earlier stage. They adopt a strategy of active investigation and use more sophisticated learning strategies. 4.2 Field dependence and metacommunication Metacommunication refers to all activities aiming at communication about the conceptual model underlying the design of the system [21] [27]. These include navigation aids which support orientation in hypermedia documents and help to overcome the negative influence of disorientation These aids inform
the user about the current state of the system or provide reminders of system functions. They could be inside the system (help and prevention messages) or outside the system (documentation). In an "open shop" environment favorable to exploration and self-learning, how can the system support the users to help them construct an adequate mental model? It is clear that metacommunication form and content should be adapted to fit the various usersÕ needs, particularly their cognitive styles and learning strategies. In a study that presents the relation between system and user characteristics, Van der Veer et al. [27] have defined guidelines on the metacommunication structure according to field dependence, based on MoranÕs four interaction of command language grammar levels (task, semantic, syntactic, interaction). Van der Veer et al. point out that experienced FD users may have more problems with the learning of a new system. They did not tend to refer as easily to similar and more familiar systems. They may need extra help in transfer from a level to the next or a system to another by adequate help explaining the relations between MoranÕs levels. The study showed that FD users tend to develop a fixation on correct solution and so encounter difficulties in adapting to similar but different situations. FI users can overcome these difficulties without support. 5. HYPERMEDIA-BASED LEARNING HypermediaÕs potential for exploration and self-learning makes it interesting in education [11] [15]. Hypermedia gives users the opportunity to control their learning and to progress according to their abilities. This individualized instruction environment should be studied to see how they encourage navigation, learning strategies and information processing. Hypermedia makes it easy to freely access large amounts of information which may be accessed using different pathways. Access to the information is possible from different locations, not only in the sequential manner as traditional written documents. Hypermedia also has the potential to provide a linear access. However, hypermedia has drawbacksÊ: freedom is associated with cognitive overload, disorientation and distraction problems [11] [15] [18] [26]. This fluid environment puts the user in a situation where he or she must constantly decide where to go and evaluate where he or she is. Freedom requires a lot of involvement, such as how and when to change location? What are usersÕ goals in the interaction? This forces them to think rigorously. Although hypermedia lets the users control navigation, they must then construct both a mental model of the system structure and the knowledge to be assimilated (the double learning problem). Users must also take decisions on which path to follow and then evaluate their choices [16] [25]. McDonald & Stevenson [17] studied the effects of three hypertext topologies on navigation performance: hierarchic and nonlinear, compared to a linear version. Subjects used the document to answer 10 questions. After a distraction period, subjects returned to the document to locate five specified hypertext nodes. Speed and accuracy were measured and the subjects' own evaluation of their performance was assessed using a questionnaire. The results showed that subjects performed better with the linear text than with the nonlinear text, while performance on the hierarchical document fell between these two extremes. Analysis of the questionnaires confirmed these differences. 6. FIELD DEPENDENCE AND HYPERMEDIA-BASED LEARNING Hypermedia gives the freedom to explore and foster exploration and self-learning by adaptation to different types of interactions. In a study on hypermedia learning environments, Stanton & Stammers [25] have demonstrated that non-linear representation is superior to a linear one for knowledge transfer, because it allows greater individuality on the behalf of the learner. Individual differences are observed in the way users interact with hypermedia system. It is important to study how the organization of information and navigation aids may contribute to a better learning, in relation to individual differences. Liu & Reed [14] examined the different learning strategies of FI and FD people in a hypermedia-assisted instructional setting. Results from the study suggested that different learning styles groups employed different learning strategies in accomplishing the same task. FI learners used more navigation aids than FD learners. The FI learners used the index tool more often to move around in the courseware whereas the FD learners preferred to follow the sequence of the courseware. Although the learners had different learning styles, they performed equally well. These results indicate that hypermedia-assisted instruction can accommodate learners with different learning styles. The large amount of research on navigation problems in hypermedia environment makes it clear that these problems are not resolved yet. Concerning hypermedia tutorials, it seems that cognitive requirements and manipulation problems in hypermedia may have a negative impact on comprehension [18]. Even after the development of sophisticated navigation aids to facilitate interaction and lighten cognitive effort required by the user [21] [24] a doubt is cast upon the efficiency of these environments to support navigation and learning. We believe that a solution to these issues relies partly in the study of users' interaction and their individual strategies to cope with complexity. 7. THE EXPERIMENT We postulate that there is a relation between navigation problems encountered on hypermedia document and cognitive abilities of the user in terms of the restructuration of information, representation and
restructure, and the difference between FD and FI people on information processing in a less organized environment, we made the hypothesis that there would be a relation between cognitive style and the ability of subjects to learn in a less structured environment. In hypermedia environment, a more flexible access to information would leave the user to better organize it, making it more difficult for FD users. Since FD users tend to employ a passive approach to learning and prefer to be guided in the interaction, they would react negatively to non-linear hypermedia. On the other hand, FI users would react negatively if they had no freedom to discover and restructure information themselves. The aim of the study was to examine how the degree of linearity and constraint in a hypermedia environment was perceived by FD and FI people and how it was related to their understanding of the content. 8. METHOD 8.1 Task Description We developed a tutorial prototype that teaches how to build a database using Microsoft EXCELä, Version 2.0 for Macintosh (define database, define criteria, search and extract) using hypermedia explanations and EXCEL simulation for demonstration, exercices and tests. The system is in Hypercard and use a task and user model to give the user feedback and guidance during his interaction progression. [6]. Database's functions in EXCEL are difficult to assimilate in an exploratory manner. They involve planning and coordinating many operations following goals that are abstract to users. They require the execution of elementary actions, which have to be made in the right combinations to achieve a specific goal. For example, sub-actions "write", "select" and "choose an option in the Data menu" is required for each step in a specific order. No feedback is given until all sections are completed. This lack of "closure" makes interpretation of errors difficult. 8.2 System Interface The tutorial contains over 86 screens of instructions, examples and tests. It is built following a GOMS model of the task, instruction pages (23) showing procedures to follow to succeed a goal and sub-goals in the database construction task. The hypermedia hierarchy presents the information following the structure of goals and sub-goals. Instruction pages have historical cues (see metacommunication aids below) to help navigation and give access to examples, exercices and test pages. Instruction pages are the first to be consulted by the user, they may be seen going forward or backward, as users back up from examples, exercices and tests. Two hypermedia interface versions were designedÊ: Restricted access: Page access is controlled by the system which limits a consultation order according to the procedure. In the instruction pages of the restricted version, all buttons are visible, but only one is active going forward, depending on the user progression. As soon as the users have accessed all instruction pages of a specific section, the button for the next section becomes available. All sections already seen may be consulted freely. When all sections had been accessed, example, exercice and test buttons become available. This structure is close to a linear structure at first, though it highlights the hierarchy and offers hypermedia facility after a first introduction. Open accessÊ: This version allows users to select all pages without any constraints in the order they wish. All buttons are accessible at anytime. 8.3 Metacommunication Aids Both interfaces have the following aids for metacommunication : Historical cues were introduced to give users an intelligent feedback about their progression. This added feedback serves two purposesÊ: 1- it reduces users' memory load so they do not have to remember what has been seen and what is still left to do, 2- it helps motivate users to complete their learning. Buttons that give access to information could be in four states : unused (white), explored incompletely (gray), explored completely (black) or unavailable because prerequisites have not been seen (dotted). Buttons' states reflect progression according to global coherence. For example, instructions could be accessed through different paths (index, as a subtask of another goal). In such cases, the task model was used to change all buttons leading to an information [7]. Index page shows main sections and sub-sections of the tutorial in a hierarchical way and gives an overview of the entire tutorial structure and content. Index also provides intelligent footsteps so users could jump to uncompleted sections from the index. Index is available from all instruction pages. User Guide is located at the beginning of the interaction. It explains the meaning of buttons and footsteps states used in the tutorial. It may be accessed from the main page.
9. EXPERIMENTAL HYPOTHESES The hypothesis of the study was that there would be an interaction effect between users cognitive style and interface version on navigation strategies, learning and usersÕ satisfaction. Because FD people build up their mental model by a longer restructuring process, we predict that FD subjects in the open version would have a mean time per page higher than FI. Since FI people browse more easily in an unstructured
context, they should access a higher number of pages in the open version than FD people, who would be less motivated to explore the tutorial, and would suffer of more disorientation. We supposed that FD people would report more "the feeling of feeling lost" in the open version. Finally, they should go more often and should stay longer in the index page and user guide than FI. 10. PROCEDURE The study was conducted in five parts: 1-pre-test, 2-Embedded Figures Test /group assignment, 3experimental session using the tutorial, 4-performance tests (EXCEL assignment & interactive test), 5usability questionnaire. Subjects ProfileÊ: Fifty-five volunteered university students, taking a New Technology course took part in this study. The grouping of users into FI and FD was based upon the mean score (mean =11.83) and the standard deviation (SD=4.56) of the GEFT test [22]. There was an equal number of FD and FI subjects. Trace AnalysisÊ: A log of users' actions was installed in the tutorial in order to track the navigational sequence and timing. It showed which sections of the system users had gone and which messages the system had shown. It recorded the time in seconds at each step. Trace analysis is used to bring to light on possible links between aspects of the interface and the users' strategies. 11. RESULTS Table 2. Results on navigation strategies, satisfaction and performance by groups. Significance at the ANOVA test on the influence of cognitive style (C.S.), access structure (A.S.) and interaction effects, with the observed differences between Restricted (R) vs. Open (O) version and Field Independent (FI) vs. Field Dependent (FD) subjects. C. S. A.S. C.S. vs Interaction Cognitive Style FI FD A.S. Effects
Access structure TOTALÊ: n=55 Overall tutorial Total amount of time spent using the tutorial Mean time/page (sec) Instruction pages Mean time / page (sec) Total number of access Total number of come back Navigation Aids Mean time spent in the index (sec) Total time spent in the user guide (sec) Users impression Orientation Satisfaction Performance Total time spent on Interactive test * p£ < .05, ** p£ < .01
Restricted n=16
Open Restricted n=12 n=14
Open n=13
1:24
1:23
1:23
1:21
32
31
33
43
.006**
.085
.02*
O : FI < FD**
17.2 100 4.3
15.5 103 4.4
19.4 98.5 4.2
22.3 78.4 3.4
.0007** .047* .05*
.646 .206 .206
.078 .076 .08
FI < FD FI > FD FI > FD
14 5.25
29 3.5
12 4.5
20 5.25
.180 .437
.004** .446
.326 .05*
RFD O: FIO R>O
14.9
16.8
12.5
23.8
.272
.003**
.027*
OÊ: FI