|
|
This paper investigates what aspects of a pupil's interaction with educational software are determinants of their learning. The work reported here considers whether the computer interface can be designed to encourage people to plan, to think more deeply about relevant information, and hence to learn more successfully. Findings reported here challenge the universal welcome given to graphical user interfaces. A number of pedagogical issues involved in designing educational software are raised. These suggest that designing with considerations other than ease-of-use is paramount.
The contention is this: The way that the learner physically manipulates objects on-screen will affect what they are mentally attending to. If the learner processes relevant domain information in the right way, this will enhance concept formation, collaborative dialogue and planning. The extent to which a particular interface manipulation style encourages learners to process relevant concepts will be used as a yardstick to assess its effectiveness. Support for this perspective comes from Larkin & Simon [2]. Studies of students using diagrams to understand pulley-systems led them to state the need for an "attentional control system" so that students can make best use of a diagram's features. Van Lehn [6] found that the way in which physics examples were processed is crucial for learning; students who gave concurrent self-explanations while studying example problems performed better, and persisted with more effective learning compared with controls. Self-explanations involved greater effort at the time, but benefited students in the long term. Learning is dependent on the depth to which information is processed and simply presenting the right information is insufficient; key concepts and relationships must be made salient.
The universal welcome given to direct manipulation graphical user interfaces is inappropriate for educational systems. Direct manipulation makes interaction a pleasant experience, with easily recognisable objects moved around using familiar pointing movements. It minimises both the effort needed to learn how to use the interface and the effort needed for moment-to-moment user actions. However, the literature indicates that there are manipulation characteristics which go beyond transparency and ease-of-use. The current research analyses the cognitive demands imposed on users by an interaction style. Manipulation costs will be beneficial to learning when they induce cognitive mechanisms, behaviours and strategies which are needed for expert-like problem solving.
The present evaluation study was a controlled comparison of two manipulation styles in terms of learning effects. Undergraduates did identical tasks, but one interface required the learner to select bears and transport from a separate radio-button screen, away from the map, while the other allowed moves to be made on the map itself by dragging objects around. The first of these was identified as more costly to the user since actions were a more lengthy diversion. User actions were automatically recorded during training and transfer, giving quantified measures of search, planning, proportion of legal moves made and overall task success.
Study 1 identified a whole range of manipulation costs that are not dichotomised by the traditional command versus direct manipulation distinction. Each of these costs have pedagogical significance because they encourage the learner to engage in beneficial problem-solving behaviours: Successful route-planners need to gain understanding of domain constraints. To do this they need to notice the outcomes of their actions, remember and evaluate action sequences which have led to any current configuration and finally they need to plan action sequences that will lead to desirable configurations. Hence constructive interaction costs for this system were those that encouraged the learner to focus on and process each of these kinds of event. So, for example, information located away from the main screen caused subjects to seek items of information directly, at times when the knowledge would be of significance to them. Secondly, lack of visual access to object configuration on the map during move-taking encouraged subjects to make the effort to remember the screen configuration for themselves. This meant that they were in a better position to evaluate configurations already tried, and to design new, more effective courses of action. Thirdly, when more keystrokes were required to make a move, or to recover from an undesirable move-state, subjects were encouraged to reflect longer on domain constraints before making moves, in order to avoid wasted effort.
The first of these allows interactive manipulation of graphical representations for perfectly elastic collisions. The user can try out different configurations of the representation, and request a simulation of the collisions for each configuration. This system, written in CT to run on Macintoshes, was developed by Cheng [1] and the actual choice of graphical representations was based on a set of principles outlined by Cheng. Studies have shown that undergraduate subjects learn from this system and they develop expert-like problem-solving behaviours.
Three variations of the present direct manipulation style are being examined for this system based on the three law-obeying constraints that govern the behaviour of the representation. Each variation allows the user to manipulate one of these constraints at a time. It is expected that when learners use each of the styles in turn, their attention will be drawn to the domain constraints more effectively than using the original manipulation style. Threefold predictions are made about student progress using these styles: improvement in pencil and paper tests, distinct strategies for exploration of the system, and a change in the nature of the dialogue between pairs of undergraduates using the system. The manipulation styles will provoke students to ask different questions, to form different hypotheses, each of which are constructive for learning.
The second system, allows interactive exploration of the behaviour of a balance-beam. It is written in SuperCard to run on Macintosh computers, and will be used with 10 -12 year old children. The pupil can try out configurations of weights and distances, and request a simulation of the tip. Direct manipulation of weights will be contrasted with a style of interaction where the student fills out a table of values in order to configure the beam. Formal informational equivalence is maintained by all pupils having visual access to the table, but only one interface uses the table as a manipulation tool. It is expected that pupils using the table-style interaction will learn more effectively, and will be more able to verbalise their findings. The role of collaborations will be analysed; it is expected that the table-style will promote more domain-centred dialogues.
The framework adopted for this research will give design recommendations for both command and direct manipulation styles. The interface should facilitate task focus which involves more than accepting the direct manipulation interface as a universal best. Instead, we should broaden our notion of the user's task beyond Human Computer Interactions's traditional concern with routine performance to consider how manipulation affects learning processes and the nature of collaboration.
1. Cheng, P.C.H. (1994) An empirical investigation of law encoding diagrams for instruction. In Proceedings of the 16th Annual Conference of the Cognitive Science Society (pp.171-176). Hillsdale, New Jersey: Lawrence Erlbaum Associates.
2. Larkin, J. H., & Simon, H. A. (1987) Why a Diagram is (Sometimes) Worth Ten Thousand Words Cognitive Science 11: pp. 65-99
3. Littleton, K., Light, P., Joiner., Messer, D., & Barnes, P. (1994) Gender and Software Interactions in Children's' Computer-based Problem Solving Technical Report 17, Psychology Department, Nottingham University
4. O'Hara, K., and Payne, S. J. (1994) Cost of Operations Affects Planfulness of Actions, School of Psychology, University of Wales, College of Cardiff
5. Svendsen, G. B. (1991) Influences of interface style on problem-solving. International Journal of Man-Machine Studies 35: 379-397
6. Van Lehn (1992) A Model of the Self-Explanation Effect. The Journal of the learning Sciences 2(1): 1-59