Attention guidance during example study via the model’s eye movements

Abstract

Research has shown that guiding students’ attention guides their thought, and that attention can be communicated via eye movements. Therefore, this study investigates whether such a procedure can further enhance the effectiveness of examples in which a solution procedure is demonstrated to students by a (expert) model. Students’ attention was guided by showing them not only the model’s problem-solving actions on the computer screen, but also the model’s eye movements while doing so. Interestingly, results show that combined with a verbal description of the thought process, this form of attention guidance had detrimental effects on learning. Consequences for further research on attention guidance and instructional design are discussed.

Introduction

Eye tracking research has shown that students’ attention can be guided on the basis of an expert’s eye movements. This can be done directly, that is, by showing students the expert’s eye movements (Velichkovsky, 1995) or indirectly, that is, by a cueing procedure based on the expert’s eye movements (Grant & Spivey, 2003). This study investigates whether showing the model’s eye movements in examples in which a solution procedure is demonstrated to students, can enhance their learning.

Eye tracking provides insight in the allocation of visual attention and therefore can provide very useful information on multimedia learning processes. However, in the field of learning and instruction, eye tracking has not been applied much as a research tool. When it was used, it was primarily in reading research (see Rayner, 1998, for a review), with some notable exceptions in other areas such as text and picture comprehension and problem solving (e.g., Hannus and Hyönä, 1999, Hegarty and Just, 1993, Verschaffel et al., 1992). This has changed over the last years. Eye tracking is applied more and more in research on instructional design. Especially in research using multimedia learning materials, that is, materials that consist of spoken or written text and pictorial information (e.g., picture, diagram, video, animation), or materials consisting of multiple representations (e.g., formula, diagram, graphic), eye tracking can provide unique information concerning what medium or representations are visually attended to, in what order, and for how long (see e.g., Holsanova et al., 2009, Louwerse et al., 2009, Schwonke et al., 2009).

However, eye tracking offers more than just a research tool. Modern eye tracking technology allows not only for recording of eye movements, but also for replaying this record integrated with other actions such as mouse and/or keyboard operations visible on the PC screen (cf. Van Gog, Paas, Van Merriënboer, & Witte, 2005). This provides the opportunity to apply it as a tool to enhance learning processes, by using eye movement records in the design of instruction. For example, by developing modelling examples in which the model’s performance as well as his/her eye movements during performance, are captured and replayed to students.

A large body of research has demonstrated that for novices, engaging in problem solving is not an effective way to acquire problem-solving skills. In the initial phases of cognitive skill acquisition, it is far more effective and efficient to study a good problem solution (this is known as the ‘worked example effect’; for overviews, see Atkinson et al., 2000, Paas and Van Gog, 2006, Renkl, 2005, Sweller, 2006, Sweller et al., 1998). Studying a good problem solution can be obtained by different means. It can take the form of modelling examples, in which a solution procedure is demonstrated to students by a model who is often an expert or an advanced peer (e.g., Braaksma et al., 2004, Kitsantas et al., 2000), which play an important role in Bandura’s (1977) social learning theory. It can also take the form of worked-out examples, in which students are given a written account of a model’s solution procedure to study (e.g., Carroll, 1994, Cooper and Sweller, 1987; Paas, 1992; Paas and Van Merriënboer, 1994, Sweller and Cooper, 1985, Van Gog et al., 2006), which play an important role in cognitive load theory. In this line of research, the ‘model’ is often a didactically behaving expert, that is, the examples contain an ‘ideal’ solution procedure detailing how students should learn to solve a problem, rather than a reflection of a “naturally” behaving expert’s solution procedure (as experts have automated procedures and are likely to skip certain steps). Finally, with current technology, animated models can be created in which the solution procedure is not demonstrated by the model or written out, but is provided in the form of an animation (Wouters, Paas, & Van Merriënboer, 2008).

Example-based learning as applied here in this study combines elements of the worked examples and the modelling examples tradition, in the sense that a solution is demonstrated to students, as in modelling examples, but the expert is not visible (only the actions s/he does are visible on the computer screen) and more importantly, the expert is behaving didactically, in other words, is performing the task not as s/he would normally do, but as the student should learn to perform it. As such, the situation is a bit less social as it normally would be in modelling examples and the examples resemble worked-out examples in the sense that an ‘ideal’ solution is shown that could also have been written out for the student to study, although this would have been rather impractical.

Cognitive load theory (Sweller, 1988, Sweller et al., 1998, Van Merriënboer and Sweller, 2005) explains the effectiveness of worked examples in terms of reduced extraneous, or ineffective cognitive load during training. The theory distinguishes cognitive load inherent to the task, and cognitive load imposed by the instructional design. The former is called intrinsic cognitive load, and results from the number of interacting elements in a task. The latter is called extraneous cognitive load when it is ineffective for learning and germane cognitive load when it is effective for learning. The aim of instructional designers should be to minimize extraneous and enhance germane load imposed by instruction (Sweller et al., 1998).

Instructional measures that successfully induce a germane load, stimulate learners to invest mental effort in the development of rich cognitive schemata during training, that subsequently allow for effective and efficient test performance. Examples of instructional measures that are known to induce a germane cognitive load in studying worked examples are for example increasing the variability (Paas & Van Merriënboer, 1994) or contextual interference (Van Merriënboer, Schuurman, De Croock, & Paas, 2002) in series of worked examples, or prompting students to self-explain the rationale behind solution steps (Chi et al., 1989, Renkl, 1997).

However, even though self-explaining can be beneficial for learning, this is not necessarily the case for all learners; it rather depends on whether they are able to self-explain, as well as on the quality of their self-explanations (Chi et al., 1989, Renkl, 1997). Nevertheless, typical examples usually require some self-explaining, as they show students only the steps required to reach the solution (i.e., the end product; these can therefore be called product-oriented worked examples; Van Gog et al., 2006, Van Gog et al., 2008). However, these solution steps are consequences of inner thought and attention processes, and these processes are usually not made explicit in the examples. That is, product-oriented examples only show what steps are taken, not how these steps are selected (e.g., is this the only option or are there different options?) or why they were selected (e.g., why is this option better – more likely to lead to a good solution – than the other one?). Not presenting this additional information can compromise students’ learning and understanding of the solution procedure (unless they are able to self-explain this correctly). Understanding a solution procedure is required to be able to flexibly apply learned procedures in novel situations (i.e., transfer; Detterman & Sternberg, 1993). Process-oriented examples (Van Gog et al., 2006, Van Gog et al., 2008) do provide additional information on the rationale that led to the solution procedure.

Research has shown that for novices, process-oriented examples that made the rationale behind the worked-out solution procedure explicit were more effective initially for learning and transfer than product-oriented examples (Van Gog et al., 2008). As yet, however, no studies have been conducted on the effects of making the model’s (visual) attention processes explicit in examples. This might be relevant, especially in the context of computer-based modelling examples, where solution steps are not written out, but are demonstrated to the learner by the model (e.g., the learner is shown a screen capture of a model’s task performance). Bandura (1977) has stressed the fact that “people cannot learn much by observation unless they attend to, and perceive accurately, the significant features of the modelled behaviour” (p. 24). However, eye tracking research on expertise differences has shown that the allocation of visual attention between the model providing the example, and the novice studying the example, is likely to differ.

Attention can shift in response to exogenous and endogenous cues (Rayner, 1998, Stelmach et al., 1997). Exogeneous shifts occur mainly in response to salient features in the environment, whereas endogenous shifts are driven by knowledge of the task, of the environment, and of the importance of information sources. In other words, endogenous attention shifts are influenced by expertise (see e.g., Charness et al., 2001, Haider and Frensch, 1999, Underwood et al., 2003). Research has shown that individuals with higher expertise are likely to allocate their (visual) attention faster and/or in greater proportion to relevant information in the task (see e.g., Charness et al., 2001, Haider and Frensch, 1999, Van Gog et al., 2005). This implies that the expert model performing the task and the student observing the performance, may not be attending to the same information. This could be problematic, as students may miss information necessary to understand what the model is doing, or why s/he is doing what s/he is doing, in particular when information is transient (there one moment and gone the next). Being able to see where the model is looking at what moment, may help students to better process the example. For example, depending on the type of task, the model’s eye movements may also make the strategy s/he is using (e.g., different options under consideration) clear to students.

As mentioned above, research by Velichkovsky (1995) on cooperative puzzle problem solving by expert-novice pairs, in which eye movements were used to demonstrate on what task aspects the partners focused their attention, has shown that attention can be guided by showing students an expert’s eye movements. In this study, the novice could control the mouse to solve the problem, and the expert indicated with his gaze what the novice should do. The question is, however, whether attention guidance can not only be used to enhance communication as in Velichkovsky’s study, but whether such guidance can also enhance learning outcomes, for example when applied while studying modelling examples. The study by Grant and Spivey (2003) suggested this might be the case. Based on eye movement patterns of participants who could successfully solve Duncker’s radiation problem in one experiment, Grant and Spivey developed a cueing procedure (visual highlighting) implemented in a second experiment, which enhanced learning outcomes. However, Duncker’s radiation problem is an insight problem. Modelling examples typically focus on problems that require a series of solution steps, that is, are dynamic and involve steady progression towards the goal state. Thus, the question remains whether attention guidance could also be effective in studying modelling examples to enhance learning, and whether eye movements could be used directly to guide attention, rather than being translated into a cueing procedure.

Therefore, this study addresses the question of whether attention guidance by showing students the model’s eye movements can enhance their learning in combination with product-oriented examples that show only the solution steps, and process-oriented examples that show the solution steps and contain a verbal explanation of why these steps are taken. It is hypothesized that this form of attention guidance would help learners to better attend to and encode the relevant features of the examples, which would contribute to learning. This effect is expected to be larger in combination with process-oriented worked examples, that convey both thought and attention processes to the learner.