Learner engagement with visualization technology

Following up on the meta-study by Hundhausen et al. (Hundhausen et al., 2002), and based on a review of experimental studies of visualization effectiveness, Naps, Röβling et al. (2003) defined the Engagement Taxonomy (ET) of six levels of learner engagement with visualization technology. This taxonomy is shown in Table 2.2. Their hypothesis is that higher levels of learner engagement produce more effective learning results.

Table 2.2 The Engagement Taxonomy (Naps, Röβling et al. 2003) 1. No viewing. There is no visualization to be viewed

2. Viewing The visualization is only looked at, without any other form of engagement. This is therefore a passive action, but required by all higher forms.

3. Responding Learners use the visualization to answer questions, possibly followed by more viewing.

4. Changing Learners modify the visualization, for example, by changing input values.

5. Constructing Learners create new visualizations of the algorithm either manually or by using a PV or AV tool, without necessarily coding it.

6. Presenting The visualization is presented for discussion and feedback to an audience.

Lauer (2008) refined both the viewing and constructing levels of the ET (Naps, Röβling et al., 2003) as described in Table 2.2 (Naps, Röβling et al., 2003). Viewing was subdivided into passive viewing and active viewing. In passive viewing learners have no control to rewind the animation and there are no fixed breakpoints after intermediate steps. In the category active viewing, users may return to earlier stages of the algorithm and/or stop the animation at break points provided, to better highlight intermediate steps. Similarly, the author subdivided the constructing category into constructive simulation and code-based simulation (Lauer, 2008). In the constructive simulation category, students

construct an animation by simulating an algorithm on a re-defined visualization, while in the code- based simulation category they code the algorithms using a visual tool. Based on the results of several other studies that investigated the relationship between learning result and levels of engagement, Lauer (2008) concludes that active viewing is on a higher level than passive viewing, and that changing and constructive simulation fits into the same level (level 4 in the ET (Naps, Röβling et al., 2003)), while code-based construction is on a higher level.

Both the ET (Naps, Röβling et al., 2003) and Lauer’s (2008) refinement thereof are based on work in AV research and the types of engagement found in AV tools. This research report’s focus is on PV. Since PV tools support other types of engagement, Myller, Bednarik, Sutinen and Ben-Ari (2009) extended the ET for PV. In particular, PV tools may provide opportunities to offer interactive input and modify the source code used for the visualization. The extended ET is shown in Table 2.3 below. Levels marked with an asterisk (*) are from the original ET, although some definitions were adapted slightly. Changing has been divided into two categories, changing and modifying. Three new categories have been added, namely controlled viewing, entering input, and reviewing. Some of these correspond to some extent with Lauer’s (2008) refinement in that controlled viewing is similar to active viewing, constructing to constructive modification, and modifying to constructive simulation. Table 2.3 The Extended Engagement Taxonomy (Myller et al., 2009)

No viewing (*)

There is no visualization to be viewed, but only material in textual format. For example, the students are reviewing the source code without modifying it or they are looking at the learning materials.

Viewing (*) The visualization is viewed with no interaction. For example, the students are looking at the visualization or the program output.

Controlled viewing The visualization is viewed and the students control the visualization, for example by selecting objects to inspect or by changing the speed of the animation.

Entering input The student enters input to a program or parameters to a method before or during their execution.

Responding (*) The visualization is accompanied by questions which are related to its content.

Changing (*) Changing of the visualization is allowed during the visualization, for instance, by direct manipulation.

Modifying Modification of the visualization is carried out before it is viewed, for example, by changing source code or an input set.

Constructing (*) The visualization is created interactively by the student by construction from components such as text and geometric shapes.

Presenting (*) Visualizations are presented and explained to others for feedback and discussion. Reviewing Visualizations are viewed for the purpose of providing comments, suggestions and

feedback on the visualization itself or on the program or algorithm.

Sorva, Karavirta & Malmi (2013) produced a comprehensive survey of PV systems for teaching novices about the run-time behaviour of programs. To assist with this process they extended the taxonomies proposed by Naps, Röβling et al. (2003) and Myller et al. (2009), in two dimensions to

reflect direct engagement with the visualization by the learner (vertical) and content ownership (horizontal). They call their taxonomy the 2DET Engagement Taxonomy.

Figure 2.2 shows the two dimensions in the 2DET taxonomy. Direct Engagement corresponds to a large degree with the taxonomies by Naps, Röβling et al. (2003) and Myller et al. (2009), while Content Ownership refers to the learner’s relationship with the target software. Table 2.4 describes the categories in the 2DET. The categories in the 2DET taxonomy as shown in Table 2.4 are self- explanatory. However, it should be noted that the category own content on the Content Ownership dimension, refers to using a PV tool to examine a learner’s own program, which differs from the creating level on the Direct Engagement level (Sorva, Karavirta & Malmi, 2013).