Review by: Dreier, Matthias (2005-07-27)

Purvi Saraiya examined pedagogically effective features of algorithm visualisations in her master thesis at Virginia Tech. In this paper she and her co-authors present the results of two experiments with heap sort visualisations.

Algorithm visualisations and animations are widely used in undergraduate computer science courses. It is commonly believed that visualisations and animations support the learning of algorithms. However, apart from the Richard Mayer’s Dual Coding Theory little is known about the way the learning is affected by such visual aids.

In order to identify basic pedagogically useful features of algorithm visualisations the authors first conducted an expert study to collect a list of features used for heap sort visualisations. The features were:

• the ability for users to enter their own data sets
• an example data set provided by the visualisation
• a pseudo-code display
• a back-button to go back any number of steps of the algorithm
• a guide which presented a series of questions related to the content

Experiment 1 with 66 participants showed some counter-intuitive results. Visualisations with more features yielded poorer results, though not significantly. Therefore, the authors concentrated on three features for the second experiment: allowing the user to step through the algorithm manually, providing an example data set, and pseudo-code display. The ability to step through an algorithm manually and the availability of a good example data set resulted in significantly better student performance. The pseudo-code display failed to provide any significant result.

The results suggest that especially the control of the animation pace and having a good example data set increase the pedagogical value of an algorithm visualisation. Neither the question guide nor the pseudo-code display provided any significant results, not even in pseudo-code related post-test questions. However, the time a student spends learning an algorithm is affected by the features. Students provided with the pseudo-code spent twice as much time than the others but did not perform any better. This means that additional features may not only impact the effectiveness but also the efficiency of learning.

The authors showed that some features of algorithm visualisations are significantly more effective than others and that providing too many features decreases the efficiency. The article covers other questions of visualisations too, for example the usability of visualisations and assessing of the students’ knowledge of algorithms. Therefore the article is worth reading for everyone who is involved in animating and visualising algorithms.