The recent move to online learning prompted academics teaching business analytics to find ways of explaining concepts that maintained some of the advantages of face-to-face teaching and learning. This included inventive ways of rigging up iPads and WACOM tablets to serve as digital whiteboards – to demonstrate analytical tools or to step through or annotate examples of how to apply formulas to solve business problems.
In the process, teachers came across tools they may not have considered previously, and started to think about how such tools might provide new ways of promoting interaction with content in the discipline. On this blog site, we recently shared an example of how combining two applications, Python and Ed, was used to provide programming tasks in tutorials where student progress could be monitored in real-time and lecturers could jump into the student’s challenge and type alongside them in a google doc.
One of the biggest challenges in creating online content for students has been to find ways of supporting learning that is not just interactive, but dynamic. In a first-year unit on Data Science in the Master of Commerce program at The University of Sydney Business School, a small team has been experimenting with a variety of ways to engage students online. In this post we share some of those experiments. But first, let’s consider what we mean by ‘dynamic’…
Dynamic Representations of Thought
Interface designer, computer scientists and electrical engineer, Bret Victor, has spent many years thinking about and supporting new representations of thought. While many representations, such as written text, mathematical notation and information graphics, “have been responsible for some of the most significant leaps in the progress of civilisation”, they have been invented for static media. Victor argues that as a result, they only “tap into a small subset of human capabilities and neglect the rest” (Victor, 2014). We know however, that there are other important modes of thinking and understanding which are incompatible with static media.
“the dynamic medium offers the opportunity to deliberately invent a humane and empowering form of knowledge work. We can design representations which draw on the entire range of human capabilities – all senses, all forms of movement, all forms of understanding – instead of straining a few and atrophying the rest.”
Victor, 2014
Dynamic pictures are one way of helping students understand a system, and how the parameters of that system, when manipulated, impact on the system as a whole. While we know that visual explanations are vital for helping us understand many concepts, we often “resort to describing when we should be depicting” (Victor, 2011). Static images, or “one-off pictures” can be limiting. Dynamic pictures on the other hand are “ideal for visual explanations, because the parameters can represent information to be conveyed. As the information changes, so does the picture” (Victor, 2011).
Below we introduce you to a range of interactive tools we’ve been using in online modules to promote information engagement in our Data Science unit. You will see that while some of our earlier tools provided ways for students to interact with content, more recent designs have started to move towards supporting more dynamic representations of thought.
Interactivity: towards more dynamic representations
Our approach to the design of this unit of study took into account the selection of tools that could make the content engaging, interactive, and interesting to students. This goal was easily achieved using Genially, H5P, Padlet, and Atomic Discussion. Essentially, these digital tools are used to boost student engagement, reinforce learning, enhance the learning experience, personalise and sustain learning, and simplify concepts to facilitate hands-on learning.
The first tool we’ll discuss is Genially. This tool can be used to transform text-based illustrations, descriptions, and complex ideas into visually engaging and interactive diagrams with minimal cognitive load. In this unit, it was specifically used to visually illustrate model processes and principles, align multiple perspectives, reinforce and revise concepts, and so on. It was also used to condense text-heavy pages into visuals, diagrams, and interactive activities. For instance, a text-heavy description of the six-phase CRISP-DM process model was transformed into a simple, engaging and interactive diagram. This visually boosts the learning experience and reinforces the stages of the model process and how each stage connects. Below is a visual depiction of the model process.
Padlet and Atomic discussion tools were also utilised to facilitate collaboration and student engagement with the unit’s material. These tools were used to increase students’ interaction through sharing information. Students could share their ideas and like or comment on the posts of others on these platforms. In our unit design, we used these tools to engage students in brainstorming on a given reading material. The Airbnb articles assigned to students as additional reading were transformed into activity-based reading where students were to read, respond to, and share their thoughts on a Padlet board or Atomic discussion based on a question prompt. The showcase of thoughts on these tools becomes a motivation for other students to read and participate. Also, students then become active contributors to learning rather than just active consumers of information when sharing ideas on a given material.
Another interactive tool used in this unit was H5P. The content types in the tool like multiple-choice questions, drag-and-drop, documentation, and dialogue cards provided immediate feedback and reinforced students’ learning. Considering the big size of our student cohort (>1000), this tool was effective in personalising and making learning engaging to everyone because of the instant feedback they received on the activity. For example, to highlight some key concepts of studies, the H5P drag and drop activity was set up for students to check their understanding. Students, after providing their answers, received instant feedback about the correct and incorrect responses and further explanation about their choice of answers. This makes learning very satisfactory, personalised and interactive for students. More so, the H5P documentation tool was set up for students to create a reflection portfolio on this unit by compiling and exporting weekly reflections based on some guided questions for future study and reference.
So, inspired by the level of interactivity in the unit design, we recognised the need to transform static graphs into interactives, and even better, dynamic graphs that could allow students to manipulate variables in a chart. This insight led us to plotly, which is a Python library that allows for the efficient creation of interactive charts.
With the assistance of the Unit Coordinator (Stephen Tierney), two interactive line graphs were created to demonstrate and explore the concept of Maximum Likelihood Estimation. The first graph accompanied a written example and students were able to drag the slider and interactively confirm the results of the example. The graph is shown below.
BUSS6002 interactive example: Optimisation – Step 3: Finding the maximum likelihood
The large black dot on the line turns red when the correct coordinates on the graph have been reached. The parameter value (x-axis) and the likelihood value (y-axis) can then be checked by the student against the example.
The second graph was used as part of an exercise where students were asked to solve a Maximum Likelihood Estimation problem. By manipulating the graph students could solve the problem and then share their answers with others in a separate comment box.
BUSS6002 interactive example: Optimisation – Apply knowledge
Where to next?
Throughout the semester a number of static graphs were presented to students, which are opportunities for greater engagement and understanding. For example in our module on Regression we displayed how the nature of random sampling leads to inaccuracy of estimates. These graphs could be replaced with dynamic versions that allow students to control the sample size and number of samples to observe the relationship between these parameters and accuracy.
Our current approach with plotly requires that the visualisations are pre-computed. This means that we cannot for example, generate random samples on the fly. In future, however, we hope to leverage advancements in technology such as PyScript, which enables Python to be run in a web browser and allow an unprecedented degree of dynamic interaction.
Featured image by Markus Spiske on Unsplash
About Authors
Stephen Tierney
Stephen is a lecturer in Business Analytics at The University Sydney Business School. His research interests include machine learning, computer vision, image processing, data visualisation and recommendation systems.
Enosh Yeboah
Enosh is an Assistant Learning Designer at University of Sydney passionate education sector professional with 10-years’ experience in diverse roles (including teaching, education support and, most recently, learning design), all underpinned by dedication to student engagement, inclusion, accessibility, and understanding.
Stephanie Wilson
Stephanie is a Senior Lecturer and Deputy Director (CLaS) with the Business Co-design team at Sydney University and Senior Fellow of the Higher Education Academy (SFHEA). She enjoys working with others to explore new approaches to learning and teaching inspired by design practice and the arts.