Dashboards in educational contexts are usually comprised of visualizations of educational data. For example, dashboards inside the LMS are often comprised of line charts, bar graphs, and pie charts of data like logins and quiz scores.
The primary goal of educational dashboard design is, ostensibly, for them to be “actionable.” In other words, a teacher, student, or administrator should be able to take an action after spending time interpreting a dashboard. For that to be possible, three questions must be answered in the affirmative.
- Does the visualization communicate something useful? (design relevance)
- Can faculty figure out what that useful bit of information is after studying the visualization? (data literacy)
- Can faculty figure out what to do based on that useful bit of information? (instructional design literacy)
Let me use course design as a metaphor for educational dashboard design. All too often, course design starts with selecting a textbook. The course syllabus is then designed around the contents of the book (week 1 = chapter 1). Hopefully, the textbook comes with some homework problems, a quiz bank, or some other assignments that can be used to assess student learning. These also get scheduled on the syllabus. If it’s an accreditation year, faculty may need to reverse engineer the contents of their book and the assignments they’ve chosen to come up with some “learning outcomes” they can list on the syllabus.
This sequence of course design decisions is so common that the more appropriate way of designing a course is popularly referred to as “backward design.” Backward design begins by deciding what students should learn (choosing learning outcomes), then deciding what evidence they would have to present for you to believe they had learned it (designing assessments aligned with your learning outcomes), and then choosing activities for students to engage in that will support their learning (like reading a chapter, watching a video, participating in a debate, etc. aligned with your learning outcomes). “Backward design” begins from the goal rather than ending with it.
Similarly, dashboard design frequently begins with the question “what data are available to be visualized?” Once that question is answered, the discussion moves on to questions like “what are reasonable ways to visualize those data? Which would be easiest to understand? Which would be most beautiful?” Only rarely – if ever – is the question asked, “what do we hope a teacher will actually do after looking at this dashboard?” We seldom begin with the end in mind.
Even when educational dashboard design begins with the end in mind, the effective use of dashboards still depends entirely on faculty’s levels of data literacy and instructional design literacy. Believe it or not, not all faculty are comfortable interpreting graphs. There’s no guarantee that even a majority of faculty can read “through” a visualization to grasp the useful information it’s trying to convey. And while all faculty will have had at least one math or quantitative literacy course, the overwhelming majority of them will have never received any instructional design training. This means they’re unlikely to be familiar with research suggesting which responses might be most effective in the context of whatever useful information they’ve gleaned from the dashboard.
What is an “Action Dashboard”?
Rather than making dashboards that are “actionable,” why not make dashboards of actions? In other words, what if we didn’t convert raw data into visualizations that we hope faculty have sufficient data literacy to interpret correctly, and then further hope that they also have the ID literacy necessary to do something useful with that information? What if dashboard designers applied the necessary data literacy and ID literacy upstream, in the design of the dashboard itself, and simply presented faculty with a list of specific actions they might consider taking? Here’s one example, presented two ways:
Current model: (1) Show a line graph of course activity.
(2) Hope that faculty can interpret it accurately. (3) Hope that faculty understand that a student not logging in regularly could indicate that they’re having some kind of trouble. (Yes, I realize that’s not what the data in my sample visualization show. But work with me here.) (4) Hope faculty decide that a reasonable thing to do would be to reach out to students who haven’t logged in for a while and check on how they’re doing.
New model: Create a list of students who faculty should check in on because they haven’t logged in for a while.
|Contact These Students|
|David Wiley||email / dismiss suggestion|
|David hasn’t logged in for 14 days. Consider checking in today to see how they’re doing.|
|Elaine Wiley||email / dismiss suggestion|
|Elaine hasn’t logged in for 8 days. Consider checking in soon to see how they’re doing.|
What information should be included in an action dashboard?
Everyone knows I’m a fan of five letter frameworks! The 5Ws you learned in school provide a great starting place for thinking about action dashboards.
- What? Describe the specific action should faculty consider taking.
- Why? Explain why they should consider taking it.
- When? How soon should they decide whether or not to take action?
- Who? Name each student with regard to whom the faculty should consider taking action.
- Where? In what format, with what tools, or in what place should faculty take the action?
From my personal perspective (and based on the research on the impact of teacher-student relationships), I think action dashboards should be filled with suggested actions that help faculty proactively express care, support, and encouragement for their students. An action dashboard might suggest faculty Contact These Students because they haven’t logged in for over a week (“Everything ok?”), because they scored above 95% on a recent quiz (“You’re crushing it!”), because they’re struggling in class (“Could you come to office hours so I can help you?”) and for a range of other reasons. If these messages were templated, faculty could send more messages of care, support, and encouragement to more students more quickly.
An action dashboard might also suggest that faculty Focus on These Topics During Class based on questions a majority of students missed on the homework. Or that faculty Review These Topics During Class to help students implement spaced review of critical concepts. Or other things.
Students, Administrators, and Power Users
There is nothing in this discussion of action dashboards from the faculty perspective that doesn’t also apply to students or administrators. It is completely unrealistic to expect them to have all the expertise necessary to translate a dashboard full of visualizations into effective actions they can take in support of student learning. Students and administrators need action dashboards, too.
It’s also true that, for faculty, students, and administrators who do have higher levels of data literacy and instructional design literacy, access to the visualizations (or even raw data) could be more powerful than an action dashboard. And I would advocate for providing those in an “advanced” view of the dashboard. My main argument here has been that, if we want dashboards to be widely used so they can improve outcomes for more students, the default view shouldn’t require high levels of data literacy or instructional design literacy in order to be usable. The default view should be immediately usable by everyone.
(It’s a bit like a Mac – if you have a high degree of computer literacy, you can open the Terminal app and type Unix commands to your heart’s content. But that shouldn’t be the default user experience. The default UX should be – and is – point and click.)
An “action dashboard” is a dashboard filled with specific actions a user might consider taking, presented in the context of the 5Ws. Action dashboards can be used effectively by all faculty, students, and administrators, regardless of their levels of data literacy or instructional design literacy. While action dashboards can feel restrictive to power users, power users can be provided with different views more suited to their literacy levels.
This is an updated version of a post that originally appeared on opencontent.org.