Online Learning Environments: Design in the Age of Big Data

An evolution in learning and teaching has been enabled by providing access to online, content rich, interactive, personalized, meaningful, and timely tools through online courseware, learning management systems and Massive Open Online Courses (MOOCs).

The Key Challenge for Online Learning

Engaging students has been a major challenge for online learning. Data shows that only a fraction of online course enrollees actually begin their courses; an even smaller number finish them. (See Figure 1 and The MOOC Revolution That Wasn’t for more details.)

Graph of statistics: 35% never engaged, 56% viewed less than half course content, 4% explored more than half, 5% earned certificates

Figure 1. In the first year of HarvardX and MITx, only 5% of the students completed the course. (Credit Ho, Reich, et al, HarvardX and MITx: The first year of open online courses 2014)

Teachers we have interviewed report a related behavior with their students in that a significant portion (about 30%) of students in a class cannot be depended upon to have reviewed online material on their own and return to class “prepared” and with questions about the material. Teacher time is then spent “catching them up,” delaying the rest of students.

Some of these engagement challenges might be rooted in the experience offered by online learning systems. A “paradox” in this experience, pointed out by critics, is worth mentioning: while such systems are touted as offering content more “interactive” in comparison to older forms of content distribution (for example, books), often the content is found to be less interactive in an important way that arguably facilitates learning. For example, if a student has a question about a concept in a traditional classroom, the teacher might attempt to discuss the concept in a different way. In contrast, in today’s online learning systems, when a student doesn’t understand a concept, the only immediate recourse has been to replay the content. Thus, it is noted that repeating it does not necessarily make it more understandable.

Evolving Online Learning Environments

The engagement and participation challenges faced by existing online learning environments call for evolution of such systems. The experiences need to be enriched through personalization, leveraging of analytics, and more spontaneous experiences more aligned with methods users have been evolving over the past 15 years to learn from networked content—the most obvious example being the wealth of content on the Internet.

What follows is a set of concepts for which we have prototyped experiences to address said challenges.

Real-time data

Real-time data is the anchor point for our prototype. Data drives and triggers key experiences. The screens in Figure 2 illustrate how a student progresses through course material, machine learning algorithms can evaluate metrics, such as longer time, touches, verbal responses and affective signs and other interactions to identify and offer relevant related content from both within and outside of the course to inform on how they are doing, enrich their study, or to present similar material in a different way.

Two screens. On the left, a reading about a silver coin that is a portrait of Julius Caesar. On the right, the self-evaluation appears next to the text. It is multiple choice with 4 choices.

Figure 2. Students can trigger a self-evaluation at any time

Improved personalization

During a digital learning experience, a student will leave traces of his or her progress and success—aka data. The analysis of this data can be further leveraged to automatically tailor a student’s experience on the fly. The more relevant to each individual student, the more engaged they can become. The prototype supports and encourages students exploring parallel paths supplementing the one proscribed by uploaded course materials.

“You don’t understand anything until you learn it more than one way.”
–Marvin Minsky, legendary AI scientist and Turing Award winner

From Book to Network: Multimodal Learning

Course content placed online need not remain isolated from related materials. Existing systems are designed to enable teachers to upload materials such as articles and books, but there are few if any provisions to enable uploaded materials and their constituent components (for example, paragraphs, images, graphics, chapters, notes) to relate or connect with each other. One can sketch out a hierarchy of relatedness:

  • Within-document
  • Cross-documents
  • Within-course
  • Cross-courses

External systems (for instance, YouTube, EdX, Moodle)

As material is added to the course, the system, using tracking data, can surface suggested content, supporting multimodal learning. In addition, fellow students can suggest useful materials that are also displayed (see Figure 3).

A list of links appears in the right-hand column.

Figure 3. Uploaded course material appears on the left side. The right two-thirds features an offering of supplementary material managed using the proposed hierarchy.

From Walled Garden to Permeable Environment

Most existing learning systems have been designed (purposefully or incidentally) to effect a cloistered model. This is changing with the ability of systems such as EdX and Moodle to share and export content to other systems. The prototype shows how to take advantage of a more “permeable” model.

Bite-sized content

Most existing course material tends to be adapted from existing books, requiring sustained attention spans and strict, linear paths. Online courses require designing with more digestible, bite-sized content in mind so that it can be assembled and presented more flexibly. The prototype supports this additional requirement by integrating designing for publishing (teacher) and consumption (learner) of componentized content.

Visualized progression

As students accept (or reject) recommended content into their individual paths, they create an individualized experience that offers students a way to visualize this progress. Visualizing their progress, students can reflect on not only their success but also on where more effort is recommended. We demonstrate how this can be achieved (see Figure 4).

Screen shot showing how these elements are displayed.

Figure 4. The student journey is tracked by activity, showing the time spent and results of quick evaluations.

Social channels

In today’s classrooms, social media is a reality and educators have to acknowledge social media as a communication channel and operational tool.

Accordingly, we should experiment with how to best integrate and leverage social media with online learning. While recognizing that some systems already include this experience, our research suggests that there is opportunity to evolve, improve, and further leverage to enhance learning via social media. For example: integrating Facebook messenger to enable peer-to-peer communication between learners, teachers, and subject matter experts.

Principles of Designing for Learning Environments

Talk with the users

This is a basic UX requirement, but with the evolving education and technology landscape it bears reiterating that a clear focus on who you are building the products for is vital—the users (students, teachers, administrators, web designers, and system administrators). Understanding what tasks users in the pedagogical space are trying to complete, the pains they have in doing so, and how your product can answer their needs is critical. This principle will not only create value for users and clarity around what your product is meant to be, but will also build confidence toward the product and brand across users.

Fit the model to the users, not the users to the model

Understand the data and how it can flexibly support self-learning or the various teaching and learning styles of teachers and students respectively. Allow course material to be published and consumed flexibly by teachers and students, respectively.

Access is not enough: focus on the end-experience

Too often content is treated as water running through a plumbing system. Unfortunately, system designers focus on the plumbing and not the experience of how the water is delivered. Stream velocity, water temperature, and volume are apt analogies for how system designers can enhance their understanding of user’s concerns. Thus, with each recommendation and predictive algorithm comes a user-facing effect. These algorithmic outcomes need user-facing stories to capture the motivation of the user and what action they might take from the information. The right hint, content, or assessment at the right time is the experience that we are designing for our users. Students, teachers, and administrators have specific needs for certain information, but fine tuning how that information is made accessible to those users should be considered more than a “nice to have.” In fact, it goes to heart of the issue of how to create systems for learning, rather than simply systems that to enable pedagogical institutions to publish content to their students.

Make actions and decisions by the system transparent

Whenever data is being collected—even to help personalize users’ experiences—it is important to let users know. This principle is arguably even more important in educational contexts, where trust and communication are central to the experience and the overall goals. Thus transparent presentation of how data collected is being used to drive and augment the content offering should further enhance and engage users (as another type of content), rather than create suspicion and reason to disengage.

Remember that learning is fun

Fun and the pleasure we get from discovering and achieving is a key component of learning. Experiences support any pedagogical style should be promoting fun engagement.

Design for a human experience

As we design and build our digital education platforms of the future using personalization and predictive technologies, we need to design with the human experience in mind. Digital learning experience should work symbiotically with the learner, improving and increasing capabilities to learn faster and deeper. The best outcomes for learners are created through more immersive teaching methods, which rely on interaction with others and application of learning as much as effective exposure to relevant content. The learning platforms should be there as a support to learners and existing teaching methods both outside and inside the classroom.

As more advanced machine learning technologies are implemented, over-automation should be avoided and peer-supported, customizable systems should complement and underpin them. As the evolution of digital products continues with new use-cases and opportunities emerging, it will be our principles and methods using research and real-world data to guide us that will create successful and exemplary experiences.

In Alpha, Beta, and Beyond

Building a data-enabled learning product will release a treasure trove of learner behavioral data to inform future releases and also the symbiotic relationship with learning content and its efficacy. Use data to track changes over time, explore new patterns, and dig deeper on problems and opportunities.

From the start, create meaningful categories that let you make sense of the data, tell a story about the experience, and ensure you develop a way to share and discuss data in your organization; start by defining the basics together.


More reading

Why I Won’t Flip My Classroom by Todd Pettigrew

Flipped Learning: A Response To Five Common Criticisms by Alan November and Brian Mull

A Critique of Student Centered Classrooms –

The MOOC Revolution That Wasn’t­ – Dan Friedman, Tech Crunch

Dean for graduate education to take leave, start new university – MIT

Stephen Greenblatt will not take questions – More or Less Bunk


Williams, J., Brooks, S. (2016). Online Learning Environments: Design in the Age of Big Data. User Experience Magazine, 16(2).
Retrieved from

Comments are closed.