Melting snowman with chalk outline

Lossless Learning: an Interview with Jared Stein

 Published on September 4, 2014 /  Written by and /  “departed” by Robert S. Donovan; CC BY 2.0 /  1

The following is an interview with Jared Stein, Vice President of Research and Education at Instructure, the makers of the Canvas LMS. Following a press release in June that announced a suite of digital products for the hybrid classroom, we caught up with Jared to get a little insight into the pedagogy behind Instructure’s new tools.


1. What inspired the idea of lossless learning?

The idea of “lossless learning” was inspired at first by a desire to think differently about some of the fundamental concepts we take for granted in education, like transmission and reception of information, in order to help teachers and technologists find new ways forward.

Like most ideas, we arrived at this metaphor from many different conversations and research threads serendipitously coming together over an extended period of time. I do remember Josh Coates and I talking about the potential of big data – truly big data from a cloud-native learning platform like Canvas. Canvas has a tremendous amount of data, more than we currently know what to do with. So how do you make that much learning data actionable in a way that is both reliable and meaningful? How do you know which data is important and which is not? Is it even the right data? I’d been reading and writing on blended learning for a while, and the lack of data in face-to-face was foremost on my mind. Josh related the challenge of lossiness in data storage, situations where the quality of information is lost — sometimes inadvertently, but sometimes to gain a benefit elsewhere, like in size or speed. This idea of educational lossiness — accidental or planned — lined up with the notion in blended education that you lose something when you move from teaching face-to-face to teaching online — and vice versa. And we were off.

The important thing about the idea of lossless learning is that it’s not just about some new tools or feature’s we’ve added to Canvas, it’s about how technology in general can help capture important information that would have been otherwise lost, and thereby lead to improvements in the quality of the learning experience. My hope is that by paying attention to education’s tendency toward lossiness, educators and technologists will find a fresh way to reflect on the information that is either captured and sacrificed in any learning experience in order to re-evaluate and iterate learning design for greater effectiveness and efficiency.

2. Instructure is primarily an LMS provider, right? So, why the turn toward in-class technology?

Over the past few years it seems education technology has been obsessed with the potential of learning analytics. I think that’s with good reason, but often conversations about learning analytics presume that online learning is the only game worth paying attention to, because that’s where the data is. While Canvas certainly recognizes the power of online learning — it’s hard to dispute the internet has been the transformative cultural phenomenon for education in the 21st century so far — we also know our teachers and students, and we know that most teaching still happens face-to-face. Why? Not because some teachers hate technology, but because face-to-face is special, and will continue to provide things that online can’t well into the future.

There’s another side to it, too: For example, I had been teaching exclusively online as an adjunct for about five years before I returned to face-to-face classroom to teach a blended course, and it was hard. I was a fish out of water, and I felt the same thing that most new online teachers feel: Teaching in this new environment meant I gained some capabilities, but I had to sacrifice others.

So we thought, we’ve given tools to online teachers that help them capture some of the sensory richness of face-to-face inside Canvas; wouldn’t it be brilliant if we could do something similar for face-to-face teachers? We should figure out a way to diminish the lossiness that happens in many face-to-face classrooms with some clever technology. But not just by capturing data (and not at all by passively monitoring students’ physiology) but by helping teachers get more of their class participating and interacting using simple tools that capture data to reflect those interactions.

3. You say in the press release for the new suite of products from Instructure that “We are excited to see the impact of this new functionality in thousands of classrooms around the world. Lossless learning and personalized, iterative instruction have been practically impossible in physical settings until now, as educators would have to spend hours trying to retroactively document the learning that takes place in their classrooms.” I’m not sure that every instructor would agree that personalized, iterative instruction has not been possible, or cannot be possible without technology. How would you respond to a skeptic?

There are always teachers who defy our predictions and make amazing things happen without special tools or resources, and we all can learn a lot from them. So, regardless of how we interpret “personalized, iterative” I think you’re right: not every teacher will agree, and not every teacher needs these particular tools. But many classroom situations make it difficult to engage all students in ways that are personalized and individualized and gather meaningful data on those interactions that promote iterative design. We’re taking a baby step toward creating new ways of doing this with Canvas Polls and MagicMarker. Polls is especially useful when your goal is to increase conceptual understanding; MagicMarker is especially useful when your goal is to dynamically assess student skill or ability in the present moment.

Having said that, I want to point out that there’s a subtext, too: “Personalized” instruction has been used almost as a synonym for “adaptive learning”, which requires lots of data to trigger automated intervention. In a sense, we’re trying to return that word “personalized” to the broader conversation about teaching and learning by suggesting the data-driven, automated approach is practically impossible face-to-face. Indeed, the automation required in many interpretations of “personalized” is antithetical to what face-to-face is best at delivering on: dynamic, human interaction.

“Iterative” is a little more clear-cut, I think; however, on reflection we probably should have added the phrase “data-driven” to that sentence. To iterate means that you change a design based on the effectiveness of previous versions. But I myself, like many teachers, have found the face-to-face experience to be so engrossing that I can’t assess its effectiveness objectively – at least not without watching a video of the class and making guesses based on future student performance. Certainly I can iterate my teaching based on my personal impressions and whatever opinions I may get from students, but it’s not data-driven. Our aim is to begin to help teachers gather data from the face-to-face environment without inhibiting what they normally do in class so they can learn more about what’s effective based on how students are performing in a class session – and even comparing that to how they perform on related projects or assessments.

Regardless, we recognize that teachers are going to establish different learning goals for their students, and will approach the act of teaching and learning from different perspectives and with different methods. So while I think the idea of lossless learning can be useful to every teacher, I wouldn’t say every Canvas tool and feature will be. We like Canvas as a platform where you pick and choose what works best.

Oh, and press release quotes are hard 🙂

4. The new tools focus on “active learning that itself generates data on student participation and performance”. I’d love to hear your thoughts on this statement. For me, active learning goes hand-in-hand with critical pedagogy, which asks that the teacher step away from the podium, as it were, and allow the classroom to be full of teacher-students and student-teachers. Does a focus on data generated by student participation keep the focus on the authority of the teacher?

There are certainly two schools of thought on the matter and many gradations in-between. I won’t take the position that the authority of the teacher must necessarily be re-defined or relegated in all learning situations, though I certainly acknowledge benefits in re-shaping the role of both teacher and student in order to better achieve desired learning outcomes.

That debate aside, the reality is that many teachers who have settled for more traditional models can find the whole idea of incorporating active learning overwhelming. Our approach is to meet teachers where they are today, and not force them to take large leaps or incorporate radical changes if they aren’t ready for that. We want to provide technology that gives more benefit than the amount of effort they put into it — and also encourages good practice. That’s why I think Polls is exciting for courses that aim to develop conceptual understanding. Polls’ simplicity and availability may help teachers who have rejected similar technology in the past think about incorporating active learning models. Polls can be used for question-driven instruction or peer instruction models, both of which change the balance of authority. The teacher remains a domain expert, but is no longer the only source of information; instead, students’ own developing expertise is acknowledged, respected, and empowered as they interact with and teach each other.

In the end, unless institutions decide to eliminate teachers, I think teachers will always play a central role as organizers, mentors, and engineers of learning. To become most effective in those roles — especially the latter — teachers can benefit from data. But students can benefit from the same data. So regardless of the position of authority that teachers hold in a classroom, we believe strongly in providing the same meaningful data to students so they can reflect on and change their own approach to learning.

5. Purdue recently innovated Course Signals, which bears some relationship to the suite of tools just introduced by Instructure. In his blog post on Inside Higher Ed, John Warner addresses the goals of the use of these sorts of software. He says that, “the goal is to improve grades and increase retention.” And then he goes on to ask some very good questions about that goal. “But what if we have different goals for our students? Or maybe additional goals beyond college persistence and good grades? For example, what if one of our goals for students is the development of agency, the ability to negotiate and exert control over their own lives?”

What are your thoughts about this? Feel free to talk about Course Signals, if you want to, but I’m mostly interested in a discussion of the way technology interacts with, supports, or interrupts the goals that teachers set for their students.

I think the answer must depend in part on the role assigned to teachers, as you mentioned above. We’re living in an exciting time where many in education and society are rethinking our assumptions about not only the process of education, but the outcomes as well. We do seem to have some general consensus around the primary goals for public education K-12, but higher education is still up for debate. What’s most valuable in a college or university experience? Is it domain expertise? Is it a broad understanding of our diverse cultural heritage? If so, traditional methods apply. Is it career-ready skills? That’s where competency-based education models get a lot of attention. I used to hear people say we go to college to learn to “think critically”. Maybe that was a boojum, because now I hear that it’s “learn how to learn”. I don’t list these facetiously — not at all; rather, my point is our values for  higher education are shifting as we figure out what’s most important in concert with a rapidly changing world. My instinct tells me that we need a combination of these goals, organized from the concrete toward increasing abstraction as learners develop advanced understanding and ability.

Technology should be flexible enough able to address a variety of educational goals and philosophical approaches, and should be deft at meshing different approaches in order to encourage individual teacher experimentation even if we must operate within more traditional institutional bounds. It’s not a new feature, and it’s not mind-blowing, but I often go back to this because it’s so simple and yet still rather novel: Canvas accepts URLs for submissions instead of just file uploads. This simple idea adds flexibility and pokes a hole in the wall of what’s historically been a closed environment without compromising the stability of that wall: It lets teachers use the LMS as an institutional system without trapping students within its walls, supporting both traditional record keeping and non-traditional student expression and ownership.

I don’t think that idea is necessarily at odds with what may seem like peripheral goals such as “increase [student] retention”. Sure, the subtext for institutions is that increasing student retention stabilizes the bottom line and makes it easier to efficiently schedule resources — very important administrative responsibilities. But increasing retention is itself a worthwhile learning goal, too, because if you don’t persist in any practice you never advance on the path toward mastery. George Leonard wrote, “Goals and contingencies are important. But they exist in the future and the past, beyond the pale of the sensory realm. Practice exists only in the present.” The hardest part of practice? Persistence in the plateaus, the doldrums.

Technology can help here as well, whether it’s by directing teacher attention to students at-risk, reminding students that they have reached a plateau or are falling behind, or suggesting learning resources or practices that have been shown to be helpful. All of these can be automated thanks to the magic of computers with useful data, but let’s not forget the automations are only as good as the data they feed off of, and they can only ever lead us, human beings with personal agency, toward the choice to engage in interactions that lead to learning. Any technology company has to make decisions about what to prioritize or they’ll end up trying to be everything, not succeeding at anything. Right now, we’re prioritizing those human interactions.

6. More generally, I’d be interested to hear your thoughts about how instructional design and critical pedagogy can mesh. I believe they can, but they’re not usually all that friendly. When you design a course, for example, what sort of pedagogy is behind it? Do you see this new suite of tools as flexible enough for multiple types of pedagogies?

For many teachers I think traditional, systematic instructional design is about as appealing as going to the dentist. You know you should undergo the procedure, and even though they try to make it as pleasant as possible the experiences still feels a little restrictive and uncomfortable. I think this is true whether a teacher adheres to a philosophy of critical pedagogy or not. But instructional design of any flavor focuses on just three questions that I think all educators can acknowledge are central to their role:

  1. What should a learner look like at the end of the experience?
  2. How can I know if a learner has achieved that state?
  3. What activities will best help the learner achieve that state?

Notice I didn’t mention “learning outcomes”, “instructional objectives”, or “assessments” at all. At its heart instructional design is a toolset, and instructional design processes provide frameworks that should make it easier to create more effective and reliable learning experiences. I sometimes get in debates over the usefulness of learning outcomes, which to me are simply a way of identifying what the purpose of the learning experience is, and to give some sense of how it might best be accomplished. Novices will benefit from more concrete, less generalizable learning outcomes; developing experts will benefit from a path toward more abstract goals.

So to get back to the tools we’re releasing in support of this ideal of lossless learning, obviously there’s limited but compelling research for certain instructional models using student response systems like Polls, and so we certainly expect to see Polls used for peer instruction and established approaches. But we’re also hopeful that by keeping Polls and MagicMarker simple and as non-prescriptive as possible teachers will try them out, adapt their usage to their students’ goals and learning outcomes, and share their results.

Add to the Conversation

1 Response
  1. […] Lossless learning is a rethinking of the traditional method of transmission and reception of information. Research shows that the traditional lecture leads to a loss in learning. We know that effective education sessions require some type of feedback to know if the audience understands the topic and has identified smart wise applications of it. Lossless learning is a pursuit of perfect feedback loops between learners and speakers. Canvas is a tech tool in the K-12 market that is leading the way between education and technology to promote lossless learning. Watch for the jump to adult education and conferences! […]

Leave a Reply

Explore Related Articles from Hybrid Pedagogy

journal logo (two nested mathematical Unity symbols in light and medium blue) above the following text: “Hybrid Pedagogy: An open-access journal of learning, teaching, and technology”

Open to Chance?

Hybrid Pedagogy on Twitter

Support Our Work