In an earlier post, I talked about how education at Oxford has remained remarkably unchanged over the course of centuries.  As an example of an “evolved” learning style, I provided the “American” higher education system, with its investment in lecture-based pedagogy and social learning.  Really, though, all that represents is a sideways move in learning evolution.  American colleges didn’t advance learning, they just borrowed elements from less advanced learning environments, namely the K-12 classroom.

This is sad because the traditional classroom developed not out of any particularly brilliant insights on how best to teach students.  Instead, it developed from a need to educate as many students in the shortest amount of time and with the fewest consumption of resources.  Kids have required education since the dawn of civilization and parents figured out that it was more efficient for them to throw all the kids together in a classroom and use only one adult to teach all the kids at once.  Efficient in terms of time commitment for parents?  Yes.  But no one can possibly defend this as the optimal path toward academic efficacy.

Adaptive learning, then, marks the dawn of the next stage of education’s evolution.  Guided by the foundational premise that every student learns differently, adaptive learning uses technology to understand something about how a student learns and then provide content differently based on how it qualifies a student’s knowledge level.  A basic example is the GMAT; as you answer questions on the GMAT, it adjusts the difficulty of each successive question it serves up to you.  It helps the GMAT create gradations of knowledge by constantly subdividing test takers into more refined groups of “knowledge isobars,” if you will.

Adaptive learning today

Adaptive learning today is much more than Computerized Adaptive Testing (CAT), though.  What adaptive learning platforms currently offer is not just assessment based on adaptive tests, but content delivery based on adaptive assessments.  In other words, rather than just more intelligently figuring out how much you know, we’re actually using that understanding to serve up adaptive content to optimize your learning path.  For example, a math student who struggles to learn concepts when tested through word problems would receive more non-word problems when being taught math, but would simultaneously receive more basic critical reading content to improve his ability to answer word problems.  (The cross-disciplinary ethos this evinces is immensely exciting.)

The reason we’re finally able to evolve this way is, obviously, because of technology.  First, because content is now being consumed on iPads and Google Chrome rather than static print media, the content delivery system can adapt to the individual learner.  Second, what makes adaptive learning so potentially powerful is that the technology-based delivery of content means that we can get smarter about assessing students by collecting constant data on their learning experience.  It’s not just about using the interface to personalize content, it’s about turning the interface into a real-time  testing system that’s always running in parallel.

Big Data.  Boom.

</facetious>  Gone are the days of event-driven learning assessments.  You don’t need to wait for midterms and finals or even weekly pop quizzes to figure out how well your students are learning.  You can use the content delivery interface as a real-time dashboard to visualize what, how, and how well every student is learning.  Does anyone doubt what this could mean for educational outcomes?  (I personally have visions of sugarplums dancing in my head.)

The future of adaptive learning

But for all this bombast, adaptive learning is still a far cry from what it ought to (and will eventually) be.  Think about the current state of affairs like a music recommendation app (say, Songza or Pandora).  You give the app a general theme and it starts serving you songs which other people who liked your specific theme have enjoyed.  As you thumbs-up or thumbs-down successive songs, you are, in theory, making the app smarter about what songs to play for you in the future.

But there is so much more depth Songza or Pandora can achieve in terms of understanding how much you’re enjoying their playlist.  A binary thumbs-up or -down doesn’t capture how much you like the song, for example, or if you hate it in the current context but want it in another one.  It doesn’t capture what you like about a song.  Worse still, if a song is anomalous relative to the artist’s entire oeuvre, maybe you don’t want the dialectic of metacarpal semiotics to categorically judge many decades worth of artistic endeavor.  (Sorry – I just happen to really love “Radio Nowhere” but only kinda like the rest of Springsteen.  Not that I’m bitter.)

Similarly, all adaptive learning does today is use the same type of testing we’ve always used to determine if students understand material.  It still comes down to submitting an answer to a predetermined question and get graded on whether it’s right or wrong; the only difference is that adaptive learning presents this evaluation constantly throughout the learning experience and uses the responses to change up the content.

There are other – arguably deeper – ways of understanding “learning.”  Without straying too deep into epistemology, how can you truly understand the rewiring of a person’s neurons after a lecture or reading?  (Those who say “multiple choice exam” should stop reading.)  To choose one direction specifically, I’d argue that “learning” correlates with biological indicators like the release of endorphins or other “rewarding” neurotransmitters.  For me, it’s sometimes accompanied by an elevated heart rate, and, in rare cases, a tear.  Learning is one of the most basic traits rewarded by evolution and consequently, its status as a Darwinian imperative ought to be reflected in a living being’s biology.

Future adaptive learning platforms will measure more quotidian statistics like time spent on a problem or where your mouse hovered before you clicked on an answer.  They’ll only get smarter as more and more students are loaded onto a platform and the time-series data grows exponentially.  But they’ll also start measuring things like facial expression and eye movements.  It’s not inconceivable to me that they’ll measure heart rate and your cerebral chemical composition, though admittedly, only when technology improves to the point where such biometric data can inexpensively and non-invasively be tracked.  (So, I guess next year?)  These platforms will be built by a collaboration of philosophers, linguists, psychologists, and Education PhDs, before even getting to the engineers and teachers.  They’ll start not with the content to be delivered but by asking an the most basic question of all: what does it mean for a person “to learn”?

Here are some rather leading questions I have when thinking about education many, many years down the road:

1. Will there be special adaptive learning workstations that are focused on collecting as much data as technologically possible on the engaged learner?  Potentially.
2. Will there be APIs to integrate with biometric apps so that adaptive learning platforms can connect biometric outputs to their recommendation engines and serve up content still more intelligently?  Likely.
3. Will the traditional four-wall classroom and the perpetually underfunded school district cease to exist as independent entities?  Almost certainly.

(On the school district point, as society blends together, the need for local culture simultaneously increases and decreases and instead of school districts, I’m guessing the LEAs will end up being arbiters of content that ought to be taught.  There’s still going to be local culture and rather than LEAs needing to administer tax dollars to different schools, they’ll be responsible for versioning a universal curriculum to preserve local culture to the extent necessary/possible.  It’s admittedly the premise of a dystopian novel right there but it’s a pretty logical extrapolation of the trends in education, globalization, and technology right now.)

Ay, there’s the rub

The “rub,” or at least the biggest challenge for me to get my head around is how this movement towards personalizing learning seems to run counter to social learning, an absolutely critical component to fulfilling the obligations of education in society.  What we love about school – those of us who loved it – was that we had friends around us to learn from and, in turn, teach.  We loved the repartee between teachers and students, each one demonstrating in turn a slightly different view of the world.  We loved that not every student was as moved by Keats as he was by, say, Bukowski or Wallace Stevens; that in US History, some students argued over the course of a year consistently against the principles of big government helping you come closer to understanding their position, even if you never fully agreed; or that there was a bolder student toward whom you were eternally thankful for his sparking a discussion on why post-Algebraic math was in fact a necessary skill for people who have only ever aspired toward relationships with keyboards not calculators.

None of this could happen in an adaptive learning environment – or could it?  Obviously, there is the potential that learning evolves toward a bifurcated model where the first part focused on individual content consumption, optimized at a personalized level and the second part was a group discussion of the content consumed.  The “flipped-classroom” movement is a good example of what this might look like.  Students go home to consume content and use the valuable class time not on moving at the pace of the slowest person (or fastest person) but rather to re-learn the same content in a group setting with the support of having individually learnt it first.

Another, more ambitious, theory is that adaptive learning groups learners together across the school, district, county, state, country, or, dare I say, the world, just based on how each one learns.  Across 700 million+ primary school learners in the world, you’d imagine learners can be subdivided into even the most granular archetypes and still find huge sub-communities that learn across “personalized-enough” content delivery patterns.  What might this look like?  Online learning today, and not even just MOOCs like Coursera or Udacity but with mandatory asynchronous and synchronous modules and Adobe Connect support for web interaction.

Conclusion

As I mentioned earlier, learning is a basic evolutionary requirement.  Animals that didn’t learn to stop eating poisonous berries died out; those that didn’t learn to build shelter in rough climates dropped out of the gene pool.  To have kids in school not excited about learning, not willing to engage with their subject material, is maybe the worst crime that society hasn’t actually criminalized yet.

I’d be remiss if I didn’t link to Knewton, the main thought-leader in the space today.  Knewton’s website – unabashedly data science-y when describing what the company does – is a great source for learning exactly how adaptive learning platforms can (ought to?) be built.  Rather than focus on building adaptive content themselves, they’ve skipped that part and moved to the intermediary model.  They provide a platform and an API for which 1) content providers like Pearson can tailor their content for adaptive environments, and, 2) educators can deliver content to students and track the personalized learning outcomes.  Knewton gets its own “deep dive” post in the future, but for now, check out their Adaptive Learning Intro for more info.

Technology finally allows us to push education to its next evolutionary stage.  Forgetting the characteristics of what this will look like, from an outcomes perspective, the most exciting aspect of it is its promise that every student can be as engaged a learner as the most engaged among us were growing up.  It’s not just an imperative for their own personal growth, it’s important toward the growth of their peers and the growth of society itself.  More productive social learning, more productive members of society is universally a good thing.  Adaptive learning – the first time we can ever say that we are, categorically, teaching every student the way he or she ought to be taught – is going to have a massive impact on societal productivity and, dare I say, happiness.