How we move fast to improve education
One feature of most grade schools is that much of the curriculum is static. Schools typically purchase curriculum in September, then the teachers use it throughout the year. The following year they maybe buy something else, but most likely they will reuse textbooks and materials for a few years. The average turnover time for textbooks is about seven years.
At eSpark, we believe that we can use technology to improve academic and life outcomes for kids. One advantage of software is that it lets us iterate much more quickly than traditional textbooks. We can change what content kids are learning from in response to data about how they are using the app in schools. Instead of iterating on a textbook every few years, we can customize the learning path for each child in response to their needs in real time.
Of course, doing this is easier said than done. There are a few technical hurdles that we have been working on leaping in order to enable the world real time education. I wanted to share some of the technical hurdles we have faced on the road to speedy iteration, and how we have adapted our technology and culture to solve them.
Make our iPad app out of mostly JavaScript
Because our product is run on the iPad, we distribute the app through the Apple App Store. One challenge is the difficulty of keeping our customers’ iPads up to date. Most schools do not empower teachers to update devices themselves - instead, as in many large companies, devices are centrally managed by the IT department. This means that updating the eSpark app, as well as loading on all of our partner applications, is a big hassle. As a result, we may have the opportunity to update the iPad app no more than once or maybe twice in a given school year.
However, we are a startup - we want to watch how our customers use the software, learn from their behavior, and release features and bug fixes to improve student learning throughout the year. How do we do that when we only get a few iterations of the native app each year?
To resolve that issue, we switched the majority of the student-facing portion of the app to JavaScript, wrapped in a thin native app written in Objective C. As a result, we can now push very frequently - sometimes multiple times a day, but usually once or twice a week. This lets us respond to feedback very quickly.
Not that this was particularly easy. Ringmark, the mobile benchmarking suite, defines four categories of behavior that make up complex mobile apps: 2D Games, Audio Apps, Video Apps, and Camera Apps - and eSpark hits on all of them!
Our app is a side-scroller game with lots of integrated audio (because Kindergarteners can’t read yet) and video (all of our instruction is presented in video). At the end of each “quest”, students record a video of themselves teaching the content they learned - so we use native hooks to access the camera as well.
Despite all of the effort of doing this in Javascript, the speed of iteration outweighs the costs and is definitely worth it.
Iteratively launch to a few classrooms, then expand out
If we are pushing out code every day, then how do we make sure that we don’t disrupt a bunch of classes experience when we roll it out?
The answer is to roll out all new features on a small scale with willing and technically savvy teachers, get their feedback, fix bugs, and ensure the product is in a good spot. Then, once it looks good, we can roll it out to a wider audience.
The majority of teachers are excited to try something new, as long as they have a choice in the matter. So for those that want to give us feedback, we are happy to give them new products early on. For those that would prefer not to deal with it until it’s well baked, we leave the current version running and only roll out new features once they have been well tested.
We are very respectful of our students’ and teachers’ time, and we want to try very hard to prevent any disruption to their learning. After all, a learning tool is only effective if it works consistently.
Because of the regimented structure of the school day, the impact of an outage varies widely by time of day. Students usually use eSpark for only 20-30 minutes in a day. If eSpark is down or fails to work during that time, then it’s an entire day of lost instruction time - whereas if eSpark is down after school, then it’s still a problem, but not as big.
You can see this in our usage graph. We have a huge spike in the morning, but it trails off in the late afternoon. By restricting our pushes to the afternoon after 3pm, we reduce the risk of disrupting a school’s learning time even as we change the product very quickly.
Through these and other cultural and process changes, we believe we can strike a good balance between moving fast while still respecting the primary goal of student learning. Our teachers and students are excited when they see new features showing up, or when bugs are fixed very quickly. We are continuing to work on making our process better in order to serve students well.
At Mineola Public Schools in New York, students using eSpark showed concrete gains of 2-5x in their targeted subject areas. The superintendent has been impressed, such that he expanded the pilot to 600 students this year. Check out the details in this video.
The author of this article offers a decent overview of the state of the world and motivations behind personalized learning. However, he concludes that kids may suffer from too much personalization. I couldn’t disagree more.
From the article:
All of this can be good, of course, but have you noticed that after you have bought lots of books on Amazon, the titles it recommends all come to sound pretty much alike?
As a gamer I want a video game to hold my hand when I begin, but I do not want it to customize the boss battle for me. I want the boss battle to test me, and I want to feel a sense of growth and accomplishment when I slay him.
What happens when people with different “sweet spots” have to learn, solve problems, and collaborate with others who have different “sweet spots,” as people so often have to do in modern workplaces? I wonder what would happen should, God forbid, children run into learning situations in the world that cannot be optimized for them individually. What if the world changes and the problems that arise just do not afford solutions that fit their sweet spot? What if their sweet spot is just no good for certain types of learning and problem solving?
Sure, a poorly designed personalization algorithm could be difficult - but for the vast majority of students, education is currently way too standardized. Research has repeatedly shown that when you tailor learning for the child, it helps them grow faster so that they can get on to the harder problems.
In their focused subject areas, students using eSpark have shown quantitative gains of twice the control group. It’s hard to argue that those kids would be better off had they been stuck learning content inappropriate for their level.
I asked myself, why do I want to decentralize education? Why? Oh, it’s obvious! To Unleash Latent Human Potential!
Well put, Dave.
This is a somewhat rambling article, but I do agree with this sentiment:
Without a trace of sarcasm or cynicism, I tell you honestly that I don’t think we’ve even begun to scratch the surface when it comes to education in the era of the Internet. And for all the millions of teachers and students using amazing platforms for organizing classes, managing classrooms, connecting to resources, and delivering content, I really don’t think we’ve begun to even imagine what the future of learning looks like.
There is so much that’s left to do, and this is a long game. In thirty years, I bet we’ll still be debating what’s next for edtech (although it will probably have a different name by then). I’m just excited to be part of it.
I had a great conversation this week with Chris Liang-Vergara from First Line Schools, which is a charter organization helping New Orleans schools recover post-Katrina.
For Chris, the quality of an edtech product is impossible to separate from the quality of the accompanying teacher dashboard.
Using the weather channel as an analogy, he argues that teacher dashboards should:
- Give a big-picture view of the state of the classroom
- For each student, give a high level forecast (“sunny”, “rainy”) and suggested actions (“wear a coat”)
- Draw clear lines between the suggested actions and the underlying evidence.
- Expose the data not only to the teacher, but directly to the student, so they can track their progress. Even Kindergarteners do better with clear goals and the ability to see how they are doing against them.
These are powerful suggestions that we are working on implementing at eSpark. I highly recommend his post for more insights.
It seems that most good teachers spend a lot of time trying to differentiate their instruction - offer a customized learning path for each kid. But this is really hard to do with only one teacher:
When I first started teaching, I went overboard with differentiation. Quickly realizing the uniqueness of every student in my class, I started creating individualized homework packets and personalized assignments. Instead of ending up with high-performing students, I ended up with a perpetual stomachache. Figuring out how to differentiate without killing yourself can be perplexing.
The promise of technology is that it can make differentiation easier and more accessible for all teachers. With the aid of computers or iPads, one teacher can offer a customized plan for each student without bending over backwards.
Robert Provonost from Belle Haven Community School in Menlo Park, CA is using various online learning tools in his classroom, such as Planet Turtle and Dreambox. He really hits the nail on the head:
Really trying to address giving every student what they need and giving them that differentiated instruction is where the technology really comes in handy. Not only are they getting immediate feedback and they are able to move at their own pace, but then I’m also able to support the students who really need my support.
Watch the full story here:
(via ChicagoUTEP)
For the past several months, many entrepreneurs have been readying brand new releases of software for the upcoming school year. I’ve talked with several of the companies mentioned in this article, including Innosight Institute, Junyo, Clever, and Learnsprout. It’s pretty cool to think that what they have been building towards is about to launch in the next week or two, and then we get to see how it all plays out. I think this year is going to be a big one.
Introducing the new, searchable Innosight Blended-Learning Profiles
Here’s a problem that is vexing to many working on education technology: how do you find out what’s actually going on in innovative schools? Educators, administrators, policy makers, and entrepreneurs spend a lot of time looking for answers to questions like:
- How do schools choose which technology to use to advance student learning? How do administrators sort through the pile of ed tech software to choose the diamonds in the rough?
- How do they actually implement it? Do they have a computer lab, laptop carts, iPads? Do they use centers, one-on-one devices, issue to students to take home? What professional development works best?
- What kind of results are they getting? Who’s actually making progress with this stuff? What are the duds?
Turns out, it’s pretty hard to find these answers today. Many education technology blogs cover companies and products, but don’t focus as much on schools, so there aren’t a lot of high quality, open, unbiased resources in this area.
But that is changing. Have you read Disrupting Class? (If not, please go read it - the book is a fantastic how-to for the ed tech entrepreneur). The authors of that book, Clayton Christensen and Michael Horn, founded Innosight Institute to track progress in blended learning. They have published several influential white papers about blended learning that have really defined how that term is understood in the field.
Besides writing books and whitepapers, the folks at Innosight Institute have been spending a lot of time on the phone - calling up innovative schools around the country and asking them questions like those above. They then published detailed profiles for these schools. Until now, these profiles have been available, but in an unsearchable format, which has made them a good, but not great tool for answering the questions above.
Opening up the doors
Over the past few months, I’ve been working with the Innosight Institute (through the Full Circle Fund) to make blended learning easier to understand. Today, we launched the first iteration of the searchable database of school profiles. Now, you can browse through a nationwide map of the profiles that the Innosight Institute has collected. Filter by the type of blended learning or type of school, grades and subjects served - and most importantly, the education technology tools and products they are using!
We hope that this resource can help educators and policy makers answer the practical questions about how best to implement blended learning in their own schools.
Code Details
The app is written pretty much entirely in Javascript, using a Cloudant backend. It’s hosted on Heroku. The code is all open sourced - check it out on Github. This is just the first iteration - but we need plenty more help to make it awesome. Let me know if you’re interested in chipping in.
