Of the 4.67 gazillion pages on the Web, exactly 1.87 nano-bazillion are understandable by children. Suppose there were a convention and a service for making child-friendly versions of any site that wanted to increase its presence and value?

That was the basic idea behind our project at the MindCET Hackathon in the Desert a couple of weeks ago.

MindCET is an ed tech incubator created by the Center for Educational Technology (CET) in Israel. “Automatically generates grade-specific versions? Hahaha.”Its founder and leader is Avi Warshavsky, a brilliant technologist and a person of great warmth and character, devoted to improving education for all the world’s children. Over the ten years that I’ve been on the CET tech advisory board, Avi has become a treasured personal friend.

In Yeruham on the edge of the Negev, 14 teams of 6-8 people did the hackathon thing. Our team — to my shame, I don’t have a list of them — pretty quickly settled on thinking about what it would take to create a world-wide expectation that sites that explain things would have versions suitable for children at various grade levels.

So, here’s our plan for Onderstand.com.

Let’s say you have a site that provides information about some topic; our example was a page about how planes fly. It’s written at a normal adult level, or perhaps it assumes even more expertise about the topic. You would like the page to be accessible to kids in grade school.

No problem! Just go to Onderstand.com and enter the page’s URL. Up pops a form that lets you press a button to automatically generate versions for your choice of grade levels. Or you can create your own versions manually. The form also lets you enter useful metadata, including what school kid questions you think your site addresses, such as “How do planes fly?”, “What keeps planes up?”, and “Why don’t planes crash?” (And because everything is miscellaneous, you also enter tags, of course.)

Before I go any further, let me address your question: “It automatically generates grade-specific versions? Hahaha.” Yes, it’s true that in the 36 hours of the hackathon, we did not fully train the requisite machine learning model, in the sense that we didn’t even try. But let’s come back to that…

Ok, so imagine that you now have three grade-specific versions of your page about how planes fly. You put them on your site and give Onderstand their Web addresses as well as the metadata you’ve filled in. (Perhaps Onderstand.com would also host or archive the pages. We did not work out all these details.)

Onderstand generates a button you can place on your site that lets the visitor know that there are kid-ready versions.

The fact that there are those versions available is also recorded by Onderstand.com so that kids know that if they have a question, they can search Onderstand for appropriate versions.

Our business model is the classic “We’re doing something of value so someone will pay for it somehow.” Of course, we guarantee that we will never sell, rent, publish, share or monetize user information. But one positive thing about this approach: The service does not become valuable only once there’s lots of content. “Because sites get the kid-ready button, they get value from it”Because sites get the kid-ready button, they get value from it even if the Onderstand.com site attracts no visitors.

If the idea were to take off, then a convention that it establishes would be useful even if Onderstand were to fold up like a cheap table. The convention would be something like Wikipedia’s prepending “simple” before an article address. For example, the Wikipedia article “Airplane” is a great example of the problem: It is full of details but light on generalizations, uses hyperlinks as an excuse for lazily relying on jargon rather than readable text, and never actually explains how a plane flies. But if you prepend “simple” to that page’s URL — https://simple.wikipedia.org/wiki/Fixed-wing_aircraft — you get taken to a much shorter page with far fewer details (but also still no explanation of how planes fly).

Now, our hackathon group did not actually come up with what those prepensions should be. Maybe “grade3”, “grade9”, etc. But we wouldn’t want kids to have to guess which grade levels the site has available. So maybe just “school” or some such which would then pop up a list of the available versions. What I’m trying to say is that that’s the only detail left before we transform the Web.

The machine learning miracle

Machine learning might be able to provide a fairly straightforward, and often unsatisfactory, way of generating grade-specific versions.

“The ML could be trained on a corpus of text that has human-generated versions for kids.”The ML could be trained on a corpus of text that has human-generated versions for kids. The “simple” Wikipedia pages and their adult equivalents could be one source. Textbooks on the same subjects designed for different class levels might be another, even though — unlike the Wikipedia “simple” pages — they are not more or less translations of the same text. There are several experimental simplification applications discussed on the Web already.

Even if this worked, it’s likely to be sub-par because it would just be simplifying language, not generating explanations that creatively think in kids’ terms. For example, to explain flight to a high schooler, you would probably want to explain the Bernoulli effect and the four forces that act on a wing, but for a middle schooler you might start with the experiment in which they blow across a strip of paper, and for a grade schooler you might want to ask if they’ve ever blown on the bottom of a bubble.

So, even if the ML works, the site owner might want to do something more creative and effective. But still, simply having reduced-vocabulary versions could be helpful, and might set an expectation that a site isn’t truly accessible if it isn’t understandable.

Ok, so who’s in on the angel funding round?

Be the first to comment »

I’m at an education technology hackathon — “Shaping the Future” — put on by MindCET, an ed tech accelerator created by the Center for Educational Technology in Israel. MindCET’s headquarters are in Yeruham in the Negev, a small-ish town that’s been growing as tech companies migrate there.

Our group created — in a demo hackathon-ish way — a tool that helps teachers create workgroups for collaborative learning based on information gleaned from machine learning about learning capabilities. The judges are four young people who are prodigiously talented computer developers. We named it Sort_ed because my team did not appreciate the sheer (shear?) genius of Zissorz. (My team was awesome.)

“Our business plan: Mexico will pay for it.”Our business plan: Mexico will pay for it.

Here are some of the projects presented at the end of the 36 hours of development. Each group has two minutes to present, ruthlessly enforced.

Interest In: A platform for students sharing their interests by learning or teaching. They can create tutorials and list them. They get badges.

Escape the classroom “Classrooms are so boring”Classrooms are so boring. Escape the Classroom uses the power of whatsApp and escape rooms (i.e., the puzzle rooms you try to get out of collaboratively, using educational clues.

Rope. Team-based learning.”Rope Team” is a course format for Moodle that implements a unique workflow for learning a set number of topics.” There are roles and responsibilities, and a workflow with automation. (The creator of Moodle, Martin Dougiamas, is on that team.)

Snippy. Every child has a passion for something. Snippy lets students create content, share it, and share the content of others. A chatbot interviews you and presents relevant materials from what other students have uploaded. You can create a multimedia object to share your passion.

Clash of Brains. No one (hardly) likes tests. This team wants to bring fun and sociality into assessments. Teachers create a quiz and the app sends a code to students. Students can “duel” other students.

Edventure — a tailor-made education adventure. In the example, a friendly monster asks for help with a question. It’s a collaborative RPG for 3-5 players.

Playful — “promoting education through play.” “They introduce RRS: Robot Rewards System.”They introduce RRS: Robot Rewards System. You get real-world rewards from a robot: perhaps art, maybe it does a dance, etc. You can also be challenged to hack the robot.

Disruptive text. “For students who hate to read.” For 7-9th graders who struggle to read long texts. The text becomes a riddle they need to decode. They use several techniques to challenge the reader: Difficult fonts. Blurred text until you click. Mirrored words that reverse when clicked.

The Words and Image Challenge. “Students from a local Bedouin school are wearing a word and a drawing of an object.”Students from a local Bedouin school (unfairly adorable) are wearing a word and a drawing of an object. They throw a ball to the person with the name of the object on her or his shirt. You have to throw the ball as quickly as you can, in “hot potato” style.

ReflectMe. A team from the Israeli army has created an app that enables students to give one another feedback. (They contrast this with top-down military structure.) It has a simple, intuitive UI. In the example, students can leave feedback on a video, tied to the time code.

Peerz. Standardization misses individual passion. The future is individualized passion-based learning. But teachers can’t scale for this. A student asks Peerz a question, with hashtags. Other students can respond. The system suggests resources, better questions, etc. The questions are rated. “Peerz monetizes talent discovery.” “Co-creative learning in your pocket.”

EdMarket. “The Amazon of Education.” It gives teachers the ability to choose the best products. EdMarket is a marketplace of learning resources, sponsored by the govt (or so their business plan says). The students and teachers can reference the market.

Owie. “An AI best friend.” It will help students talk about emotions, especially when the situation is stressful. Owie is a chatbot that lets 8-12 year olds communicate with other friends and play emotionally-supportive games.

Shape on You. A virtual reality experience that teaches geometric figures. It aims at making it easier to grasp abstract concepts. You can manipulate figures, see the dimensions, alter them, and see the results. You can share your figure with other students.

Action Learning. They show a robot (a bit Lego-like) that models a robot for delivering water in the desert. They programmed this with the Creative Learning Lab. They created a space, physical and digital, where you can meet others and learn life lessons. “Solving problems that you couldn’t solve in school.”

Who Am I?. How to encourage creation within children, and how to motivate them to be interactive and really invest in the process. Who Am I? is a mini-quest game where you try to discover who is hidden in the room. It’s a mobile app that you navigate by moving the phone. You find clues. Students can make their own puzzles.

DPlay — “Democracy Playground.” “How do you liberate learning for self-reflection.” They created a platform for debating issues and reflecting on one’s own positions. Students fill out a little survey about the opposing positions, reflecting on why they react against it. These surveys are compiled over time. Is a student changing her vote often? Is she always voting with her friends?

OwnEd. They created an app that takes away the stress from students (12-13yr old) who are unsure what subjects they should be taking. It lets them design their own learning program. How do they want to learn? When do they want to learn? An “intuitive app” visually stimulates them to say that they’re most interested in. The backend uses this to suggest areas. The app suggests a time structure for their program. “Breaking the rules around space and time.”

Imagibate.com “Free learners’ mind from the old structures by engaging them in debate.” They use imaginary worlds to make sure the issues are not personally sensitive. The debates will be put up on line. E.g., “a world of unicorns and coffee beans”a world of unicorns and coffee beans, two tribes that have gotten along until the coffee beans learn to make a scent they find pleasurable, but it makes the unicorns sick. The team models a live debate, complete with a unicorn hat.

The winner was Who Am I?. We came in second, by one vote.

Be the first to comment »