Around a 100 million first-grade-aged children lack access to schools. A foundation is testing whether poor children who are given computers and learning software can teach themselves.
HACKED in MONTHS, with ZERO INSTRUCTION
Given Tablets but No Teachers, Ethiopian Children Teach Themselves
by David Talbot / October 29, 2012
With 100 million first-grade-aged children worldwide having no access to schooling, the One Laptop Per Child organization is trying something new in two remote Ethiopian villages—simply dropping off tablet computers with preloaded programs and seeing what happens. The goal: to see if illiterate kids with no previous exposure to written words can learn how to read all by themselves, by experimenting with the tablet and its preloaded alphabet-training games, e-books, movies, cartoons, paintings, and other programs. Early observations are encouraging, said Nicholas Negroponte, OLPC’s founder, at MIT Technology Review’s EmTech conference last week.
The devices involved are Motorola Xoom tablets—used together with a solar charging system, which Ethiopian technicians had taught adults in the village to use. Once a week, a technician visits the villages and swaps out memory cards so that researchers can study how the machines were actually used. After several months, the kids in both villages were still heavily engaged in using and recharging the machines, and had been observed reciting the “alphabet song,” and even spelling words. One boy, exposed to literacy games with animal pictures, opened up a paint program and wrote the word “Lion.” The experiment is being done in two isolated rural villages with about 20 first-grade-aged children each, about 50 miles from Addis Ababa. One village is called Wonchi, on the rim of a volcanic crater at 11,000 feet; the other is called Wolonchete, in the Great Rift Valley. Children there had never previously seen printed materials, road signs, or even packaging that had words on them, Negroponte said.
Earlier this year, OLPC workers dropped off closed boxes containing the tablets, taped shut, with no instruction. “I thought the kids would play with the boxes. Within four minutes, one kid not only opened the box, found the on-off switch … powered it up. Within five days, they were using 47 apps per child, per day. Within two weeks, they were singing ABC songs in the village, and within five months, they had hacked Android,” Negroponte said. “Some idiot in our organization or in the Media Lab had disabled the camera, and they figured out the camera, and had hacked Android.” Elaborating later on Negroponte’s hacking comment, Ed McNierney, OLPC’s chief technology officer, said that the kids had gotten around OLPC’s effort to freeze desktop settings. “The kids had completely customized the desktop—so every kids’ tablet looked different. We had installed software to prevent them from doing that,” McNierney said. “And the fact they worked around it was clearly the kind of creativity, the kind of inquiry, the kind of discovery that we think is essential to learning.”
“If they can learn to read, then they can read to learn,” Negroponte said (see “Emtech Preview: Another Way to Think About Learning”). In an interview after his talk, Negroponte said that while the early results are promising, reaching conclusions about whether children could learn to read this way would require more time. “If it gets funded, it would need to continue for another a year and a half to two years to come to a conclusion that the scientific community would accept,” Negroponte said. “We’d have to start with a new village and make a clean start.”
The idea of dropping off tablets outside of the context of schools is a new paradigm for OLPC. Through the late 2000s, the company was focused on delivering a custom miniaturized and ruggedized laptop, the XO, of which about 3 million have been distributed to kids in 40 countries. Deployments went to schools including ones in Peru (see “Una Laptop por Nino”). Giving computers directly to poor kids without any instruction is even more ambitious than OLPC’s earlier pushes. “What can we do for these 100 million kids around the world who don’t go to school?” McNierney said. “Can we give them tool to read anD learn—without having to provide schools and teachers and textbooks and all that?”
ONE SMARTPHONE per CHILD (cont.)
Another Way to Think about Learning
by Nicholas Negroponte / September 13, 2012
Seymour Papert, a computer scientist and pioneer in artificial intelligence, once said: “You cannot think about thinking unless you think about thinking about something.” Does this apply to learning? Maybe not. Here is what I mean. As we industrialized learning and created schools, we needed to measure the system’s efficacy and each child’s progress. What you really want to measure is curiosity, imagination, passion, creativity, and the ability to see things from multiple points of view. But these are hard to measure other than one on one, and even then, the assessment will be subjective. So instead, we measure what a child knows, and from that we infer that the child has learned how to learn. This is the real aspiration we have for our children: learning learning.
I believe that we get into trouble when knowing becomes a surrogate for learning. We know that a vast recall of facts about something is in no way a measure of understanding them. At best, it is necessary but not sufficient. And yet we subject our kids to memorizing. We seem to believe that rote learning is akin to physical exercise, good for their minds. And, quite conveniently, we can test whether the facts stuck, like spaghetti to a wall. In some cases knowledge is so drilled in that you know and hate a subject at the same time. The closest I have ever come to thinking about thinking is writing computer programs. This involves teasing apart a process into constituent parts, step-by-step functions, and conditional statements. What is so important about computer programs is that they (almost) never work the first time. Since they do something (versus nothing), just not what you wanted, you can look at the (mis)behavior to debug and change your code. This iterative process, so common in computer programming, is similar to learning.
The gods must be crazy
Have you watched a two-year-old use an iPad? The meteoric rise of modern instructionism, including the misguided belief that there is a perfect way to teach something, is alarming because of the unlimited support it is getting from Bill Gates, Google, and my own institution, MIT. While Khan Academy is charming and brilliantly nonprofit, Salman Khan cannot seriously believe that he and a small number of colleagues can produce all the material, even if we did limit our learning to being instructed. One Laptop per Child (OLPC), a nonprofit association that I founded, launched the so-called XO Laptop in 2005 with built-in programming languages. There are 2.5 million XOs in the hand of kids today in 40 countries, with 25 languages in use. In Uruguay, where all 400,000 kids have an XO laptop, knowing how to program is required in schools. The same is now true in Estonia. In Ethiopia, 5,000 kids are writing computer programs in the language Squeak. OLPC represents about $1 billion in sales and deployment worldwide since 2005—it’s bigger than most people think. What have we learned? We learned that kids learn a great deal by themselves. The question is, how much?
To answer that question, we have now turned our attention to the 100 million kids worldwide who do not go to first grade. Most of them do not go because there is no school, there are no literate adults in their village, and there is little promise of that changing soon. My colleagues and I have started an experiment in two such villages, asking a simple question: can children learn how to read on their own? To answer this question, we have delivered fully loaded tablets to two villages in Ethiopia, one per child, with no instruction or instructional material whatsoever. The tablets come with a solar panel, because there is no electricity in these villages. They contain modestly curated games, books, cartoons, movies—just to see what the kids will play with and whether they can figure out how to use them. We then monitor each tablet remotely, in this case by swapping SIM cards weekly (through a process affectionately known as sneakernet).
Within minutes of arrival, the tablets were unboxed and turned on by the kids themselves. After the first week, on average, 47 apps were used per day. After week two, the kids were playing games to race each other in saying the ABCs. Will this lead to deep reading? The votes are still out. But if a child can learn to read, he or she can read to learn. If these kids are reading at, say, a third-grade level in 18 months, that would be transformational. Whether this can happen has yet to be proved. But not only will the results tell us how to reach the rest of the 100 million kids much faster than we can by building schools and training teachers, they should also tell us a great deal about learning in the developed world. If kids in Ethiopia learn to read without school, what does that say about kids in New York City who do not learn even with school? The message will be very simple: children can learn a great deal by themselves. More than we give them credit for. Curiosity is natural, and all kids have it unless it is whipped out of them, often by school. Making things, discovering things, and sharing things are keys. Having massive libraries of explicative material like modern-day encyclopedias or textbooks is fine. But such access may be much less significant than building a world in which ideas are shaped, discovered, and reinvented in the name of learning by doing and discovery.
or HOW to TEACH LEARNING
by Evan Ackerman / Oct 30, 2012
What happens if you give a thousand Motorola Zoom tablet PCs to Ethiopian kids who have never even seen a printed word? Within five months, they’ll start teaching themselves English while circumventing the security on your OS to customize settings and activate disabled hardware.
The One Laptop Per Child project started as a way of delivering technology and resources to schools in countries with little or no education infrastructure, using inexpensive computers to improve traditional curricula. What the OLPC Project has realized over the last five or six years, though, is that teaching kids stuff is really not that valuable. Yes, knowing all your state capitols how to spell “neighborhood” properly and whatnot isn’t a bad thing, but memorizing facts and procedures isn’t going to inspire kids to go out and learn by teaching themselves, which is the key to a good education. Instead, OLPC is trying to figure out a way to teach kids to learn, which is what this experiment is all about. Rather than give out laptops (they’re actually Motorola Zoom tablets plus solar chargers running custom software) to kids in schools with teachers, the OLPC Project decided to try something completely different: it delivered some boxes of tablets to two villages in Ethiopia, taped shut, with no instructions whatsoever. Just like, “hey kids, here’s this box, you can open it if you want, see ya!”
Just to give you a sense of what these villages in Ethiopia are like, the kids (and most of the adults) there have never seen a word. No books, no newspapers, no street signs, no labels on packaged foods or goods. Nothing. And these villages aren’t unique in that respect; there are many of them in Africa where the literacy rate is close to zero. So you might think that if you’re going to give out fancy tablet computers, it would be helpful to have someone along to show these people how to use them, right? But that’s not what OLPC did. They just left the boxes there, sealed up, containing one tablet for every kid in each of the villages (nearly a thousand tablets in total), pre-loaded with a custom English-language operating system and SD cards with tracking software on them to record how the tablets were used. Here’s how it went down, as related by OLPC founder Nicholas Negroponte at MIT Technology Review’s EmTech conference last week:
“We left the boxes in the village. Closed. Taped shut. No instruction, no human being. I thought, the kids will play with the boxes! Within four minutes, one kid not only opened the box, but found the on/off switch. He’d never seen an on/off switch. He powered it up. Within five days, they were using 47 apps per child per day. Within two weeks, they were singing ABC songs [in English] in the village. And within five months, they had hacked Android. Some idiot in our organization or in the Media Lab had disabled the camera! And they figured out it had a camera, and they hacked Android.”
This experiment began earlier this year, and what OLPC really want to see is whether these kids can learn to read and write in English. Around the world, there are something like 100,000,000 kids who don’t even make it to first grade, simply because there are not only no schools, but very few literate adults, and if it turns out that for the cost of a tablet all of these kids can simply teach themselves, it has huge implications for education. And it goes beyond the kids, too, since previous OLPC studies have shown that kids will use their computers to teach their parents to read and write as well, which is incredibly amazing and awesome.
If this all reminds you of a certain science fiction book by a certain well-known author, it’s not a coincidence: Nell’s Primer in Neal Stephenson’s The Diamond Age was a direct inspiration for much of the OLPC teaching software, which itself is named Nell. Here’s an example of how Nell uses an evolving, personalized narrative to help kids learn to learn without beating them over the head with standardized lessons and traditional teaching methods:
Miles from the nearest school, a young Ethiopian girl named Rahel turns on her new tablet computer. The solar powered machine speaks to her: “Hello! Would you like to hear a story?” She nods and listens to a story about a princess. Later, when the girl has learned a little more, she will tell the machine that the princess is named “Rahel” like she is and that she likes to wear blue–but for now the green book draws pictures of the unnamed Princess for her and asks her to trace shapes on the screen. “R is for Run. Can you trace the R?” As she traces the R, it comes to life and gallops across the screen. “Run starts with R. Roger the R runs across the Red Rug. Roger has a dog named Rover.” Rover barks: “Ruﬀ! Ruﬀ!” The Princess asks, “Can you ﬁnd something Red?” and Rahel uses the camera to photograph a berry on a nearby bush. “Good work! I see a little red here. Can you ﬁnd something big and red?”
As Rahel grows, the book asks her to trace not just letters, but whole words. The book’s responses are written on the screen as it speaks them, and eventually she doesn’t need to leave the sound on all the time. Soon Rahel can write complete sentences in her special book, and sometimes the Princess will respond to them. New stories teach her about music (she unlocks a dungeon door by playing certain tunes) and programming with blocks (Princess Rahel helps a not very-bright turtle to draw diﬀerent shapes). Rahel writes her own stories about the Princess, which she shares with her friends. The book tells her that she is very good at music, and her lessons begin to encourage her to invent silly songs about what she’s learning. An older Rahel learns that the block language she used to talk with the turtle is also used to write all the software running inside her special book. Rahel uses the blocks to write a new sort of rhythm game. Her younger brother has just received his own green book, and Rahel writes him a story which uses her rhythm game to help him learn to count.
After a repressive history, Uruguay’s left-wing government is paving the way for a new generation of digital-savvy, globalised nationals by being the first country to give every child their own laptop.
Ex-guerilla fighter, Jose Mujica, has been in office since 2010. He has never given up his ideals of social justice. The aim of the ambitious ‘A laptop per child’ program is to help close the divide between rich and poor, and between developed and developing countries. Many parents and teachers were initially sceptical; years without proper social policies still shape life in the slums. ”I don’t want to know anything about computers. It is a child who knows something about computers and technology that throws bombs.” Yet despite early skepticism, the program is now being widely embraced. “We need to educate the children differently so that they understand the world of today.”
a LESSON from NELL
Growing Up With Nell: A Narrative Interface for Literacy
by C. Scott Ananian, Chris J. Ball, Michael Stone of the One Laptop Per Child Foundation
Abstract: Nell is a tablet-oriented education platform for children in the developing world. A novel modular narrative system guides learning, even for children far from educational infrastructure, and provides personalized instruction which grows with the child. Nell’s design builds on experience with the Sugar Learning Platform , used by over two million children around the world.
Miles from the nearest school, a young Ethiopian girl named Rahel turns on her new tablet computer. The solar-powered machine speaks to her: “Hello! Would you like to hear a story?” She nods and listens to a story about a princess. Later, when the girl has learned a little more, she will tell the machine that the princess is named “Rahel” like she is and that she likes to wear blue—but for now the green book draws pictures of the unnamed Princess for her and asks her to trace shapes on the screen. “R is for Run. Can you trace the R?” As she traces the R, it comes to life and gallops across the screen. “Run starts with R. Roger the R runs across the Red Rug. Roger has a dog named Rover.” Rover barks: “Ruﬀ ! Ruﬀ !” The Princess asks, “Can you ﬁnd something Red?” and Rahel uses the camera to photograph a berry on a nearby bush. “Good work! I see a little red here. Can you ﬁnd something big and red?” As Rahel grows, the book asks her to trace not just letters, but whole words. The book’s responses are written on the screen as it speaks them, and eventually she doesn’t need to leave the sound on all the time. Soon Rahel can write complete sentences in her special book, and sometimes the Princess will respond to them. New stories teach her about music (she unlocks a dungeon door by playing certain tunes) and programming with blocks (Princess Rahel helps a notvery-bright turtle to draw diﬀerent shapes). Rahel writes her own stories about the Princess, which she shares with her friends. The book tells her that she is very good at music, and her lessons begin to encourage her to invent silly songs about what she’s learning. An older Rahel learns that the block language she used to talk with the turtle is also used to write all the software running inside her special book. Rahel uses the blocks to write a new sort of rhythm game. Her younger brother has just received his own green book, and Rahel writes him a story which uses her rhythm game to help him learn to count.
The interaction design of the Nell system described above is inspired by the “Young Lady’s Illustrated Primer” in the Neal Stephenson novel The Diamond Age, from whose pro-tagonist Nell takes its name. Nell’s design embodies four key ideas: it is a Narrative interface using Direct Interaction which Grows with, and is Personalized for, the child. Nell uses these four key concepts to build a novel learning platform which addresses several challenges we’ve encountered in earlier work.
Children (like all humans) are hard-wired for stories . In our earlier learning software, we’ve seen some of our favorite pedagogical activities neglected because children had diﬃculty ﬁnding their way into the material without an enthusiastic teacher’s guidance. In response, Nell uses a Narrative Interface [3, 6] to pave a path for the child user through the learning material. In the narrative interface, all system actions are shaped by a storybook metaphor, and interactions with the system are framed as interactions with one of the system protagonists. Nell’s overall story is a multi-character serial adventure, taking cues from The Diamond Age and from the UNIVERSE narrative-generation system . Each of the several characters represents a speciﬁc skill or subject area, and each adventure represents about a year’s curriculum. Adventures are further subdivided into story modules, which match the scope of a traditional lesson plan. The serialized multi-character design improves modularity and allows updates and decentralized authorship. Nell’s characters are always-available agents layered above a particular system activity (application) which provides specialized functionality. The agents are not always foregrounded: constructionist learning occurs when the child plays freely to “make things” with the base activity. The handwriting tutor is fundamentally a drawing activity; the adventure involving the magical musical lock is also a musicmaking activity. The narrative system is hooked into each activity to provide passive guidance (congratulating the child when it notices they’ve drawn a letter), active guidance (Apple-Guide–style  contextual help), or system services (switching activities; jumping into a related story module). A system-wide achievement system provides goals and rewards.
Children learn in diﬀerent ways. Nell provides multiple story modules for its plot points/lessons to engage the child’s interests and cater to multiple intelligences . Selecting between the alternatives can depend on either explicit choices made by the child (“fractions in outer space,” if the child chose a space-themed story or previously indicated an interest in space) or prior success with a given lesson style (a large number of accomplishments in musical-rhythmic tasks suggests a rhythmic approach to fractions). Similarly, several diﬀerent achievements can coexist, rewarding diﬀerent ways of accomplishing the same pedagogical goal. The record of past choices and accomplishments is stored in a Journal and can be reviewed by the child. The contents of the Journal can be rearranged to create a Portfolio  demonstrating the child’s progress. In order to encourage fearless play and experimentation, the Journal also supports pervasive undo; the child can rewind and replay the narrative starting at any past point in the journal. Content can be remixed for further personalization. The child is encouraged to rename characters and change clothing, colors, and other superﬁcial details. In the future we expect to accomplish further narrative customization by recombination of story elements. Lebowitz  and Riedl  show how a planner can be used to recombine and adapt story fragments. We have less ambition than the cited work: instead of attempting to generate thousands of stories from tens of templates, we hope to select and then modestly adapt from hundreds of story modules created in a decentralized manner by teachers—and eventually by the students themselves.
Children grow. Nell aims to provide a “low ﬂoor and no ceiling” to grow with them, along three main axes.
2.4 Direct interaction
The constructionist learning philosophy emphasizes creation and tangible interaction. Nell uses direct interaction on a tablet computer to maintain the child’s connection to their work (Figure 1). In particular, Nell uses handwriting recognition as a primary interface. The direct interaction model lowers the ﬂoor by eliminating the need to learn an abstract touchpad or mouse interface. It also supports our literacy goals by allowing direct handwriting instruction and development of motor skills.
Our implementation choices further our goals of modularity and decentralization. In this section we describe the technologies that enable Nell’s four key ideas. New story modules can be downloaded from the Internet in connected deployments. In disconnected deployments, new stories might be distributed once a year on USB sticks or shared among friends. In a classroom environment, a teacher can share stories for the day’s lesson at the start of class. Story modules can be inherited and extended. This makes it easy to take an existing story and modify it to better suit particular interests, a particular type of learner—or just to add variety. This can lead to conﬂicts: multiple versions of a story, or scenes within a story, may have their preconditions satisﬁed simultaneously. Conﬂicts are resolved by consulting a group associated with each scene. Each group has a default priority and names a scene to be invoked when a conﬂict exists; the actions associated with that scene will (eventually) alter the priorities of the conﬂicting scenes to resolve the conﬂict. The resolution scene may include any number of actions. It may select randomly among the conﬂicting scenes or base its selection on the child’s preferences, curriculum progress, or inferred learning style. The resolution scene may even initiate a new multi-scene story. For example, the story-intro-group used in Figure 2 defaults to a low priority so that continuing a story in progress is preferred above starting a new story. When a story module is completed, several new story openings will be in conﬂict.
3.2 Extensible dialog
Nell’s use as an interactive diary and portfolio  is enhanced by the child’s collaboration in the ﬁction that Nell is intelligent—what Turkle  calls the “ELIZA eﬀect.” Convincing discourse is a hard problem, but reasonable approximations do not have to be diﬃcult; even the rudimentary conversational abilities of ELIZA elicited hours of conversation.We’ve chosen to allow the child to directly converse with the Princess and other agents within Nell. Our implementation extends AIML , the markup language developed for the Loebner Prize–winning AliceBot. Each story module can include AIML fragments extending the conversational capabilities of Nell’s agents. Figure 3 shows an AIML fragment augmenting an agent with commands suitable for a text adventure story module. In such a story the child might write instructions such as go north, go east, or open door in addition to the usual dialog with the agent. These new commands ﬁre events which are referenced by scene preconditions. For example, north may trigger a scene which causes the activity to draw a new location on the page, causes the agent to write and speak a description of the new location, and ﬁnally causes a shift in the AIML topic to enable additional vocabulary suitable for the new location.
We have described the design of a novel narrative directinterface system for education, starting with literacy, which is personalized for and grows with its child owner. Personalized direct interface engages the child, and narrative guides them through pedagogic material. Low-cost solarpowered hardware allows Nell to reach further into the leastdeveloped areas of the world to help those without traditional educational infrastructure.
 A.L.I.C.E. AI Foundation. Artiﬁcial intelligence markup language (AIML) version 1.0.1. http://www.alicebot.org/TR/2005/WD-aiml/, 2005.
 C. S. Ananian. http://cscott.net/Projects/TurtleScript/, 2011.
 J. Bizzocchi. Games and narrative: An analytical framework. Loading—the Journal of the Canadian Games Studies Association, 1(1), 2007.
 B. Boyd. On the Origin of Stories: Evolution, Cognition, and Fiction. Belknap Press of Harvard University Press, 2009.
 A. Chang and C. Breazeal. TinkRBook: Shared reading interfaces for storytelling. In Proc. of the 10th Int’l Conf. on Interaction Design and Children (IDC ’11), pages 145–148. ACM, June 2011.
 A. Don. Narrative and the interface. In B. Laurel, editor, The Art of Human-Computer Interface Design, pages 383–391. Addison-Wesley, Reading, MA, 1990.
 R. E. Fikes and N. J. Nilsson. STRIPS: A new approach to the application of theorem proving to problem solving. Artiﬁcial Intelligence, 2(3–4):189–208, 1971.
 H. E. Gardner. Frames Of Mind: The Theory of Multiple Intelligences. Basic Books, Dec. 1983.
 M. Lebowitz. Story-telling as planning and learning. Poetics, 14(6):483–502, Dec. 1985.
 N. Montfort. Curveship’s automatic narrative variation. In Proc. of the 6th Int’l Conf. on the Foundations of Digital Games (FDG ’11), pages 211–218, June 2011.
 J. Powers. Giving users help with Apple Guide. develop, 18, June 1994. Apple Computer, Inc.
 M. O. Riedl, A. Stern, D. Dini, and J. Alderman. Dynamic experience management in virtual worlds for entertainment, education, and training. Int’l Trans. on Systems Science and Applications, 3(1), 2008.
 M. O. Riedl and R. M. Young. Narrative planning: Balancing plot and character. Journal of Artiﬁcial Intelligence Research, 39:217–267, 2010.
 M. Samuel. Ecmascript quasi-literals. http://wiki.ecmascript.org/doku.php?id=harmony:quasis, 2012.
 J. Skorupski and M. Mateas. Interactive story generation for writers: Lessons learned from the Wide Ruled authoring tool. In Proc. of the 8th Digital Art and Culture Conf. (DAC), Irvine, CA, Dec. 2009.
 E. Stefanakis. Multiple Intelligences and Portfolios: A Window Into The Learner’s Mind. Heinemann, 2002.
 Sugar Labs. http://sugarlabs.org.
 S. Turkle. Alone Together: Why We Expect MoreFrom Technology and Less From Each Other. Basic Books, 2011.
 B. Wilcox. Beyond Fa¸cade: Pattern matching for natural language applications. http://chatscript.sourceforge.net/Documentation/Pattern_Matching_for_Natural_Language_Applications.pdf, Feb. 2011.
SEE ALSO: EMPTY LOT, OCEAN VIEW