A blog post in three parts.
It’s not a secret that education is they key to success in life and just maybe changing the world. But unlike almost everything else, teaching methods haven’t changed in the last century. We still educate using the factory model, or as Ken Robinson puts it, “in batches”. Students sit at the same desks using largely the same tools to learn the same material in the same way. The old system of two-dozen kids and one adult never worked particularly well. The classroom isn’t good at accommodating personalization, even though every child learns differently. Not just at quantitatively different rates but in qualitatively different styles. The only way to bring to classroom into the 21st century is to use – gasp! – technology.
I am a little concerned with handing over our youth’s education to a machine. Isn’t the transfer of knowledge from generation to generation one of the core ideas that make us human? We could invest in hundreds of thousands of tutor-teachers to personalize lessons for each child. It would certainly create jobs, but at a cost of millions of dollars and years of training. So that’s out. But whatever technology we use needs to be intelligent and capable. The best way to do that is with an algorithm.
“Cook for five minutes or until golden brown” is a example of a simple algorithm. Actually, it’s two algorithms: “cook for five minutes” and “cook until golden brown”. The two stopping conditions are crucially different. The former is an inflexible process, while the latter requires a change in behavior based on observation. In a word, it’s adaptive, which is the difference between watching the timer and watching the toast. So for education, we want an algorithm to focus on the brains, not the bits.
The big name is virtual schools is the Khan Academy. There is a fixed tree of topics and prerequisites, instruction comes from founder and polymath Sal Khan drawing on a virtual whiteboard on YouTube, and assessment is solving ten cookie-cutter problems in a row. There’s been a fair amount of wrath against Khan, which boils down to this: he’s using new tools to do old things. The academy doesn’t observe the students’ actual comprehension, only how much time spent jumping through technological hoops. His algorithm doesn’t adapt.
I’d be remiss to discount virtual schools in general because of one bad apple. School of One is a NYC-area startup that is managing the classroom differently. In addition to large or small group instruction, School of One add other “modalities” like virtual instruction, independent practice, and and group collaboration. Teachers have tried (most of) these techniques before, but School of One’s algorithm tracks how each student performs in each modality and creates a personalized lesson plans accordingly. Beyond structuring what is learned, School of One cares about how the students learn it. Freakonomics likens it to Pandora internet radio, which learns a user’s music preferences and plays only songs they like.
While the algorithm does its best to adapt to student needs and provide lesson plans, at no point in the process are teachers unavailable. Students can always ask questions, and teachers can override the algorithm’s proposed schedule. (You would’t begrudge your captain the use of autopilot, but you still want him in the cockpit.) However, as much as School of One empowers teachers and extends their reach, it can never be fully automated as the Khan Academy is. Their website claims anyone can “learn almost anything for free”, which apparently means “watch a video and take a test on almost anything for free”. School for One is currently available only to New York elementary schoolers. Meanwhile, the jury is still out as to whether any algorithm can be sophisticated enough to adequately replace a teacher.
School of One exemplifies not only the future of education, but the future of a computer-aided — but not computer-replaced — workforce. The algorithm helps us focus on the important things, the ones that make us human.
* * *
Tom Vander Ark (Getting Smart, 2011; TEDx talk) is an advocate of “personal digital learning” in both the U.S. and the developing world. He’s praised School of One for its customization of learning, and looks forward to the equalization of education borne on the wings of tablets. The third -ation trend he’s interested in is motivation, which Vander Ark seems to largely equate with video games.
I won’t deny that games can be tremendously motivating — too motivating. Ian Bogost created Cow Clicker as a satire of Zynga’s social games, and hundreds of people dutifully clicked cows every six hours for months until he pulled the plug. As the excellent Wired article somberly concludes, “we were clicking nothing the whole time. It just looked like we were clicking cows.”
All too often, video games become become a black hole for time while bearing no relation to real-world problems. Then, once the artificial motivation is removed, kids go through a drug-like withdrawal period where they are unable to motivate themselves. Dan Pink (Drive, 2010; TED talk) writes about extrinsic vs. intrinsic motivation. Although he focuses on human resources and management, his ideas are also applicable to games. Without a lot of careful planning, video games and badge systems (Khan Academy, I’m looking at you) actually decrease motivation as soon as the challenges get mildly difficult. This revenge effect is related to the well-documented psychological phenomenon that students praised for the intelligence (how many points they earned) have less persistence than those praised for their effort. Students may fall into the trap of “I beat the game, leave me alone!” and miss that the educational material is important outside the game. Games present learning as a means to an end, not an end in itself, and the games may themselves be gamed.
What Pink calls the “mechanistic, reward-and-punishment approach” that is inherent to video games often doesn’t work. Specifically, it falters when problems are open-ended and vague (wicked problems, engineering projects) instead of formulaic and well-defined (worksheets, multiple choice).
Another way that video games fail to motivate us after the game is over is reward inflation. Once you’ve saved the earth from hordes of alien invaders, doing a problem set is beneath you. It doesn’t matter if the game made you solve a math problem to fire your gun – “I beat the game, leave me alone!”
Using shoot-em-ups to funnel pubescent male exuberance into math is a mistake. Any shooter involves an element of split-second timing. The ticking clock causes students to fall back on their incorrect instincts and preconceptions rather than replace them with something new, i.e. learn. Eliminating timed trials from games is one way to promote the element of forgiveness. (Axing streaks of ten correct problems is another, Khan.) STEM fields can be surprisingly forgiving of mechanical mistakes at the professional level. Even the most seasoned programmer gets compiler errors.
Math is not a summer action flick; it’s more like a drama or mystery. Math games should be of a genre that promotes calm planning and critical thinking – puzzles. The MIND Research Institute (TEDx talk) has a series of puzzle games that present math problems wordlessly and graphically. Students use math to help a penguin traverse an obstruction. For example, students choose the right shape to fill a hole. “The results are consistent and pretty amazing,” Vander Ark writes. Use of the games for 90 minutes a week “typically doubles math gains.”
Pink can help explain why. He writes that the most powerful intrinsic motivators are autonomy, mastery, and purpose, which the MIND puzzles offer. Students have the autonomy to try different solutions and see what works, without fear of upsetting a teacher or being made fun of. The forgiveness creates an academically safe environment for students take risks. Mistakes are a natural part of puzzles, and the learning process. This leads students to pursue mastery of the material. They are not discouraged by wrong answers. While they are given the superficial purpose of helping a penguin cross the screen, the game creates the mathematical atmosphere of “try something and see what happens”. The visual puzzles help make a notoriously abstract subject concrete. When the symbols are introduced, students gratefully accept them as labels for concepts. Those funny squiggles, they realize, explain why and how the math works as it dances in front of their eyes.
* * *
I wasn’t taught using puzzles or videos or modal learning. My inspiration for being an engineer comes from doing FIRST robotics in high school. FIRST is an engineer recruitment program thinly disguised as a sport for robots. Scenes from a typical work session:
Tools, lubricants, scrap metal, and engine parts line the walls and overflow the shelves. The “wall of shame”, the collection of parts from prior robots, hangs from the ceiling. The kids can’t keep quiet long enough to let the team leader explain what’s going on. Half of them are sent inside to read the website with posts about what happened at previous sessions, which they should have read at home. We had a piece machined from two years ago to hold a part in place, and we get halfway through making another before we find it. The welder stands around, waiting for metal to be cut (accurately). Everyone exchanges playful insults and dirty puns. In the corner, a NASA engineer talks about the engineering problems with the shuttle’s replacement after management chose a poor design. At one point, the team leader’s wife brings out old photographs of him with hair down to his shoulders in front of a science fair backboard. One group uses general purpose tapping lubricant instead of the aluminum-specific kind. Inside, I instruct the 3D modeling team how to take a sphere, squish it, and cut it into the shape they want. The kid with the mouse makes giant spheres of arbitrary size and position. When I tell him to make a guide to help place the sphere’s center, he says he doesn’t know how to do that. One of the programers is giving his variables meaningless names using Greek characters. The CAD crew can’t keep track of what has and has not been modeled already. During break, the kids play Yu-Gi-Oh cards while the adults discuss the robot design. One kid won’t give up the idea of building a helicopter. In summary: it’s a stew of chaos and inefficiency that I find cringe-worthy.
Notice that I called FIRST “my inspiration”, not “my education”; an “engineer recruitment program”, not an “engineer training program”. FIRST is not an educational program. Namely, it’s For Inspiration and Recognition of Science and Technology (emphasis mine, capitals theirs). It inspires young people to achieve excellence, and recognizes those that do, but it doesn’t provide instructions to get from one to the other. As such, it works opposite the educational programs discussed above. Instead of Pink’s autonomy, mastery, and purpose, FIRST runs on teamwork, discovery, and fun.
There’s a balance to be struck. If the robot can’t move, that’s no fun, but if the kids never touch the metal on “their” award-winning bot, that isn’t any fun either. If the robot doesn’t work, that’s because no one achieved mastery of a crucial component, but if the kids don’t get their hands dirty they’ll never make discoveries. Someone needs to have enough autonomy (and initiative) to set up a file-sharing system that facilitates teamwork.
That kind of work ethic (and play ethic) does not come naturally. Even though AutoDesk or LabVIEW won’t take your finger off like a bandsaw or mill, students should learn to treat them with the same respect. FIRST teaches the engineering mindset more than any particular skillset. Moreover, it teaches how to balance productivity with fun – and shouldn’t fun play a role when deciding what children are going to be doing?