The Misunderstanding About Education That Cost Mark Zuckerberg $100 Million

Personalized learning can feel isolating. Whole class learning can feel personal. This is hard to understand.

Last week, Matt Barnum reported in Chalkbeat that the Chan Zuckerberg Initiative is laying off dozens of staff members and pivoting away from the personalized learning platform they have funded since 2015 with somewhere near $100M.

CZI’s shift in approach marks something of a coda to an era when various advocates and funders believed that computer-based “personalized learning” could dramatically improve education. Summit, CZI’s pet project, has not spread as far as once hoped, and there’s little evidence that it or similar efforts have led to the large learning gains that Zuckerberg envisioned. This gap between ambitions and results underscores the difficulty of using technology to dramatically improve America’s vast system of decentralized schools.

On Twitter, the social network where people are particularly sensitive to the hubris of the billionaire class, commenters have received Barnum’s article fairly gleefully. I’d like to go farther than glee, however, and name the single misunderstanding about education that, more than any other, led Zuckerberg to make this investment which he is now writing down.

This is the Summit model, which I have witnessed firsthand:

Summit also featured 16 hours a week of “personalized learning time.” Students worked at their own pace on a computer, which fed them a “playlist” of content where they learned specific skills. Students could move on once they got eight of 10 questions right on an online quiz.

That seemed to be the biggest draw for Zuckerberg, who contrasted the approach to “having every student sit in a classroom and listen to a teacher explain the same material at the same pace in the same way.” He suggested this could lead to transformational improvements in student learning. The goal, he wrote in 2017, was “scaling this approach to every classroom.”

There it is. You see this misunderstanding time and again from the people who influence whatever passes for a national strategy in education technology.

In 2016, for example, the Gates Foundation announced a fund for personalized learning under the same premise:

[Personalized learning] allows students to progress through content at their own pace without worrying about being too far behind (or ahead) of their classmates.

Sal Khan echoed this premise in 2018:

For us, personalization is — and we could talk about the different flavors of personalization that people use out in the world — but for us, it is, you learn at your own time and pace.

I have tried to illustrate as often as my subscribers will tolerate that students don’t particularly enjoy learning alone with laptops within social spaces like classrooms. That learning fails to answer their questions about their social identity. It contributes to their feelings of alienation and disbelonging. I find this case easy to make but hard to prove. Maybe we just haven’t done personalized learning right? Maybe Summit just needed to include generative AI chatbots in their platform?

What is far easier to prove, or rather to disprove, is the idea that “whole class instruction must feel impersonal to students,” that “whole class instruction must necessarily fail to meet the needs of individual students.”

For that proof, I only need to raise the existence of a single classroom where the teacher interacts with the whole class in ways that feel personal and responsive to the individual students. Here is one of thousands operating daily.

---

In our take on the classic Pool Border problem, students eventually write an algebraic equation to represent the number of tiles around a pool of any size. Liz Clark-Garvey of New York City Public Schools starts the lesson with a whole class move that supports the personal learning of every student in the class.

She flashes a pool and its border to the class for less than half a second¹. Then she asks, “What did you notice?”

Students respond in ways that are, again, very personal, and they benefit from seeing from one another’s vantage point:

• It was pink around the edges.

• It had gray inside.

• It looked like a reverse steak.

• There were 16 squares.

• It was a square on the outside.

• It was a four by four square.

Liz then asks the class to use these new shared experiences and ideas to “try to come up with a shortcut to figure out how many blue tiles there are.” Students then take time on their own to figure out how many tiles are around several different pools, each time getting automated feedback on their efforts. 

Using those experiences, students come up with different shortcuts that are, again, personal to them. Liz spends the rest of class:

• crediting students for their shortcuts,

• grouping students together under different shortcuts, 

• helping students see the connections between their shortcuts

• helping students turn their shortcuts into algebraic notation.

She periodically moves between desks but the majority of this work happens from the front of the room with Liz speaking to the whole class in ways that still feel personal.

---

Whole class work can feel personal. I consider it proven by Liz. QED. 

More students prefer learning in classes like Liz’s than plugged into their laptops with headphones, alone together. I have not proven this. It isn’t obvious to me how one would prove or disprove this, but I find evidence and experience pretty compelling here.

Would Zuckerberg have invested $100 million into Summit’s personalized learning platform (and then more into AltSchool’s personalized learning platform, now dissolved) had he simply visited Liz’s class instead of a class at Summit? My guess is, yes, he would have still invested in personalized learning anyway.

It is an unfortunate fact of modern life that such a small number of people who know so little about the wide range of learning experiences in the world can so drastically affect their course at so little cost to themselves. $100 million was less than three tenths of one percent of Zuckerberg’s wealth in 2015.

It doesn’t matter that this particular marriage of technology and schooling has always ended in divorce, dating back to early programs like TutorTexts, which Audrey Watters describes in her book Teaching Machines:

The book provided lessons in “small units,” followed by multiple choice questions “which the reader must answer in order to proceed further in the book. A wrong answer leads to more discussion of the same point of information; a correct answer leads to the next unit of information and the next question.”

That was 1958. A CEO named Tigran Sloyan predicted in Fast Company last week that generative AI will finally (this time!) “make personalized learning accessible for all.”

The idea that computers should personalize instruction, flattening the human differences that learners would much rather see celebrated and developed, maintains an absolute stranglehold on the imagination of the billionaires who fund education technology. Small children will touch a hot stove only once yet billionaires will fund personalized learning initiatives again and again and again and we might wonder why. What accounts for the appeal of this idea? 

We might also wonder why we’d let any one person have such power to transform the landscape in which teachers teach and students learn. I can console myself somewhat when the whims of these funders align with what I, Dan Meyer, normatively think is good for teachers and students. But I’d be far happier if far fewer teachers and students were subject to those whims at all.

1. This is a routine called Contemplate then Calculate from Grace Kelemanik and Amy Lucenta, one which marries pedagogy and technology together in a way that is simple and thoughtful and reliable and absolutely ruthless in its effectiveness. It is so hard to screw this up.