The First Wave of Personalized Learning
“No two students are alike so no two students should be learning the same thing or way.” That’s the key statement underpinning the $180 million raised by the NY-based Knewton since its founding in 2008. Knewton founder Jose Ferriera described his company’s concept in a 2015 story at NPR, “We think of it like a robot tutor in the sky that can semi-read your mind and figure out what your strengths and weaknesses are, down to the percentile.” A few weeks ago, Knewton was acquired for less than $17 million.
Meanwhile, Silicon Valley tech billionaires, including Mark Zuckerberg, poured $174 million into AltSchool. Last valued at $440 million, AltSchool folded this summer and is rebranding as Altitude Learning, re-shifting focus to selling software and providing professional development services.
Why is this happening? Ferriera and others treated education as a tech problem: the human mind as a bunch of algorithms to be optimized. This led to the emergence of new blended learning models such as the flipped classroom and station rotation. It also led to a big push for self-paced student-centric learning, or as I call it, the first wave of personalized learning. What these solutions failed to take into account, however, was an a priori understanding that learning is a social activity — most students learn best from human-to-human interactions with teachers and peers. No matter how powerful your algorithms are and how advanced AI becomes, the human interactions and relationships in the classroom are the most important pieces of learning.
ChalkTalk and the Second Wave of Personalized Learning
When I first started ChalkTalk, my engineering background led me to fall into the same trap as others. We set out to build the best adaptive robot overlords in the sky, treating the personalization of classroom learning as a complex math problem. Ultimately, we built a software that administered highly personalized practice questions and instructional multimedia content to students based on their performance on an initial summative assessment and ongoing formative assessments. The goal was to bring the experience of a 1-on-1 tutor for every student, despite their socioeconomic backgrounds and starting proficiency levels.
There was one problem with our idea: when we deployed it, it worked well for only about 8 percent of students.
Obviously, this was not the success we had envisioned. But as it turns out, we were still doing better than Massive Open Online Courses (MOOCs), the education movement of the future. 3.13 percent of MOOC users completed their courses in 2017–18, down from about 4 percent the two previous years and nearly 6 percent in 2014–15—and that didn’t take into account whether or not they did well in the course. This is perhaps why after the New York Times trumpeted 2012 as “the year of the MOOC“, it was only a year later that the same publication declared MOOCs had “turned into a flop”.
We could have moved forward on this vanity metric and stated that our software “delivered results at twice the rate of MOOCs,” but instead decided to step back and identify the problem at its root cause.
Meeting Teachers Where They’re At
After our first setback, instead of approaching the problem of personalized learning with my training in computer engineering — with algorithms, metadata, tags, taxonomies, and AI — I decided to try to solve it as a teacher. I reached back to my experience teaching courses in my high school, university, and after-school learning centers and asked myself: “if our goal is to improve education in classroom education, then what are the ‘building blocks’ that make up a classroom course?“ Once we had those identified, I knew we could start optimizing them.
To answer that question, I applied “first principle” thinking to break down a “course” into three constituent blocks:
- Whole group instruction: the teacher lectures to the entire room, often while writing or projecting instructional notes on the board;
- Collaborative learning: group projects, worksheets, or something as simple as a single walkthrough example that the teacher jots on the board and asks students to solve in pairs or small groups during class time; and,
- Individual practice: homework assignments or in-class practice.
These blocks were glued together with:
- A schedule and lesson plans to organize the scope and sequence; and
- A set of formative and summative assessments to measure progress.
…and there you have it, the key building blocks that make up a classroom course and the adhesive glue that holds them together.
It was becoming clear to us that we didn’t need to reinvent the wheel; rather, we needed to optimize each of these three building blocks and the glue that connected them with technology.
So we built on top of our existing platform, created something brand new to replicate this experience, and launched it at our first pilot site, East Boston High School, the largest high school in Boston Public School District. The first class ever, an SAT prep class for 11th graders, was chaotic. Teachers didn’t know how to use the instructional materials and quickly got lost, students were confused on how and when to use the software, and we were running behind schedule. So I jumped in and got permission from the school administration to teach the class myself. Every day after teaching, I’d speedily jot down notes on my phone in the Uber back from school citing user experience improvements, software bugs, and content delivery methods.
After 3 weeks, the software was nearly unrecognizable from our starting point and I decided to let the teachers teach the rest of the course on their own. I shadowed them for 3 days and then was able to stop coming to school for the course period. I also rolled out the program to our second pilot site, Summit Academy Charter School in Brooklyn. At the end of the year, East Boston High School moved from the 3rd to the 22nd percentile on the Massachusetts Comprehensive Assessment System and had the highest score improvements in the SAT since the school’s founding. In our second pilot site, Summit Academy, 100% of the students got admitted into college and the school had more students scoring greater than 1000 on the SAT since the school’s founding (even with an SAT prep class in place for 5 years prior).
Through this experience and insight, we designed ChalkTalk to optimize the current classroom experience, starting with assessments, and then generating curriculum designed for the classroom as a whole, small groups, and individuals. More than 80 percent of teachers who have tried ChalkTalk end up utilizing it in more than 90 percent of their classes. Why? Because we aren’t asking them to change the way they do things — instead, with ChalkTalk, teachers teach the way they want to teach and watch their results go up! Most notably, those results include student score improvements 2–6 times greater than those of students who don’t use it.
What we learned at ChalkTalk is a lesson that sounds so simple and easy: learning is social and education is human. Applying a Silicon-Valley-driven engineering mentality — turning education solely into 1s and 0s — just won’t work; it’s much harder than that. We had to find a way to create a personalized learning solution for every student without having students’ eyes glued to screens and losing the social nature of learning. Human motivation, discipline, and encouragement: those are all crucial things that teachers bring to the table and things that can’t be replaced by algorithms, gamification systems, badges and points, remote Skype calls, or MOOCs. If education was simply about maximizing content distribution, whether software or printed books, we would long ago have sent students home with 12 years worth of books and invited them to take their finals at the end of their schooling.
Clearly, learning is about more than content administration. We don’t remember our favorite textbooks and publishers — but we do remember our favorite teachers because ultimately, education is not content, education is not technology: education is human.
Also published at https://medium.com/@moarbaji