The 2 Sigma Problem in Education

Students who receive one-to-one tutoring, on average, perform at the 98th percentile of students who learn in conventional classrooms.

A tutor in a one-to-one setting more fully understands a student's misconception from their deeper interaction and then can address it immediately and directly. Students get corrective feedback for their efforts instead of getting a score a few days after a formal assessment.

If a student has gaps in required prior knowledge, a tutor addresses those gaps before teaching more advanced concepts that build upon such knowledge. This personalization of pace and content allows students to build upon a solid, not a shaky, foundation.

But one-to-one tutoring is expensive for most societies to bear, and there aren't enough educators to provide one-to-one tutoring on most subjects. It also lacks the rich engagement between students in a classroom that offers a diversity of perspectives for students.

How can we create a group instruction method that is as effective as one-to-one tutoring?

This is the “two sigma problem in education,” where two-sigma refers to a statistical measure equivalent to the 98th percentile difference in outcomes.

Benjamin Bloom, an eminent academic researcher, known to every educator for creating a common language and framework for learning and assessments known as the Bloom Taxonomy, published the above conclusion in 1984 in his landmark paper.

Unfortunately, the modern education system, developed at the dawn of the industrial age, is like a conveyor belt. We expect a large group of unique individuals with significant differences in prior knowledge, interests, and needs to learn in unison under the guidance of one teacher. It is not hard to imagine the mediocre outcome for most students, yet it is the only affordable and scalable education system globally.

Imagine the impact on a nation's human capital if every student was able to learn to their full potential.

My personal and organizational north star goal is solving the two sigma problem in computer science education. The strong and growing need for computer science education combined with a shortage of qualified teachers creates a natural inequity in access. It provides the perfect socioeconomic incentive to find a scalable solution.

Computer Science education feels personal. I was fortunate to have access to computer science starting in middle school in a small remote town in India. It changed the trajectory of my higher education choices and career. I became the first in our extended family to pursue a rewarding career in engineering instead of inheriting a small family business.

2Sigma School has been teaching blended classes in computer science at several high schools as their external instruction partner over the past year. These classes have provided our instruction team and me with critical feedback on how to address student learning and their engagement and motivation.

We are reimagining conventional education!

Our reimagination of the classroom is starting to take shape as we challenge ourselves to address the unique needs of every student. Identifying and solving scalability bottlenecks in personalizing to every student across a diverse range of backgrounds has been humbling. It doesn't have an engineering-only solution, but one that has to emerge from real-world experience.

Creating a personalized experience like that of one-to-one tutoring in a group setting requires that the teacher have time to engage with every student and address their unique pacing and learning needs. It is impossible if the teacher delivers lessons as lectures more than 80% of the time. There must be ample time for the teacher to walk around (virtually), help every student make real progress, and be comfortable managing every student's learning at a different pace. The sheer overhead of tracking progress can be overwhelming for the teacher. Existing learning management systems aren't designed for every student to take an individualized learning path at their own pace.

Timely constructive feedback to students is one of the critical reasons for the effectiveness of one-to-one tutoring. It requires the teacher to be highly tuned to the student's work so they can proactively ask the right questions and provide the necessary support. Such real-time observation of every student's work is all but humanly impossible in a group setting.

Now is the perfect time to solve the elusive 2 sigma problem in education.

There are two significant inflection points, one in enabling technology and another in user behavior, that will allow two sigma improvements in education over the next decade. First, artificial intelligence has matured sufficiently to provide a human-like understanding of complex, unstructured tasks, as we see with autonomous driving and conversational chat-bots. It is critical for real-time inference of when and where a student needs support and automatically delivering human-like support where it can be effective.

Second, the recent global experience with remote learning has changed everyone's imagination about what is possible in education, with the accelerated adoption of digital technologies inside and outside the classroom. It will pave the way for AI to seamlessly integrate with students' learning and will be critical to elevating the role of the teacher to more impactful tasks.

So what will a two-sigma classroom look like in a decade?

We are actively imagining and shaping the future of education, starting with computer science, and are privileged to be working with students in traditional classrooms where 99.9% of students are today.