Will AI solve the 2-sigma problem in education?

This article ties in closely with my previous blog called Learning Agency: The Role of Students, Educators, and EdTech Providers.

But first, a little introduction to the 2-sigma problem, for those of you not familiar with it.

In a landmark 1984 paper [1], educational psychologist Benjamin Bloom along with his team of graduate students demonstrated that 1-on-1 tutoring significantly improved student outcomes, with those receiving 1-on-1 tutoring achieving nearly two standard deviations above the average student in a conventional classroom setting (without 1-on-1 tutoring).

In other words, this means the student group receiving 1-on-1 tutoring performed above 98% of the non-tutoring group.

This. Is. Huge.

However, we know all too well that such 1-on-1 tutoring is very expensive and therefore not accessible to all pupils.

So with the advent of chatbots like chatGPT passing high-stakes exams with flying colours, it is exciting and motivational in some way, to think about the prospects of inching closer to solving the 2-sigma problem by building AI tutoring systems.

Fair enough.

Gen-AI based tutoring systems can offer an engaging and highly personalised, conversational learning environment.  Indeed similar to chatting with a tutor. It has the potential to support the idea of ‘sustainable education’ by improving access to personalised learning without the need for a tutor and improving targeted interventions based on data-driven decision-making. Some features that replicate tutoring would be:

• Keeping students engaged in the session through frequent prompting, and responding in a customised manner
• Reinforcing learning with immediate feedback on both what has been learnt and where the gaps in learning are
• Recommending learning materials according to the student’s ability
• Allowing practice in their own time, at their own pace.

But I see a few big misconceptions brewing here and there’s a big missing piece that’s being overlooked.

Over 60% of school going children globally are not connected to the internet at home.

Teaching is not about passing the right information on to students or guiding students to the right answers to questions.

Teaching is about leaning into pedagogical principles to promote lifelong agency for learning among individuals.

Teaching is about relationships, being human and providing that personal touch to mould a successful professional, a responsible citizen, and a passionate human being.

Some of the people who had the biggest influence on my academic and professional life were teachers from my school and university lecturers.

I don’t remember them or look up to them for how well they imparted subject knowledge but instead, it’s for supporting me personally, when I needed additional emotional support, helping me believe in myself and my abilities, and somehow cultivating a sense of wanting to learn - learning agency - which has stayed with me till today! They became life-long mentors.

Will having access to an AI tutor app anywhere a student goes be helpful?

Absolutely.

However, using such a tool is closely tied to the student’s agency to learn. Pupils who are already motivated to excel in exams will easily take advantage of these tools to learn anywhere, anytime. This will not lead to a true democracy in learning. And I’m pretty sure that this was not the original intent of these tools.

Ultimately, while AI can certainly augment and enrich tutoring sessions or indeed provide a more personalised experience within the classroom, it cannot fully replicate the irreplaceable role of a human teacher. Educators bring a unique blend of empathy, adaptability, emotional support, and holistic guidance that AI simply cannot emulate. They inspire, motivate, and mentor students in ways that extend far beyond the confines of academia.

[1] The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring on JSTOR

[2] Benjamin Bloom

‍