A Northstar for the 20-Year Future of Education with AI

A WORKING PAPER TO OPEN THE CONVERSATION

Ikhlaq Sidhu, Dean, School of Science & Technology, IE University, Madrid

October 1, 2023

Thank you to many from past and recent conversations including Esther Wojcicki (Woj), John Maeda , VP Design & AI, Microsoft, Edo Segal , Founder CEO of Touchcast, and IE University's Executive leadership including Santiago Iniguez and Diego del Alcázar Benjumea , for everyone's collective support and insights.

Introduction

This year, virtually everyone was hit by the shockwave of AI and large language models. Since then, we have seen a lot of discussion about the future of education with AI, ranging from blocking its use to the more progressive AI tutor recently introduced by Khan Academy.

However, by looking only at what technology can do today doesn’t fully answer the question of what will happen next. To understand the future of education, we must consider scenarios where humans can no longer distinguish between AIs and other humans, even while acting as classmates in study sections, offering role play as experts, and augmenting the teaching staff. In that model, humans and AIs would achieve a symbiosis further fueled by gamification, curated content, and greater spatial immersion - all of which would lead to a new paradigm in lectures, discussions, and projects. Of course, this is only one scenario.

Educational Northstar

In this working paper, we extrapolate the future trajectory of education, drawing our projections from the dual perspectives of historical retrospection and advanced technical possibilities that have surfaced in recent years. We intend to identify a 'North Star’ as a guiding beacon to shape and navigate the evolution of educational paradigms over the next two decades.

The Challenge:

In May 2023, Sal Khan of Khan Academy challenged us all about this graphic of the two sigma problem, which states that 1-1 tutoring has been the most effective teaching method but is not possible for mass education due to the cost of instruction. He then asks whether technology could now make it possible to achieve the 1-1 learning outcome at the cost of the mass education model. Although he offers a point solution as an impressive AI tutor, this paper offers a more multi-dimensional analysis of that question and seeks to understand where we are likely to be headed in the next 20 years and in which ways these technologies will provide the most benefit.

By analyzing our educational past, we gain insights into patterns, successes, and failures, leveraging them to anticipate and innovate for the future. Concurrently, new technologies provide unprecedented opportunities to reimagine and reconstruct educational frameworks, tools, and environments. However, a striking contrast exists between the rapid pace of technological advancement and the slower, more gradual adaptation of human behaviors and societal norms.

This paper aims to merge past perspectives with progressive innovation to develop a visionary path for education with AI.

Historical Context in Education:

Let's start with the Roots. We started 1:1 and have since then moved to a factory model:

We begin by examining the historical evolution of educational models. The earliest models from institutions like Oxford, dating back to the 1200s, fostered close, personalized interactions, employing a faculty-to-student ratio of 1:1 or up to 1:6, prioritizing individualized learning experiences.

Conversely, in Germany, Frederick the Great, with Johann Julius Hecker, established the foundations of the Prussian primary education system in 1763 with the Generallandschulreglement decree; Hecker founded Prussia's first teacher's seminary in 1748. This has influenced the American education system and has independently evolved to an approach emphasizing efficiency, sometimes called the “factory model.”? Here, the technical education model boasted a 300:1 student-to-teacher ratio, primarily emphasizing delivering lectures by distinguished educators.? This method was grounded in the philosophy that dialog was unnecessary at the earlier educational stages, proposing that effective knowledge transfer could occur through listening to accomplished lecturers.

Today, the educational spectrum has broadened significantly, incorporating diverse options. This range spans from intimate, dialogue-rich small classes in private schools to expansive, lecture-based learning environments, embodying a synthesis of historical educational philosophies. This multiplicity of choices reflects an evolved understanding of educational delivery methods.

Not Every Classroom is the Same:

Education is multifaceted, comprised of various class types, each designed with a unique purpose, reflecting diverse learning needs and objectives.

• Lectures, reminiscent of the German model, are apt for delivering fundamentals and theories.
• The Socratic method caters to discussions and critical thinking, often employed in settings like case studies, allowing up to 50 participants per class to engage deeply. *
• Labs focus on the practical application of theories, and their capacity is typically constrained by available equipment.
• Research papers and theses undertaken under faculty advisement, align most closely with the Oxford model, emphasizing one-to-one learning experiences.
• Additionally, project-based learning spans from synthesis to creating new ventures, fostering a blend of skills and knowledge application. Projects add a team dimension to the work, including hard and soft skills to deliver the project results.

Each model will continue to evolve, contributing to the holistic educational ecosystem that accommodates varying pedagogical goals and learner preferences.

The Horizon of Technological Change:

Technological advancements have the power to reshape the educational landscape monumentally.

• Video: We can start with the inception of video platforms like YouTube.? When used for education, the platform has been pivotal in democratizing learning, marking the initial transition from traditional models.
• The MOOC: Going beyond watching video lectures, there has been the evolution to MOOCs (Massive Open Online Courses), offering broader access, though they often grapple with low completion rates. A pivotal factor in enhancing student commitment is a faculty member who maintains relationships and serves as a role model, altering students' engagement substantially.
• The Internet as a library: The Internet serves as a colossal repository, housing millions of human-authored pages, offering in-depth insights on myriad topics, which are now effortlessly searchable by both popularity and relevance.
• Generative AI has produced an Oracle that has read the Internet library: This immense web can be encapsulated into generative AI models, trained on vast textual data, enabling them to respond based on the information they’ve processed. These AI entities can function as coaches, assistants, or experts. They can translate knowledge from multiple functional perspectives. They can possess controllable memory capabilities, allowing continuous, coherent interactions to form.
• Multi-agent AIs: AIs Can Interact with each other and with humans. The advent of sophisticated AI technologies enables these models to interact with one another and participate in group discussions, paving the way for multi-agent models to function with specific objectives. This interaction isn’t confined to AI entities alone; humans can join these group interactions, merging education, work objectives, and entertainment.
• Gamification: The Internet is rife with various algorithmic models, meticulously designed to track, predict, and gamify services. We must also consider that gamification in social media has been addictive and potentially leads to mental stress. However, these models hold significant potential in testing and applying within the educational domain.

Of course, this only captures what has happened up to the time of this writing. If we look forward 20 years, we can expect over 20 product cycle iterations and perhaps 2-4 disruptive waves. Our estimation of capabilities in 20 years is likely to be that:

1. Web page search, Video, MOOCS, and Gamification are already in a moderate to mature phase of the technology S-curve (10-20 years old), so the increased capability will likely be in processes or economics. Meanwhile we expect internet coverage to incrementally strengthen in emerging and established markets.
2. Language Models, AIs, and multi-agent capabilities are very early in their innovation cycles, so in 20 years, we will likely expect near perfect presentation of human characteristics, difficulty in distinguishing humans from AIs, more real / immersive experiences, and greater cognitive powers that will include self correction, self reflection, objective-orientation, synthesis of logic with language, new policy regulations, and more.

These ongoing innovations signify a harmonious integration between humans and AIs, depicting a future where learning is not just confined to human interaction but extends to include interaction within groups of intelligent agents. The convergence of all these technologies leads to uncharted territories in learning methodologies.

Before we Reinvent it, let us Understand what Education is:

It would be wrong to think that education is simply a matter of watching a lecture and learning something about it.? Education is a multifaceted entity, traditionally perceived as a journey through school or college. However, this conventional view often overlooks the holistic value beyond academic learning, which imbibe the following:

• A designated culture
• A social environment fertile for forging lifelong connections
• An employment-oriented hub
• A custodianship that allows students to be safely cared for and educated during large portions of time and
• A realm for sports and physical activities.

The purpose of education has also been a moving target, ranging originally from serving the elite thinkers and then broadening to include a vocational audiences:

Should the University be designed to educate an elite set of deep thinkers, or should it be a mass market system that provides vocational training for the masses? Historically, education’s primary goal was cultivating a select few as deep thinkers, and not focused on those requiring professional training.

However, contemporary education seeks to amalgamate job-oriented training with profound intellectual development, blurring the lines between vocational and philosophical education.

It is yet unclear if there will be a segmentation of the educational system or whether universities will have to continue to serve both the elite thinker and mass vocational market together.

Before we understand the Future of Education, let's predict the future of Work. The knowledge worker’s job will likely remain the same, but empowered with staff of AIs.

Indeed, in crafting an education that will effectively serve future thinkers and workers, an understanding and anticipation of the evolving nature of work are crucial, particularly in a landscape incessantly shaped by daily technological advancements.

The hypothesis here is that behaviors of people and their job functions may change less than we expect, however the amount of interaction and assistance from cognitive machines will change the nature of the work and the expected output dramatically. It will change the core skills that people will need to be competitive with each other. And will in some manner set up a competition for complimentary skills between humans and machines. This introduces a new dynamic where humans and machines coexist in a competitive yet complementary symbiosis.

It’s imperative to align educational and training initiatives with these evolving requirements, preparing individuals for this future landscape where cognitive machines are integral components of the professional ecosystem.

What Can We Expect?

The Education North Star guides us toward a future where education is a harmonious blend of evolving technologies and human adaptability, promoting lifelong learning, critical thinking, and innovation. It fosters a holistic, inclusive learning environment, adapting to each individual's diverse, changing needs and potentials.

Will the Model of Education Change due to Technology?

The onset of technological advancements raises contemplations about their potential impact on these traditional student experiences. The uncertainty revolves around whether technology can significantly reshape or help disaggregate the business model of the traditional educational institutions, which have withstood the metamorphic tests of over 800 years. Our hypothesis is ‘No”, because of the non-academic societal expectations on education. In this case, technology is more likely to be incremental than completely disruptive to the educational model.

A Key Question: Will Everything become Virtual or Will the Campus Continue?

Hypothesis: Maintaining a physical presence and a traditional campus in the future of education is pivotal. A wholly virtual university model seems unsustainable.

Although a virtual institution may manage to formulate student cohorts and maintain careful selection, it likely falls short in delivering extracurricular experiences and fulfilling the societal expectations of custodianship.

To validate this hypothesis, one could examine student engagement, satisfaction levels, and the holistic development attained in virtual versus physical educational environments, comparing outcomes, societal integration, and the student's overall success rate and well-being.

As a shortcut, our experience with the all-virtual learning experience of the pandemic of 2019 indicates virtual education's lack of success and its side effects.

A Methodology to Understand What’s Next in Education:

Several pivotal questions arise in contemplating the future evolution of each education segment like theory, Socratic, labs, and project/thesis.

• How will the professor's role transform, and what will AI's role encompass in the educational journey?
• What will be the role of AI agents in each model?
• How will assessments adapt, and which components can optimally function in physical and virtual realms?
• What parts can be physical and virtual
• What will the learning group likely become
• What tech tools are most likely to benefit the learning experience

These considerations will be integral in shaping an adaptive, inclusive educational paradigm, balancing technological advancements with human interaction, tailored to diverse learning needs and preferences.

An Analysis in Table Format

The table below characterizes each learning style in a different column with the analysis questions in each row. Entries in the table are estimates based on logic and judgement.

A list of assumptions to be tested:

• Many cells in the above analysis table are estimated by judgement and all deserve scrutiny.
• Will students accept interaction with virtual professors, virtual tutors, virtual classmates, and virtual experts? We assume so, but in varying degrees, which are not clear now.
• We assume the limiting scale of the educational outcome will depend of how much instructor’s interaction/relationship with the student can be scaled. And we will seek to understand the limit of that scaling, limited by the degree that students are able to maintain motivation.
• The fundamental education journey will be maintained, which begins with rote learning, progresses to fundamental, labs, Socratic discussion, and finally leads to open-ended projects.

Summary

The underlying issue of education up to this point has been the cost to quality trade-off: the effectiveness of education is highest in situations with 1:1 or small group instruction and mentoring, while the cost is lowest with a lower-quality educational factory model. But we are now leaving this behind as education begins a new and exciting phase of transformation due to the arrival of AI and other technologies

The three main takeaways from our analysis are:

• AI changes the cost to quality relationship in education and a new and optimal ratio will likely arise in the future, offering both high quality and lower cost.
• The changes in education that come from AI and other new technologies will be gradual and incremental rather than sudden and dramatic. Universities will experience change but they will not be fundamentally disrupted because technology alone cannot replace the social expectations of education, specifically what and how individuals should learn.
• By understanding the correlations between current educational models with the emerging AT/tech capabilities, we can begin to more accurately envision a new future of education and understand how the tech-enabled and improved cost to quality trade-off will enhance access to quality education.

Using this framework, we expect to be able to draw a vision of that journey and we will be able to estimate the financial cost savings of that education compared with the so-called factory model.

Again, this is a working paper developed to open this important conversation. Send me your comments, suggestions, and contributions. I look forward to learning where this discussion will takes us.