The Problem With AI-Driven Personalization in Higher Education

Criticism of AI-enhanced personalized education is on the rise. While technological advances in the form of GPTs, AI tutors, and even entire educational ecosystems like Khan Academy may at face value seem appealing, a high degree of customization in the learning experience is not without consequence.

Sal Khan tells us that the most successful type of learning has always been one-to-one tutoring, referring to Benjamin Bloom’s 1984 “Two Sigma” study which highlighted the benefits of one-to-one tutoring. This is true, but there’s a but - and it’s a big one: the kind of high impact tutoring we are talking about has always been human to human, from Socrates to present day tutoring in the best universities around the world.

The Limitations of AI-Driven One-to-One Tutoring

I see at least four reasons why we may want to proceed with caution when we talk about “personalised” tutoring in higher education:

First, AI-enhanced personalized education fundamentally builds on old principles of learning. While the technology may be cutting-edge, the underlying pedagogical assumptions often go back to outdated models of education. The idea that learning can somehow neatly be packaged into discrete, personalized modules ignores the complex, interconnected nature of knowledge acquisition. We risk reinforcing a reductionist view of education that prioritizes measurable outcomes over holistic understanding.

Second, it potentially creates pockets of learning at different levels, which will make transparency an issue for employers and other stakeholders. As students progress through highly individualized learning paths, it becomes increasingly difficult to compare and evaluate their skills and knowledge bases. How can we ensure that a degree or certification holds consistent meaning when each learner's journey is unique? This lack of standardization could lead to increased skepticism from employers and a devaluation of traditional educational credentials.

Third, education is and always has been a social activity as much as an academic experience. The rush towards personalized, AI-driven learning threatens to isolate students, depriving them of the rich interpersonal experiences that have long been a cornerstone of education. This is not to suggest the current education system isn’t flawed; it obviously is. But the development of soft skills, the exchange of diverse perspectives, and the formation of professional networks all rely on human-to-human interactions that AI cannot replicate.

Fourth, who decided on the underlying premise that tailor-made learning is necessarily better than group learning? Clearly, there are advantages and disadvantages to both. While personalization can address individual learning needs, it may also deprive students of the valuable experiences that come from collaborative problem-solving, debate, and shared discovery. Thinking on your feet with a chatbot is very different from doing the same in a group of people whose body language communicate endlessly more than text, speech or multimodal digital outputs. Group learning fosters resilience, adaptability, and the ability to work with diverse teams – skills that are increasingly crucial in our interconnected world.

The Promised Benefits of AI Personalization

It is clear by now that AI's ability to analyze vast amounts of data about individual students - their learning styles, pace, strengths, and weaknesses - offers unprecedented potential for customization. Adaptive learning platforms can adjust content difficulty, suggest relevant resources, and provide timely interventions. This level of personalization could theoretically lead to more engaged students, improved learning outcomes, and a more efficient educational process.

Theoretically being the operative word here.

As the hunt for AI-driven personalization continues, we risk creating an educational landscape where the very algorithms designed to cater to individual needs inadvertently narrow students' exposure to diverse perspectives and challenging ideas. The echo chamber effect, so prevalent in social media, could find its way into our classrooms, virtual or otherwise.

If an AI system determines that a student learns best through visual content and practical examples, it might prioritize such materials, potentially limiting exposure to abstract concepts or text-based learning. While this approach may lead to better short-term performance, it could hinder the development of well-rounded critical thinking skills crucial for navigating our complex world.

Moreover, as AI becomes more adept at generating educational content, we face the risk of a subtle homogenization of knowledge. When countless institutions rely on similar AI models to create course materials, we may end up with a convergence of content that lacks the richness and diversity of human-curated curricula. The nuances of different academic traditions, the idiosyncrasies of individual professors, and the serendipitous discoveries that often occur in less structured learning environments could be lost in the hunt for data driven, measurable educational outcomes.

Sticking to the Purpose of Higher Education

This paradox poses fundamental questions about the nature of education itself. Is the goal of higher education solely to impart knowledge and skills in the most efficient manner possible? Or is there inherent value in the struggle, in exposure to diverse and sometimes uncomfortable ideas, in the shared experience of grappling with complex concepts alongside peers - and in networking with real human beings as part of something larger than oneself?

The ultimate goal of education is not just to transmit and convey information, but to cultivate wisdom, encourage innovation, and prepare students to shape the world they will inherit. In navigating generative AI as educators, our pursuit of personalization must not come at the cost of the diverse, challenging, and often messy experiences that make education a profoundly human endeavour.