I’m Emil, and I recently had the pleasure of hosting a truly thought-provoking conversation on the podcast with Dimitris Tollis, a leading mind at the intersection of learning design, workplace learning, and artificial intelligence. Dimitris brings a unique perspective, grounded in philosophy and pedagogy, to the often-hyped world of AI in education. If you enjoy this deep dive into the future of learning, please consider exploring more of our content and joining our community of forward-thinkers.
Have you ever felt that AI-generated content, whilst fast, sometimes misses the mark for true, deep learning?
What’s the biggest misconception about AI in learning that you’ve encountered?
Dimitris argues that it’s the “illusion of wisdom.” We ask AI a question, get a good answer, and then have the illusion that the answer is ours, that we’ve become wise. But we haven’t. This illusion is pervasive, particularly among learners who copy-paste AI-generated responses for assignments, receive good grades, and believe they’ve improved. They haven’t. This over-reliance leads to what Dimitris calls a “critical thinking debt.”
Here are the key takeaways from our conversation:
The Illusion of Wisdom: AI can provide answers, but it doesn’t automatically confer understanding or wisdom. Over-reliance can lead to a “critical thinking debt.” Productivity vs. Pedagogy: The current AI boom often prioritises rapid content generation and personal productivity over genuine learner engagement and quality educational outcomes. Learning Should Be Challenging: True learning requires effort, reflection, and problem-solving, not just passive consumption of easily generated content. AI for Personalisation, Not Just Production: AI’s true potential lies in creating adaptive quizzes, knowledge mentors, coaching assistants, and role-plays that challenge learners and build competencies. The Evolving Role of the LMS: Learning Management Systems should transform into “smart coaches,” integrating with intelligent content to deliver deeply personalised learning journeys. Human in the Loop: Instructional designers must adapt to “design for the agent,” setting guardrails and context for AI, ensuring human oversight remains integral to the learning process. Future Focus: Beyond core Large Language Models (LLMs), innovation will centre on agent ecosystems, automation, realistic avatars, and the convergence of AI with VR/AR, all driven by personalisation. Learner-Centred Design: The ultimate goal for EdTech should be to prioritise the learner’s needs and genuine educational outcomes, co-designing solutions with educators and learners.
This illusion of effortless learning is concerning. What impact does this over-reliance on AI for personal productivity have on the quality of learning experiences and the learner themselves?
How has the drive for increased productivity in L&D, fuelled by AI, inadvertently led to a decline in genuine learning outcomes?
Dimitris observed that the focus has shifted dramatically from the learner to content production. Instructional designers and trainers, whilst seeing an enormous boost in personal productivity (creating quizzes in hours instead of days), often become overly dependent on AI. This leads to a “cognitive offload” where their own abilities to intervene or correct diminish. The quality of deliverables starts to drop, and crucially, the learner is forgotten. We’re producing much more content, but doing less for the learner. As Aristotle wisely stated, learning should not be easy; it should be challenging. How can you truly learn if you don’t push yourself to think, reason, and reflect?
So, if simply churning out more content isn’t the solution, how can we leverage AI to create truly impactful, challenging, and personalised learning experiences?
What specific AI-driven innovations can move us beyond mass content creation towards deeper engagement and competency building?
Dimitris highlighted several innovative applications, many of which he’s been piloting with European Union agencies. One such concept is “InSORM AI.” Imagine a learning package where, alongside theory and exercises, you have a knowledge mentor – an AI agent deeply familiar with the course content, ready to answer any question you might have. Or consider the adaptive and personalised quiz. Instead of a one-size-fits-all test, this AI-driven quiz is fed by a competency framework. It doesn’t stop until you achieve every single competency, adjusting the difficulty of questions based on your performance. If you’re excelling, it offers harder questions; if you’re struggling, easier ones. This is a fundamental shift from simply getting a numerical score to truly building and assessing mastery. Beyond quizzes, there are coaching assistants that provide qualitative, detailed feedback on open-ended scenarios, and AI role-plays where you can practise skills like interviewing with a real-time AI avatar, getting immediate, actionable feedback. These innovations add the necessary “friction” to the learning process, making it more challenging and memorable.
This vision of personalised, adaptive learning requires a sophisticated interplay between content and platform. How do traditional learning management systems evolve to support this new paradigm?
Should AI capabilities be embedded solely within the learning content, or should the LMS itself become a “smart coach”?
Dimitris believes it should be both. The LMS needs to evolve from being a mere content delivery system to a “smart coach,” a “smart trainer,” or a “smart mentor.” It should be able to monitor a learner’s competencies, understand their deficiencies, and tailor the learning journey accordingly. This means a collaborative relationship where the LMS, drawing from a comprehensive “organisational brain” of data (including user profiles, past performance, and company knowledge bases), interacts with smart, adaptive course content. This integration allows for automated, deeply personalised learning production, moving away from generic courses and towards training that is truly on-demand and tailored to individual needs and organisational goals.
With AI taking on so many roles, what aspects of learning design must remain fundamentally human to ensure quality and authenticity?
How does the role of the instructional designer transform when parts of the learning experience are autonomously managed by AI?
Dimitris firmly believes in “human in the loop,” meaning humans must be an integral part of the process, not merely optional. The role of the instructional designer is set to change dramatically. Traditionally, instructional designers build and test everything before it reaches the learner. Now, parts of the course – the knowledge mentor, the Socratic agent, the adaptive quiz – are out of their direct control. The best they can do is provide the correct prompting, context, and guardrails for the AI. It’s akin to “training the agent” – ensuring that when the designer isn’t there, the AI assistant performs as it should. This requires a new, almost “3D” skill set for instructional designers, transforming their role to design for the agent, not just the learner. Without this evolution, we risk many unwelcome surprises and a loss of trust from learners and organisations alike.
Looking forward, which AI innovations hold the most promise for EdTech in the coming years, beyond the current capabilities of large language models?
Where do you see the most significant breakthroughs happening in the next three to five years to truly enhance learning?
Whilst Large Language Models (LLMs) are already incredibly powerful, Dimitris believes the exponential improvement will now shift to the “ecosystems” and “wrappers” built around them. This includes advancements in agent technology – how we prompt, control, and orchestrate multiple AI agents to create intelligent behaviours. He anticipates significant progress in automation, offloading heavy workloads to agents so that humans can focus on more valuable interactions with learners. He’s also looking forward to better avatars that overcome the “uncanny valley” effect, offering more natural, real-time communication in any language. Ultimately, he foresees a powerful convergence of media, tools, and techniques, bringing AI closer to the worlds of Virtual Reality (VR) and Augmented Reality (AR), all underpinned by the relentless drive towards personalisation.
So, to finally answer that initial question about why AI-generated content sometimes misses the mark for true, deep learning… it comes down to one fundamental principle: learner-centred design. We need to prioritise the learner’s needs and genuine educational outcomes over mere technological capability or productivity gains. Technology companies, Dimitris argues, should invest more in hiring educators and co-designing solutions with them and with learners themselves. It’s about doing the right thing with the right technology, not just doing the wrong thing faster.
—
The future of learning is here, and it’s about more than just generating content. It’s about crafting deeply personalised, challenging, and effective learning experiences that truly build capability and critical thinking.
Ready to explore how Open eLMS can help your organisation move beyond basic content creation to truly transform your learning and development? Discover the power of the Open eLMS Learning Generator to create immersive, adaptive learning experiences, or delve into the capabilities of our Open eLMS LMS to become a smart coach and mentor for your team.
Contact us today to learn more about how our AI-powered solutions can help you create learning that truly resonates and delivers measurable outcomes.