Human-Centered AI: The Importance of Learning from Experience

May 14, 2024 by Tim Sarchet

The potential applications of AI and Large Language Models (LLMs) for enterprise learning, and indeed learning of all types and for all ages, is enormous. As industry analyst and friend of Nomadic ¹, Josh Bersin, outlines in a recent blog post, the use cases and implications are far reaching.

Josh envisions a world where AI is used to create new learning content on the fly, as demand emerges for new topics and at the point of need for the employee. He predicts a world where employees can simply ask an AI powered teaching assistant how to master an area of their job, and an LLM, well trained in the specific tools and operations of a business, will be able to respond with effective guidance. Not only that, it will be able to adapt to how the employee is doing and serve-up new content matched to their real-time knowledge gaps.

While the potential of AI is obvious, we also think AI has a particular weakness which is often overlooked, and which should guide how and when AI is used in enterprise learning.

Learning from Lived Experience

AI, and LLMs in particular, have no lived experience. It is ingesting the experiences and knowledge of others (the population of the entire internet!) at massive scale and it seems to be able to create something novel from that ingested experience. But it has never actually experienced or done anything first hand. It has never completed a sales conversation with a key client, it has never had to comfort a team member who was going through a rough time, it has never had to lead a team meeting after layoffs.

But learning from lived experience, either directly from your own trial and error or from those with first-hand experience, is what matters most in many subjects. The stuff that’s hard to articulate or learn from a book but can only be learned through years of experience. Trying things, failing, trying again, and learning from others who have been down the same road. This is often where the best and hardest learning happens.

In some subjects lived experience doesn’t matter that much, take compliance or how to use excel. In those subjects AI will work very well. But as a thought exercise, consider building a boat, and the boat you built had to carry you across an ocean and withstand all that nature can throw at it. Would you rather learn how to build your boat from someone who had read every book on boat building but had never actually built a boat themselves, or would you rather learn from a master shipwright, who had built dozens of boats with their own hands and sailed the boats they built across oceans?

Think about things closer to home, like setting the strategy for your team and leading them through a turbulent time where jobs might be lost. Or figuring out how to develop a creative marketing strategy that is good enough for your organization to stand out in a very crowded market. The experiences of others who have been down the same road is invaluable.

You could argue that AI has ingested 1000s of similar experiences to the one you are facing and can combine those experiences into lessons that are more effective than the experience of a handful of colleagues. But there’s a profound difference between learning from the amalgamated experiences of 1000s of anonymous people, and the firsthand stories and lessons of a few people you trust.

AI & Cohort-Based Social Learning

All this isn’t to say that AI can’t and won’t be massively helpful (and disruptive) even in the kind of subjects where lived experience is essential, It can and will be. But it will not be very effective on its own. It has to be combined with learning from your own direct experiences, in contexts where it’s safe to practice and fail, and with learning from the direct experience of colleagues, mentors and experts.

That’s why we think cohort-based social learning combined with the capabilities of GenAI - intelligent search & recommendations, real-time content creation, adaptive learning - is going to be the most powerful combination. Cohort-based social learning gives learners direct access to the knowledge and experience of their peers, in similar roles around the world. Through carefully designed conversations, learners can discover how their colleagues tackled a shared challenge, or overcame a common obstacle. And the safe environment of a cohort-based learning platform gives the learner the space to test their own ideas, and to share failures and successes as they implement what they are learning.

Alongside this, AI can find the best piece of content to supplement a case they are discussing with their peers. It can continually assess how the learner is doing and point out areas for deeper learning and improvement. And it can answer questions from the learner, which the learner can test and verify with their cohort of peers.

We’re still figuring out exactly how AI can complement cohort-based social learning but as with many other industries and business functions we are finding the best role for AI is to augment what we do, not necessarily to replace it. In many ways this is the defining challenge of AI, how do we combine human activity with artificial intelligence. We’ll keep experimenting, and maybe we’ll ask AI what it thinks too.

Interested in how AI is reshaping learning and the skills people need as we enter the age of AI? Check out a few other relevant articles:

• A human centered approach to AI
• The Rapidly Evolving Role of L&D in an AI-Powered World
• 5 Essential Capabilities to Thrive in an AI World

Want to learn more about Nomadic’s work helping global organizations adapt to the Age of AI? Get in touch here and we’ll set up a conversation!