In a recent conversation with Ben Kornell on the Edtech Insiders podcast, we explored a theme that sits at the intersection of cognitive science, education, and AI development: the profound role human expertise plays in teaching machines how to think.
Humans Train AI Through Structured, Expert Thinking
At its core, “human data” isn’t just labels or corrections—it’s the distilled reasoning of domain experts. We discussed how experts guide AI models not merely through right-or-wrong answers, but through structured thinking, showing models how to arrive at conclusions. This visibility into expert cognition is quickly becoming one of the most valuable forms of data in the AI ecosystem. (Episode marker: 00:02:48)
AI Training Now Mirrors How Humans Learn
We talked about the shift from simple binary labeling to techniques like reinforcement learning—approaches that parallel how people learn through feedback, iteration, and experience. As AI training evolves, it increasingly reflects the dynamics of human learning: try, evaluate, refine, repeat. (00:05:30)
Spiral Learning at Scale
Ben and I explored how both people and models improve through repeated exposure to concepts at increasing levels of complexity. This “spiral curriculum” isn’t just a pedagogy—it’s also how modern AI models become more capable over time. (00:07:47)
Making the Invisible Visible: Cognitive Modeling
One of our core discussions centered on revealing how experts think. By breaking down intuition, tacit knowledge, and mental models into explicit steps, we accelerate both human learning and AI performance. Cognitive modeling turns invisible expertise into teachable, scalable knowledge. (00:10:44)
Breaking Down Human Reasoning
We dove into methods for categorizing reasoning across domains—scientific, mathematical, narrative, ethical, and more. When AI systems understand these structures, they can generate or evaluate reasoning that more closely mirrors human thought. (00:13:28)
Why Interdisciplinary Thinking Drives Breakthroughs
Complex problems, whether in AI, education, or society, rarely belong to a single field. Ben and I talked about how interdisciplinary thinking gives both humans and AI the cognitive flexibility to solve problems that don’t fit neatly into one discipline. (00:16:28)
The Future of Human Data Work
As human knowledge expands, so will the demand for experts who can teach AI how to navigate new domains. The future isn’t about replacing human insight; it’s about scaling it. Human data work is becoming a dynamic, evolving profession aligned with the growth of human understanding itself. (00:18:58)
