Let's say someone asks you to suggest five global markets for a product launch. You enter it into ChatGPT, receive a response within seconds, and present it during the meeting. Six months later, when it's either a massive success or spectacular failure, someone asks, “Why did we choose these markets?” And you realize... you have no idea.
This is what our guest John Willshire calls “cognitive debt”—the cost of forgoing thinking to get answers quickly, creating a debt of understanding that, like technical debt in software, is meant to be repaid but often isn't.
John Willshire runs Smithery, a strategic design practice in the UK, and has been thinking critically about AI since 2017. In this conversation, we explore his framework for understanding how we accumulate cognitive debt, why organizations are mandating its creation at scale, and what happens when we lose the connections between our questions and our answers.
We delve into the distinctions between AI as a narrow, specific tool and the "world-eating data generalists" marketed as universal solutions. We examine why sycophantic chat interfaces make us trust statistical relationships we can't verify, especially in areas where we're weakest. And we consider what it means for agency and understanding when our thinking processes become increasingly opaque to ourselves.
This isn't just about AI; it's about the risk of prioritizing answers over understanding and its effect on human capacity.
Key Topics:
Chapters:
00:01 - Introduction & Welcome John Willshire
03:34 - What is Cognitive Debt? The Core Concept
09:50 - AI Types & Why Text Generation is Different
16:58 - AI Sycophancy: When Systems Lie to Please
24:19 - Why Organizations are Mandating Cognitive Debt
34:15 - Cultural Imagination vs. Technological Reality
40:30 - Dependency & Agency: The Real Cost
45:01 - Designing Better AI: The Path Forward
Links: