In this episode, we sit down with Keir Henderson, Financial Services Account Executive at Precisely. Keir shares deep insights into AI governance, policy challenges, international cooperation, and the philosophical tensions driving today’s AI landscape. From existential risk to frontier models, we explore the crossroads of technology, politics, and ethics in shaping our collective AI future. If you’re interested in how governments are preparing for the age of powerful AI systems—or how we should be—this conversation is for you.
Episode Summary
.
From Evidence to Action: What Makes Research Actually Useful?
In this episode, Keir Henderson offers a refreshing and deeply thoughtful perspective on what it really means for research to be useful—especially in the realm of public policy. He doesn’t stop at whether an intervention works in one isolated study. Instead, he emphasises the importance of portability—the idea that useful research must travel well across contexts, populations, and settings.
“We’re not just looking for a signal,” Keir says, “we’re looking for a signal that survives contact with the rest of the world.”
This insight cuts to the heart of what separates theoretical value from practical application. In a world eager for evidence-based decisions, Keir encourages listeners to go beyond surface-level metrics and ask whether an intervention’s underlying mechanism will hold up under different circumstances. It’s a challenge to researchers, policymakers, and practitioners alike to rethink how we assess success and build systems that are not just evidence-informed, but robust and adaptive in the face of complexity.
The IKEA Effect of Ideas: Why We Overvalue Our Own Models
One of the most memorable parts of the conversation is Keir’s exploration of the subtle biases that shape how researchers and thinkers evaluate their own work. In a moment of delightful self-awareness, he jokes, “There’s a kind of IKEA effect in social science, where if you built the model yourself, you trust it more than you should.” This metaphor—playful yet cutting—captures how personal investment in a theory or framework can cloud our judgment.
He dives into the social and psychological forces that often make scientific discourse feel less like a search for truth and more like a battleground of identities, aesthetics, and status games. Rather than condemning these tendencies, Keir invites us to recognise them—and to develop a kind of epistemic humility that makes room for uncertainty, disagreement, and the messiness of real discovery. It’s a call for a more grounded, less self-congratulatory science—one where we hold our models loosely and stay open to being wrong.
Mechanisms Over Metrics: The Deep Work of Causal Inference
Throughout the episode, Keir circles back to one of his core themes: that understanding why something works is just as important—if not more—than knowing that it works. In the current culture of evidence-based policy, randomised controlled trials (RCTs) are often treated as the gold standard. And while Keir acknowledges their power, he also highlights their limitations. “RCTs are great for telling us whether something works in a specific context,” he says, “but they don’t necessarily tell us what about it made it work.”
This distinction matters deeply when attempting to scale or replicate an intervention. If we can’t identify the active ingredient, we risk copying surface features and missing what actually drives impact. Keir’s perspective pushes for deeper inquiry, focusing not just on measurable outcomes, but on the underlying causal structures that generate them. It’s a richer, more demanding vision of scientific work—one that values clarity, depth, and a relentless curiosity about how the world really works.
.
To find out more about Keir, check out our full episode – available on all your favourite channels. Now including YouTube!
.
This article summarises podcast episode 129 “track, assess, & classify” recorded by CX Insider.
Written by Elysia Filaitis