Magnús Smári Smárason

"Treat the patient, not the monitor" is a core principle from emergency medicine. When vital signs on a screen contradict what you see in front of you—a child struggling to breathe, a patient in obvious distress—you trust the patient. Numbers guide diagnosis, but they serve care rather than replace judgment.

I apply this principle to AI governance. Right now, we are treating the monitor: optimizing for efficiency metrics, productivity rates, and capital returns while ignoring the human cost. The "patient" is the worker, the student, the citizen affected by these systems. The "monitor" is the dashboard that cannot register pain, exhaustion, or eroded capability.

Sixteen years as a paramedic and firefighter taught me what happens when systems fail people. I learned pattern recognition under pressure and diagnostic thinking when stakes are real. In November 2022, the week I moved to office work, ChatGPT launched. I started building AI systems immediately.

My research focuses on two problems. Responsibility Fog: how accountability dissolves when decisions get distributed across algorithms and committees. Cognitive Debt: how human judgment atrophies when we outsource thinking to machines. Both are symptoms of treating the monitor.

I serve as AI Project Manager at University of Akureyri. I led Iceland's first comprehensive university AI policy and oversee systems including BORG (institutional governance) and Arctic Tracker (conservation). I teach AI implementation courses, host Temjum tæknina podcast, served on the Technology Development Fund expert panel (spring and fall 2025), and am co-authoring The Irreducible Human (Brill).

The question driving my work: How do we deploy AI that builds human capacity instead of replacing it?