Icelandic patients are now regularly using AI for medical self-diagnosis and demanding specific prescriptions. Psychiatrist Ólafur Þór Ævarsson revealed the trend, warning it's already impacting doctor-patient relationships nationwide.
Ólafur wrote about the issue on Akureyri.net and discussed it in Reykjavík. He said AI is here to stay. But he stressed the critical need to distinguish between AI-generated information and doctor's knowledge gained through repeated research and analysis. He knows from personal experience that AI can return false results.
"Maybe there are two things I want to talk about," Ólafur said. "One is what appears in the article—that we differentiate between technology and information, and then knowledge, which is what's behind it and experience."
He argues that when people have such good access to information, it changes how they communicate with physicians. "They just come and tell us beforehand what they think is best to do, which isn't always," Ólafur stated.
A Systemic Shift in Healthcare
The problem isn't isolated to Reykjavík's 101 district or the larger capital region. It's a national issue affecting clinics from Hafnarfjörður to Akureyri. Ólafur's warning highlights a fundamental change in how healthcare is accessed. Patients aren't just researching symptoms online anymore. They're arriving with AI-generated conclusions, expecting validation and specific treatment plans.
This creates immediate tension. A doctor's training, clinical experience, and ethical duty to "first, do no harm" can conflict with a patient's AI-powered certainty. The authority of the diagnosis is being challenged before the consultation even begins.
The Knowledge Versus Information Divide
Ólafur's core argument centers on a crucial distinction. Information is data—a list of symptoms matched to possible conditions. Knowledge is the synthesis of that data with years of medical training, patient history, physical examination, and professional intuition. It's what happens in a doctor's office, not a chatbot.
"We make a distinction between technology and information and then knowledge," he reiterated. The concern is that patients, armed with convincing AI outputs, may not see or value this difference. They're presented with a seemingly authoritative answer, lacking the context of its limitations or the probabilistic nature of medicine.
This isn't about Luddism. Ólafur acknowledges AI's permanent role. The challenge is integration. How does Iceland's state-backed healthcare system, already strained in remote regions, adapt? Should there be public guidelines on using AI for health? Can it be a tool for triage rather than diagnosis?
Political and Professional Reckoning
The Althingi's Health and Social Security Committee hasn't formally addressed AI's clinical impact. But Ólafur's comments force the issue onto the political agenda. MPs from the Independence Party, the Left-Greens, and the Progressive Party will likely face questions about regulatory frameworks.
There's a Nordic cooperation angle here too. Sweden and Norway are grappling with similar trends. Iceland's small, connected population makes it a potential test case for broader Scandinavian policy. Will Reykjavík lead on setting standards, or follow Oslo and Stockholm?
The Icelandic Medical Association hasn't issued formal guidance either. Individual practitioners are navigating this new dynamic alone. Some may see it as a time-saver if a patient is well-informed. Others view it as a profound disruption to their professional judgment.
Environmental and Economic Parallels
Consider Iceland's approach to its other major industries. Geothermal energy projects undergo years of environmental impact assessments. The fishing industry uses precise quotas based on marine biology research. Both rely on expert analysis of complex systems, not algorithmic shortcuts.
Healthcare is another complex national system. Rushing treatment based on unverified AI output could be compared to setting a fishing quota based on a single, unverified sonar reading. The potential for systemic error and resource misallocation is significant.
What about cost? Misdiagnosis leads to unnecessary tests, wrong medications, and delayed correct treatment. That strains Landspítali, the National University Hospital, and regional health centers. It wastes taxpayer krónur.
The Human Element in Reykjavík's Clinics
Back in a clinic in Grafarvogur or Breiðholt, the dynamic is personal. A doctor spends valuable consultation time deconstructing an AI report instead of conducting their own examination. Trust erodes. The patient might feel dismissed; the doctor might feel undermined.
Ólafur's example of a false AI result is key. It proves the technology isn't infallible. But convincing a patient who's emotionally invested in that result is another battle entirely. It requires time and delicate communication—resources already in short supply.
Some see a potential upside. Engaged patients can be better partners in their care. But engagement fueled by misleading or incomplete AI guidance is different. It's a distortion of the informed consent process.
Looking Ahead: A National Conversation
This isn't a problem with an easy policy fix. You can't legislate away ChatGPT. The solution starts with the conversation Ólafur has initiated. Public health campaigns might be needed, similar to those about antibiotic resistance, explaining the limits of AI in medicine.
Medical schools in Iceland might need to adapt curricula, teaching future doctors how to collaboratively assess AI-generated patient information. Continuing education for current practitioners is just as critical.
For now, the burden falls on individual doctors and nurses. They're the front line, correcting misinformation and rebuilding the trust that an algorithm can't provide. It's a silent, daily addition to their workload.
The AI doesn't feel the weight of a wrong diagnosis. The doctor does. And so does the patient. That human reality, Ólafur suggests, is what's getting lost in the code.