AI in Spine Surgery: Preserving Clinical Judgment in a Data-Driven Era

Summary

---

Podcast Summary

Becker’s Healthcare Podcast
Episode: AI in Spine Surgery: Preserving Clinical Judgment in a Data-Driven Era
Host: Carly Beam
Guest: Dr. Morgan Loriaux, Spine Surgeon & Past President, International Society for the Advancement of Spine Surgery
Date: March 28, 2026

---

Overview

This episode delves into the rapidly evolving relationship between artificial intelligence (AI) and spine surgery. Dr. Morgan Loriaux draws on decades of clinical, leadership, and policy experience to discuss how AI tools reshape data analysis and clinical workflows—and, crucially, why the preservation of human judgment remains at the core of effective, ethical patient care. The conversation covers opportunities, risks, and frameworks for integrating AI into complex medical environments, with a particular focus on the subtle, essential cognitive work that precedes surgical intervention.

---

Key Discussion Points & Insights

The Scale and Speed of AI in Healthcare

• AI’s explosive integration versus the gradual rise of the internet:
  • “Artificial intelligence is arriving very differently. Its growth isn't gradual, it's almost vertical. And now that capability sits inside the devices we carry in our pockets.” — Dr. Loriaux [01:33]
• Real risk is not simply errors, but skill erosion:
  • High-value work happens before surgery, when physicians synthesize imaging, history, comorbidities, and risks.
  • These cognitive steps are vital and not easily measured or reimbursed.

Fears of AI Replacing Physicians

• Replacement is not the main threat:
  • “AI will not enter medicine neutrally… Accuracy can improve while human judgment quietly declines.” — Dr. Loriaux [03:27]
• True risk lies in governance and subtle erosion of professional skills.
  • AI’s integration into existing systems (algorithms, prior authorizations) may quietly degrade unique clinical insight.

The Imperative of Physician Oversight

• Governance and ‘surgeon in the loop’:
  • “Maintaining physician oversight therefore becomes critical as AI becomes embedded in those systems.” — Dr. Loriaux [04:26]
• Complex patients need flexible, reasoned care:
  • Human judgment is often required to override standard algorithms when patients defy typical patterns.

Memorable Case Example: Asking the Right Questions

• AI can deliver accurate but incomplete answers:
  • Dr. Loriaux recounts analyzing cancer trends with AI regarding post-COVID incidence versus mortality:
    • “The AI system hadn't really made a mistake per se… Physicians still have to decide which questions matter.” [05:44]
• Insight: It’s sometimes not the answer but the question that reveals the overlooked truth.

Efficiency vs. Blind Spots: The ‘GPS’ Analogy

• AI optimizes based on predefined filters; the “route” is only as good as the user’s intent:
  • “AI can optimize the route, but someone still has to decide where we're going as we move forward.” — Dr. Loriaux [07:05]
• Overemphasis on efficiency risks missing individual patient realities.

The Physician’s Role: Stewardship & Transparency

• Physicians must be ‘stewards’—not mere users—of AI-driven systems:
  • Complexity and unpredictability in outcomes map to the most vulnerable patients.
  • There’s a risk that over-prioritizing predictability may neglect those who need nuanced care the most.
• Dubai Future Foundation roundtable and the “collaboration framework”:
  • Levels of human and AI collaboration must be transparent—e.g., human-led AI, machine-dominant tasks, etc.
  • “A clear labeling system helps make that relationship visible.” — Dr. Loriaux [09:20]
• Clinical responsibility and judgment must remain with humans.

Influence on Medical Knowledge Production

• AI shapes not only what we write, but how and even what we choose to write about:
  • Journals are codifying guidelines (e.g., at the International Journal of Spine Surgery) requiring scientific and clinical interpretation to remain human-led.
  • “…Capability will precede consent and adoption will precede reflection. Most importantly, those filters can gradually direct the flow of conversation, emphasizing some questions while quietly discouraging others.” — Dr. Loriaux [11:23]

Application to Spine Care

• Chronic low back pain is not a single disease but multiple phenotypes:
  • AI can help recognize subtypes, but if systems oversimplify pathways for efficiency, crucial distinctions may be lost.
  • “It’s whether physicians remain actively engaged in shaping how knowledge itself is generated and applied.” — Dr. Loriaux [13:23]

Machine Memory vs. Human Meaning

• AI preserves all information, but human memory is selective:
  • “That act of selection...is actually a fundamental part of how knowledge develops in medicine.” — Dr. Loriaux [13:58]
• Physicians must decide what matters enough to carry forward, preserving resilience and meaning in a sea of machine memory.
• Bureaucracy—amplified by AI—decides what becomes ‘allowed to matter’.

The Essential Takeaway

• Preserve the habits of thinking that make physicians indispensable:
  • “Artificial intelligence is going to become extraordinarily good at answering questions. The future of medicine will depend on whether physicians continue to ask the right ones... The real risk is that it will slowly replace the habits of thinking that make physicians necessary.” — Dr. Loriaux [15:19]

---

Notable Quotes & Moments

• On the evolution of clinical roles in AI environments:
  “AI answers questions very well...Physicians still have to decide which questions matter.” — Dr. Loriaux [05:58]

• On transparency in human-AI collaboration:
  “A clear labeling system helps make that relationship visible.” — Dr. Loriaux [09:20]

• On the risk to clinical identity:
  “The real risk is that it will slowly replace the habits of thinking that make physicians necessary. And in that vacuum, no one will be left to recognize or remember the meaning.” — Dr. Loriaux [15:40]

---

Timestamps for Important Segments

• [01:13] – AI’s arrival: scale, speed, and real risk to clinical roles
• [03:12] – Job security, replacement concerns, and the subtle threat to judgment
• [04:01] – Governance: The need for surgeon-in-the-loop oversight
• [04:58] – Case study: Role of questioning in interpreting AI-generated data
• [06:44] – Efficiency, optimization, and the risk of clinical blind spots
• [07:43] – Stewardship, transparency, and the Dubai AI governance framework
• [10:03] – How AI shapes scientific writing and medical knowledge
• [12:41] – Specific applications and challenges in spine care
• [13:46] – Human meaning vs. machine memory: selection as a source of resilience
• [15:15] – Dr. Loriaux’s parting lesson for physicians navigating the AI era

---

Conclusion

This episode frames AI not as a threat or panacea, but as a force that must be thoughtfully governed. Dr. Loriaux calls for active, ongoing stewardship from physicians to maintain clinical meaning and ensure the highest-value work—critical thinking and nuanced judgment—remains human-driven. In spine surgery and beyond, the future belongs to clinicians who not only use AI, but also preserve the habits of mind that add irreplaceable value to technological tools.

Transcript

[A]

This is Carly Beam with the Becker Spine and Orthopedics podcast. And joining us today is Dr. Morgan Loriaux, a spine surgeon and past president of the International Society for the Advancement of Spine Surgery. And today we're going to have a great chat about AI and spine surgery and where all sorts kind of fits in the larger healthcare equation. So, Dr. Lorio, thank you so much for joining us.

[B]

Thank you, Carly. It's a pleasure to be here.

[A]

Always great to have you back on the podcast. And before we dive into our questions, could you just give a quick overview of your background and your work?

[B]

Okay. Well, I've been a spine surgeon and a hand surgeon and past president of isas, and I did a lot of work for policy, and I chaired coding and reimbursement task force for ISAs as well. So that's pretty much been my interest. And AI kind of fits in here because of its impact on policies is it impacts both healthcare delivery and, frankly, what we write and what we do moving forward. So hopefully this is relevant.

[A]

Absolutely. And I mean, obviously, AI is everywhere in healthcare and healthcare conversations. I'd love to hear start off how you think about its role when it comes to spine surgery.

[B]

Well, I use AI and I value it. It's an extraordinary analytical tool. If you step back for a moment, the scale of this technology becomes clearer. We remember the Internet took decades to reshape society. And today we recognize the lack of Internet access represents a form of inequality. And so artificial intelligence is arriving very differently. Is growth isn't gradual, it's almost vertical. And now that capability sits inside the devices we carry in our pockets. But I think the real risk is often misunderstood by that. I mean, the real risk isn't that AI will get things wrong, that it will succeed while we slowly forget how in spine care, the highest value work often happens before any incision is made. That's when a surgeon integrates imaging, biomechanics, patient history, comorbidities, expectations, and risk tolerance into a coherent clinical judgment. Those cognitive steps are difficult to measure and are rarely reimbursed in proportion to their importance, but they determine whether the right patient receives the right intervention at the right time.

[A]

It's a really fascinating way of looking at it, too. Just the idea that AI is coming in before the actual surgery. But obviously, some people are worried that AI might replace their jobs and including physicians. Do you see that being a concern?

[B]

Well, I don't think that replacement's the real issue. Specifically for surgeons, health care already operates through algorithms, reimbursement rules, prior Authorization systems, coverage policies, and outcome reporting structures. But I can say this AI will not enter medicine neutrally. Artificial intelligence will likely arrive through those same systems I mentioned. Unfortunately, skill erosion rarely shows up in the metrics we measure. Accuracy can improve while human judgment quietly declines. So I guess the real question becomes governance.

[A]

Yeah. Can you dive into that more governance specifically?

[B]

Yeah. Complex patients don't always fit neatly inside efficient systems. Historically, the surgeon in the loop judgment has allowed physicians to recognize when a patient doesn't fit the algorithm and to treat the patient rather than the pathway. Medicine requires flexibility because real patients are real. Complicated systems, however, often prefer predictability. Maintaining physician oversight therefore becomes critical as AI becomes embedded in those systems.

[A]

It's an interesting balance, it sounds like, between the algorithms and then figuring out those one off cases. Can you give an example of where physician judgment still matters even when AI is helping out as a tool?

[B]

Sure. Well, I recently explored some cancer trend data within an AI system. And when I asked whether cancer incidents had changed after Covid, the analysis suggested that incident rates were largely unchanged. But when I examined the time series more closely, something about the pattern didn't quite fit in the data points. The COVID years were clearly disruptive, so we treated that period as difficult to interpret and step back to examine broader trends. When we then asked a different question, focusing on cancer mortality instead of incidence, the pattern appeared different. Mortality seemed meaningfully higher than earlier baseline patterns, roughly on the order of about 16%. That is a substantial change. The AI system hadn't really made a mistake per se. It answered the question it was asked, but it reminded me of something important. AI answers questions that very well. Sometimes it narrows answers simultaneously. Physicians still have to decide which questions matter. Sometimes the insight comes from recognizing that the question itself needs to change.

[A]

Really fascinating idea there. And we'll circle back to that in a moment. I also want to pick your brain on health systems and how many of them. Often they hope that AI will make care more efficient. What are your thoughts on that?

[B]

Well, efficiency is important, but efficiency can also introduce blind spots. Sometimes I think about this using a GPS analogy. You know, navigation systems optimize routes based on filters, minimizing time, avoiding tolls, reducing distance. But occasionally those filters lead you into loops or take you off the main road, only to bring you back again later. The system is optimized or optimizing based on the parameters it was given. AI functions in a similar way. AI can optimize the route, but someone still has to decide where we're going as we move forward. These Filters are going to become more and more important.

[A]

Dr. Lorio, can you talk about the role physicians should play as AI is becoming more integrated into the work they do?

[B]

Yeah. So physicians need to remain actively involved in the governance of these systems, not just as users, but as stewards. Complex patients often generate more uncertainty and variability in outcomes, but those are frequently the patients who need the most thoughtful care. If systems begin to prioritize predictability above all else, there's a risk that the most complex patients may become harder to treat. Just over a year ago, I had the opportunity to participate in a global 50 roundtable organized by the Dubai Future foundation, where experts from multiple sectors were asked to consider possible futures roughly five decades from now. My perspective in that discussion came from two places. My international policy work in spine surgery through ICES and my experience as a practicing surgeon. One idea emerging from that initiative is a classification framework describing how humans and artificial intelligence collaborate in the production of knowledge and decisions. The framework uses simple visual markers to indicate the degree of human involvement in AI assisted work, ranging from fully human activity to machine dominant processes, with several intermediate stages representing different levels of collaboration and oversight. The purpose is transparency. As artificial intelligence becomes more deeply embedded in research, analysis and communication, it will become increasingly difficult to determine how much of an output reflects human judgment versus machine computation. A clear labeling system helps make that relationship visible. From a clinical perspective, this aligns closely with something we have long emphasized in medicine surgeon in the loop governance. Artificial intelligence may assist with analysis, pattern recognition and data synthesis, but clinical responsibility and decision authority remain human. In the terminology of the Dubai framework, this approach would be classified as human led AI.

[A]

The fascinating way to describe it, human led AI. And can you talk about how AI might influence the way medical knowledge itself is developing?

[B]

Yeah. There's another dimension to this conversation that we don't always acknowledge. AI increasingly influences how we write, whether drafting clinical notes, summarizing literature, or producing scientific manuscripts. When technology influences how we write, it inevitably influences what we write. Knowing who or what produced it is therefore important. This can be categorized across six human only, human directed, human led, machine assisted, machine dominant, and machine only. In fact, journals are already starting to formalize this principle. At the International Journal of Spine Surgery, new author guidelines make it clear that AI can assist with drafting or organization, but scientific judgment and clinical interpretation must remain human led. And over that time, that shapes what we collectively know, what we believe, and ultimately how we practice medicine. AI systems apply filters, optimization rules, and administrative constraints. Capability will precede consent and adoption will precede reflection. Most importantly, Those filters can gradually direct the flow of conversation, emphasizing some questions while quietly discouraging others. I can't emphasize how important and dangerous these filters are and may become. When systems rewrite rather than carefully amend existing work, portions of the original meaning can quietly disappear. The danger isn't censorship or deception. The danger is continuous optimization, where erosion of skills slowly becomes frustration, framed as efficiency. And that matters in medicine because the way knowledge is framed ultimately shapes how diseases are defined, classified, and treated.

[A]

Definitely. And I like what you said about what you're doing at ISAs and in the journal and how you're kind of making guidelines based off of what we know. Now, with AI, can you dive deeper into how all this relates specifically to spine care?

[B]

Sure. Well, in spine care, this matters because chronic low back pain, as I mentioned, is not a single disease. For instance, it includes multiple phenotypes. Discogenic pain, vertebrogenic pain, sacroiliac dysfunction, multifidus, muscle failure, neuropathic and radicular pain, and central sensitization patterns. AI can help us recognize these phenotypes earlier and more consistently. But if surrounding systems favor simplified treatment pathways, or if they narrow outcome metrics, those distinctions can gradually disappear from the conversation. So the issue isn't simply whether AI helps analyze data. It's whether physicians remain actively engaged in shaping how knowledge itself is generated and applied.

[A]

Got it. And, Dr. Lorio, I wanted to also touch on this idea of machine memory versus human meaning. Can you dive into that and just, you know, the kind of. The tensions that exist in that realm?

[B]

Okay, well, speaking to that tension as well, Artificial intelligence systems accumulate information indefinitely. They do not forget. Humans are very different. Our memory is finite. We cannot retain everything. And so we must choose what is worth remembering. What matters is remembering the sequence, what humans did before systems, when judgment began to be deferred and when dependence quietly replaced skill. Maintaining skill will be crucial if humans are to provide resilience for this new AI ecosystem. That act of selection, deciding which observations matter enough to carry it forward, is. Is actually a fundamental part of how knowledge develops in medicine. AI may store and synthesize vast amounts of information, but physicians still determine which patterns are meaningful and which insights deserve to shape future care. AI remembers everything but bureaucracies, Bureaucracies. They decide what is allowed to matter. Artificial intelligence didn't invent these. These tendencies. It simply accelerated them.

[A]

That's very well said, Dr. Lorio. And before we wrap up, if you had to summarize a key lesson for physicians about AI, what would it be?

[B]

Okay. Well, artificial intelligence is going to become extraordinarily good at answering questions. The future of medicine will depend on whether physicians continue to ask the right ones. Technology can analyze the data. Physicians still will have to recognize the meaning. The real risk isn't that AI will replace physicians. The real risk is that it will slowly replace the habits of thinking that make physicians necessary. And in that vacuum, no one will be left to recognize or remember the meaning.

[A]

Very well said. Well, Dr. Lorio, thank you again for joining us in the podcast today. It's a great conversation and I hope to connect again in the future.

[B]

Thank you, Carly. I appreciate this conversation and time to be with you.