The Algorithm’s Shadow: Why Critical Thinking is the Only Antidote to Medical ‘De-Skilling’

February 21, 2026

NEW DELHI — At the high-tech corridors of the India AI Impact Summit 2026 this week, a sobering warning echoed through the halls of the Bharat Mandapam: the very technology designed to save lives might be eroding the fundamental skills of the people holding the scalpel. Jan Herzhoff, President of Global Health Businesses at Elsevier, cautioned an audience of policymakers and physicians that an over-reliance on Artificial Intelligence (AI) without rigorous critical thinking risks a “de-skilling” crisis in healthcare. This warning follows the release of a landmark 2025 study revealing a startling 20% drop in the ability of experienced clinicians to detect potentially precancerous polyps during colonoscopies when their AI assistants were turned off.

As AI tools transition from experimental curiosities to everyday staples in clinics and hospitals, the medical community is grappling with a profound paradox: how to embrace the massive benefits of technology while preserving the human expertise that underpins medicine.

The ‘Deskilling’ Dilemma: Evidence from the Front Lines

The core of this debate centers on a pivotal 2025 study published in The Lancet Gastroenterology & Hepatology. Researchers tracked endoscopists at four major Polish centers who had integrated AI-assisted polyp detection tools into their practices since late 2021.

The study compared Adenoma Detection Rates (ADR)—a gold-standard quality metric measuring a clinician’s ability to find precancerous polyps—across 1,443 procedures. The findings were stark: in colonoscopies performed without AI after a period of prolonged use, the ADR among experienced professionals fell from 28.4% to 22.4%. This 6% absolute drop represents a significant decline in the clinician’s independent diagnostic power.

“Clinicians are extremely busy,” Herzhoff noted during his summit address. “They send questions to an AI, get an answer, and rely on that response without critical thinking. This bypasses the natural process of knowledge expansion and building new connections.”

The concern is not that the AI is wrong—in fact, AI-assisted ADR is often 1.5 to 2 times higher than human-only rates—but rather that doctors may become “passengers” in the diagnostic process. When the technology is absent or fails, the clinician’s own “internal algorithm” may have grown rusty.

Expert Perspectives: The Paradox of Automation

Medical professionals not involved in the Polish study are calling for a proactive approach to this “automation bias.” Dr. Kosta Katsaros, a specialist in healthcare management, describes this as a “skills erosion” that can affect everything from diagnostic reasoning to physical examinations.

“AI is an incredible tool for augmentation, but if we outsource our thinking entirely, we risk losing the nuanced, ‘gut-feeling’ expertise that comes from years of manual practice,” Katsaros says.

However, the cause of this skill decline might not be the AI alone. Dr. Venet Osmani suggests that workload may play a crucial role. In the Lancet study, the total number of procedures nearly doubled during the observation period.

“It’s possible that fatigue from a higher patient volume, enabled by the speed of AI, contributes to the drop in performance more than the tool itself,” Osmani noted in a recent reaction to the findings.

A Global Perspective: From Super-Specialists to Community Health

While the risks are real, the potential for AI to bridge healthcare gaps remains undeniable. In India, Elsevier’s DIISHA project is using AI to upskill more than one million Accredited Social Health Activists (ASHA) workers. Pilot feedback suggests these community health workers are using AI to follow protocols more accurately and feel more confident in their roles.

The key, according to summit participants, is democratization with regulation. Tools like ClinicalKey AI are being designed with “trust markers”—such as journal impact factors and verified citations—to help students and professionals differentiate between high-quality data and AI-generated “hallucinations.”

Public Health: What Consumers Need to Know

For the general public, the implications of de-skilling are significant. Colorectal cancer is a leading cause of cancer-related deaths globally. If a clinician misses a precancerous lesion because they have grown overly reliant on an AI tool that fails to flag it, the outcome can be life-threatening.

What should patients do?

• Inquire about AI use: Don’t be afraid to ask your doctor if AI tools were used in your diagnosis.

• Ask for human oversight: Ensure that a qualified professional has reviewed the AI’s findings and can explain the reasoning behind them.

• Use wearables wisely: Consumers using AI-driven health apps or wearables should view the data as a conversation starter for their next doctor’s appointment, not as a definitive diagnosis.

Limitations and the Road Ahead

It is important to note that the Lancet study was observational and retrospective, meaning it shows a correlation but doesn’t prove that AI caused the skill drop. The study also focused on experienced endoscopists with a median age of 61. It remains unclear how AI will impact younger trainees who are learning these skills alongside the technology from day one.

Critics of the “de-skilling” narrative argue that if AI-assisted performance is consistently better than human-only performance, the loss of unassisted skills may be a trade-off worth making—much like society has traded mental math skills for the speed of calculators.

Strategies for the Future

To prevent a total erosion of medical expertise, experts recommend several hybrid models:

1. AI Support, Not Substitution: Mandate that clinicians perform an initial review before turning on the AI “second set of eyes.”

2. Unassisted Drills: Medical institutions should track performance metrics during occasional “AI-free” shifts to ensure baseline skills remain sharp.

3. Critical Appraisal Training: Medical curricula must evolve to teach students how to verify AI-provided references and identify “prompt poverty”—the failure to provide the AI with enough context for an accurate result.

As Prime Minister Narendra Modi urged at the summit, the goal must be to make AI inclusive while safeguarding against its unintended consequences.

The bottom line: AI is a powerful co-pilot, but the human must always remain in the cockpit, hand on the controls and eyes on the horizon.

Medical Disclaimer: This article is for informational purposes only and should not be considered medical advice. Always consult with qualified healthcare professionals before making any health-related decisions or changes to your treatment plan. The information presented here is based on current research and expert opinions, which may evolve as new evidence emerges.

References

• The Economic Times. (February 20, 2026). “Critical thinking, not reliance on AI, will protect against de-skilling: Elsevier’s Jan Herzhoff.”