The Rising Role of AI in Medical Diagnosis and Treatment
In a typical morning at a private hospital in Mexico City, an observant doctor notices a resident subtly checking their phone. In mere seconds, an AI-powered assistant provides a differential diagnosis and management plan, raising questions about authorship, accountability, and the erosion of critical thinking skills as machines mimic human reasoning.
The New England Journal of Medicine’s Perspective
A recent article in the New England Journal of Medicine proposes a pedagogical framework for supervising AI use among medical students and residents. The article argues that while large language models (LLMs) may appear to reason, their opacity and variability introduce educational risks such as de-skilling, non-skilling, and mal-skilling.
- De-skilling: Loss of skills due to over-reliance on AI for complex tasks
- Non-skilling: Failure to acquire necessary skills as AI takes over tasks
- Mal-skilling: Reinforcement of incorrect behaviors when AI provides flawed or biased outputs
Research indicates that frequent dependence on AI is linked to decreased critical thinking, and blindly accepting AI outputs leads to poorer performance in analytical tasks compared to not using AI at all.
DEFT-AI Method for Critical Thinking with AI
The article suggests that the solution is not to prohibit AI use but rather to teach critical thinking with AI. The proposed DEFT-AI method (Diagnosis/Discussion, Evidence/Evidence, Feedback/Feedback, Teaching/Teaching + Recommendation of AI use) aims to transform the “AI moment”—when a machine’s judgment requires a leap of faith—into an opportunity for guided critical thinking.
- Diagnosis/Discussion: Explore the student’s question and prompt refinement
- Evidence/Evidence: Evaluate supporting or contradicting evidence for AI output
- Feedback/Feedback: Provide specific recommendations on task delegation and co-construction with AI
- Teaching/Teaching + Recommendation: Balance centaur (human-AI collaboration) and cyborg (integrated AI) modes based on task risk
The article also highlights the importance of two missing competencies in current programs: knowing how to ask (prompt design) and knowing how to verify (confirmation bias avoidance, request for AI reasoning, verification against guidelines, literature, and clinical judgment).
Medicine and Pharmacy Reports’ Perspective
Another recent review in Medicine and Pharmacy Reports addresses the concern that AI will replace or support medical specialties.
The review concludes that AI excels in mass data and image analysis tasks (radiology, pathology, ophthalmology, dermatology) but falls short in areas requiring nuanced judgment, empathetic communication, and situated clinical reasoning.
The recommended approach is integration that supports professionals rather than replacing them.
Specialty-wise Impact of AI
The review clarifies that specialties more susceptible to partial replacement are those dominated by repetitive patterns (e.g., image interpretation). Conversely, specialties resistant to full automation include psychiatry, pediatrics, internal medicine, emergency care, intensive care, and surgery—areas where interpreting silence, negotiating uncertainty, balancing family values, and making quick decisions cannot be reduced to variables.
AI can assist through chatbots and monitoring agents but cannot replace the human relationship in psychiatry. Pediatrics and emergency care require holistic assessment, family communication, and dynamic prioritization based on incomplete information—all of which demand clinical creativity not yet automated.
The review also notes that AI is already enhancing access and efficiency through telemedicine, virtual assistants, predictive analytics for resource allocation, and low-cost diagnostic devices. When properly governed, these tools can lower costs and increase access in resource-constrained settings.
Mexico’s Context: Balancing AI Integration and Human Expertise
While the federal government of Mexico’s Fourth Transformation downplays the extent of deterioration, the nation grapples with consequences such as hasty system transitions, insufficient funding, infrastructure gaps, exhausted personnel, and digital divides leading to delayed diagnoses and public distrust.
The focus should not be on being “for” or “against” AI but rather on how to integrate it thoughtfully and for whose benefit.
AI in National Health Care Protocols (PRONAM)
The question remains: To whom are the National Health Care Protocols (PRONAM)—guidelines based on the best scientific evidence for disease prevention, diagnosis, and treatment determined by the General Health Council—directed: Centaurs (human-AI collaboration), Cyborgs (integrated AI), or do they ignore AI use altogether?
The author (www.ectorjaime.mx) is a general surgery specialist, certified in public health, with doctorates in health sciences and public administration. He is a legislator and advocate for public health in Mexico, a reelected PAN group member in the LXVI Legislature.
