Beyond Digital Skills Why AI Literacy Must Become A Core Element of Medical Professionalism
Associate Professor (Department of Pathology) Quaid E Azam Medical College, Bahawalpur
Sara Reza
Artificial intelligence (AI) is no longer a distant possibility in medicine. It is already being used to summarize clinical information, support diagnostic decisions, draft patient communications, assist research, and generate educational content. Medical students and doctors are also using freely available generative AI tools, often without formal training or institutional guidance. The question is therefore no longer whether doctors should use AI. The more important question is whether they are sufficiently prepared to use it safely, critically, and professionally.
Traditional digital skills focus on the ability to operate electronic health records, search databases, communicate through digital platforms, and protect one’s online identity. AI literacy demands something more. It requires doctors to understand what AI can and cannot do, critically evaluate its outputs, recognize bias and uncertainty, protect confidential information, and remain accountable for decisions influenced by these systems. In this sense, AI literacy is not simply a technical skill. It is becoming an essential component of medical professionalism. From technological competence to professional responsibility. Professionalism has always required physicians to maintain competence, protect confidentiality, communicate honestly, recognize the limits of their knowledge, and place patients’ inter-ests above convenience. These responsibilities do not disappear when AI becomes involved. On the contrary, AI makes them more complex.
Generative AI systems can produce fluent and authoritative answers even when the underlying information is incomplete, inaccurate, or fabricated. The World Health Organization has highlighted risks related to false or biased outputs, automation bias, cybersecurity, privacy, and the inappropriate use of patient data1 A well-written response may appear clinically convincing, but fluency should not be mistaken for accuracy. The World Medical Association has therefore emphasized the concept of augmented intelligence: AI should strengthen rather than replace human judgement.
Its 2025 statement affirms that physicians must retain final authority over AI-supported decisions and remain accountable for patient care2 This is a crucial professional principle. A doctor cannot transfer responsibility to an algorithm, just as responsibility cannot be transferred to a laboratory report, clinical guideline, or decision-support system. Recent evidence illustrates why this matters. In a study evaluating large language models for drafting emergency department encounter summaries, only 33% of summaries produced by GPT-4 were entirely free of errors across all evaluated domains. Hallucinated information appeared in 42% of summaries, while 47% omitted clinically relevant information3 The study did not suggest that these tools have no value. Rather, it demonstrated that apparently useful outputs still require careful clinical review. The problem is not limited to inaccurate information. AI recommendations can also influence how clinicians think. An empirical study of AI-assisted diagnostic decision-making found that users sometimes agreed with incorrect AI recommendations particularly when they perceived the system as beneficial. Those without specialist expertise were more susceptible to this automation bias.4 Doctors therefore need the confidence and competence to challenge an AI-generated recommendation, even when it is presented with certainty.