Premed · Premed · Medical Ethics Humanities
Lecture 20: AI and Emerging Technologies in Medicine
Foundations of Medical Ethics and the Health Humanities
Learning Objectives
By the end of this lecture, students will be able to:
- Describe the current and near-future applications of artificial intelligence in healthcare
- Analyze the ethical challenges posed by AI including bias, transparency, accountability, and privacy
- Evaluate the impact of emerging technologies (telemedicine, wearables, digital therapeutics) on the patient-provider relationship
- Discuss the implications of algorithmic decision-making for autonomy, justice, and trust
- Articulate principles for the ethical integration of technology into clinical practice
Lecture Content
I. AI in Medicine: Current Applications
Artificial intelligence is rapidly transforming healthcare across multiple domains. In diagnostics, AI algorithms can detect diabetic retinopathy, skin cancer, breast cancer on mammography, and lung nodules on CT scans. Some AI systems match or exceed radiologist performance in specific, narrow tasks. AI is also being applied to pathology for tissue sample analysis and to ECG interpretation for detecting conditions like atrial fibrillation and hypertrophic cardiomyopathy.
Clinical decision support tools use AI to analyze electronic health record data in real time. Sepsis prediction algorithms, such as Epic's sepsis model, flag patients at risk. Drug interaction checkers, dosing calculators, and risk stratification tools assist clinical decision-making. Natural language processing extracts information from clinical notes and summarizes records.
Administrative and operational AI addresses scheduling optimization, billing, and prior authorization automation, with the potential to reduce the administrative burden that is a major driver of physician burnout. In drug discovery and development, AI models identify drug candidates, predict molecular interactions, and accelerate clinical trial design. Robotic surgery systems, such as the da Vinci platform, provide AI-assisted precision, tremor reduction, and enhanced visualization, though fully autonomous surgical robots remain experimental and raise unique liability questions.
II. Ethical Challenges of AI in Healthcare
Bias and fairness represent perhaps the most pressing ethical concern. AI systems learn from historical data, and if that data reflects existing disparities, the AI will reproduce and amplify them. A striking example: an algorithm widely used in US hospitals to identify patients needing extra care was found to systematically underestimate the needs of Black patients because it used healthcare spending -- which is lower for Black patients due to systemic barriers -- as a proxy for illness severity (Obermeyer et al., Science, 2019). Training data bias is also a concern, as datasets that underrepresent certain populations lead to AI that performs worse for those groups (such as dermatology AI trained primarily on light skin). Mitigation requires diverse training data, bias auditing, fairness metrics, and ongoing monitoring.
Transparency and explainability pose what is known as the "black box" problem. Many AI systems, especially deep learning models, cannot explain why they reached a particular conclusion. This challenges informed consent: how can a patient consent to a diagnosis or treatment recommendation they cannot understand? Physicians may also struggle to exercise clinical judgment when they do not understand how the AI reached its conclusion. Explainable AI (XAI) is a growing field focused on making AI reasoning interpretable.
Accountability and liability raise fundamental questions. When an AI makes an error, who is responsible -- the physician who relied on it, the hospital that deployed it, or the company that developed it? Current legal frameworks are not well equipped to assign liability for AI-related harm. The physician must remain the final decision-maker; AI should augment, not replace, clinical judgment.
Privacy and data security are critical concerns because AI requires vast amounts of patient data for training and operation. Risks include data breaches, re-identification of de-identified data, and unauthorized use of data for commercial purposes. Patients may not know or understand how their data is being used to train AI systems. Technical approaches such as federated learning and differential privacy aim to protect data while enabling AI development.
<image>A four-quadrant ethical framework for AI in healthcare. Top-left: "Fairness" (Does the AI perform equitably across demographic groups? Is the training data representative? Are biases identified and mitigated?). Top-right: "Transparency" (Can the AI's reasoning be explained? Can physicians and patients understand the basis for recommendations? Is the algorithm auditable?). Bottom-left: "Accountability" (Who is responsible when AI errs? Is there human oversight? Are there mechanisms for redress?). Bottom-right: "Privacy" (How is patient data collected, stored, and used? Is consent obtained? Are data security measures adequate?). A central circle reads: "Ethical AI in Medicine: all four dimensions must be addressed."</image>
III. The Physician-AI Relationship
The current consensus holds that AI should augment physician decision-making rather than replace it. The physician provides context, values, empathy, and judgment that AI cannot. The "human in the loop" principle means the physician reviews, interprets, and retains final authority over AI-generated recommendations.
Automation bias -- the tendency to defer to automated systems even when they are wrong -- is a real risk. Physicians may stop exercising independent judgment if they become overly reliant on AI. Maintaining strong clinical skills and critical thinking is the necessary counterbalance.
Deskilling is a related concern. If AI handles tasks previously performed by physicians, such as radiology reads or ECG interpretation, the underlying clinical skills may atrophy. Medical education must adapt to a world where AI is a partner in clinical practice.
Trust calibration is essential: physicians must learn to trust AI appropriately, neither blindly accepting nor reflexively dismissing its output. Understanding the AI's limitations, error rates, and areas of uncertainty is critical for appropriate use.
IV. Telemedicine and Digital Health
Telemedicine expanded dramatically during COVID-19 and has become a permanent feature of healthcare delivery. Its benefits include increased access for rural and underserved populations, mobility-limited patients, and reduced travel burden. It offers convenience for chronic disease management, follow-up visits, and mental health care. It also supports public health by reducing disease transmission during pandemics.
Ethical concerns about telemedicine include the digital divide, in which patients without internet access, smartphones, or digital literacy are left behind. Quality of care may be affected because some conditions require physical examination and telemedicine may miss important findings. Privacy is a concern when home consultations lack adequate privacy due to family members present or unsecured internet connections. Licensure and liability become complicated when physicians practice across state or national borders. The patient-provider relationship may be affected, as telemedicine can reduce the human connection, empathy, and trust that come from in-person encounters.
Wearable devices and remote monitoring, including continuous glucose monitors, smartwatches detecting arrhythmias, and remote blood pressure monitoring, offer benefits such as early detection, real-time data, and patient empowerment. Concerns include data overload for physicians, patient anxiety from constant monitoring, data privacy, and the reliability of consumer-grade devices.
Digital therapeutics are software-based interventions for conditions like substance use disorders, insomnia, and diabetes management. FDA-approved examples exist, such as reSET for substance use. Questions remain about who serves as the "provider," how these are regulated, and what constitutes an adequate evidence base.
<image>A Venn diagram with three overlapping circles representing the intersection of technology and patient care. Circle 1: "Access and Convenience" (telemedicine, remote monitoring, digital therapeutics -- expands reach of care). Circle 2: "Quality and Safety" (AI diagnostics, clinical decision support, precision medicine -- improves accuracy). Circle 3: "Human Connection" (empathy, trust, shared decision-making, narrative understanding -- the core of the therapeutic relationship). The overlap of all three: "The Goal: technology that enhances access and quality WITHOUT eroding the human dimensions of care." Areas where only two circles overlap are noted: Technology without humanity risks depersonalization; humanity without technology risks inequity; access without quality risks harm.</image>
V. Emerging Frontiers
Precision medicine tailors treatment to the individual based on genetics, biomarkers, lifestyle, and environment. Ethical issues include access disparities, data privacy, and the risk of reducing the patient to a data profile.
Brain-computer interfaces (BCIs) are devices that read or stimulate brain activity, with applications in paralysis, epilepsy, and depression. Ethical issues include cognitive liberty, identity, consent for brain-altering interventions, and data security for neural information.
3D bioprinting and synthetic biology have the potential to print organs, tissues, and personalized medical devices. In the long term, these technologies could address the organ shortage, but near-term ethical questions about regulation, safety, and access must be addressed.
Fully autonomous systems, including autonomous surgical robots and AI-driven drug prescribing without physician review, raise fundamental questions about the nature of medical practice and the physician's role.
VI. Principles for Ethical Technology Integration
Several principles should guide the ethical integration of technology into clinical practice. Patient-centeredness demands that technology serve the patient's interests, not just institutional efficiency or commercial gain. Equity requires ensuring new technologies reduce rather than widen disparities. Transparency means that patients and physicians should understand how technology works and its limitations. Human oversight requires that physicians retain meaningful decision-making authority. Continuous evaluation calls for ongoing assessment of safety, effectiveness, fairness, and impact on the patient experience. Inclusive design involves patients, communities, and diverse stakeholders in the development and deployment of health technologies.
