Artificial Intelligence and the Ethical Foundations of Cardiothoracic Surgery: Evidence, Accountability, and the Limits of Delegated Judgment.

چکیده اصلی

Artificial intelligence (AI) is moving rapidly from retrospective prediction and image analysis into treatment selection, operative planning, intraoperative guidance, and postoperative prognostication in cardiothoracic surgery. This transition raises an ethical problem that cannot be resolved by model accuracy alone: when an algorithm begins to shape a high-stakes clinical decision, the distribution of knowledge, authority, and responsibility also changes. This review synthesizes cardiothoracic and closely related medical evidence available through August 2026, with emphasis on quantitative performance, human-AI interaction, bias, patient autonomy, and liability. The available evidence is simultaneously encouraging and cautionary. Machine-learning approaches can improve predictive performance and AI-assisted thoracic planning can reduce errors and increase procedural consistency; however, these gains have not consistently translated into superior patient outcomes. Human-AI studies similarly demonstrate that improved accuracy may coexist with automation bias and overacceptance of algorithmic recommendations. Evidence of demographic performance disparities and limitations in the representativeness of training and validation datasets further raises concerns regarding fairness and equitable access to care. On this basis, we argue that cardiothoracic AI should be governed according to the level of decision influence rather than by technology type alone. We distinguish non-delegable professional duties, distributed system responsibilities, and non-transferable patient authority, and propose an Ethical Heart Team Framework for converting algorithmic output into ethically defensible clinical action.

متن کامل اصلی