BETWEEN ALGORITHMS AND CLINICAL DECISIONS: THE ROLE OF ARTIFICIAL INTELLIGENCE IN BUILDING A HEALTH DIAGNOSIS

Abstract

The incorporation of Artificial Intelligence (AI) in healthcare diagnosis represents a qualitative transformation in clinical practice, distinct from all previous technological innovations due to its ability to analyze, interpret, and generate inferences from large volumes of data. This chapter examines this transformation from its historical roots, tracing the paradigms that have shaped diagnosis from Hippocratic medicine, with its empirical knowledge centered on the patient's narrative, to the digital age, passing through anatomical-clinical medicine, the advent of diagnostic technology, and the consolidation of Evidence-Based Medicine. It then presents the technical foundations of the main AI approaches, including machine learning, deep learning, natural language processing, and Big Data integration, discussing how each of these approaches operates and what their specific capabilities and limitations are in the clinical context. The chapter analyzes the implications of this incorporation for the very nature of diagnosis, which shifts from a punctual interpretive act to a probabilistic and continuous process, highlighting the uncertainty inherent in clinical practice and reconfiguring the dynamics of authority between professional and system. This chapter examines phenomena such as algorithmic opacity, systematic biases arising from unrepresentative historical data, the tension between patient uniqueness and the generalizing logic of algorithms, and the risk of automation bias in decision-making. Contemporary AI applications are discussed in their most relevant contexts, including diagnostic imaging, dental practice, and clinical decision support systems, paying attention both to the performance demonstrated in the literature and to the limitations that prevent the direct transposition of research results to the real world. The chapter also addresses the ethical dimensions of AI in diagnosis, dealing with principles such as autonomy, equity, and accountability, as well as the regulatory challenges imposed by the adaptive nature of these systems, and discusses the implications for professional health training, arguing that preparing the contemporary professional requires integrating technological fundamentals with clinical reasoning, without replacing the human skills that algorithms are not yet able to reproduce. It is concluded that the future of diagnosis depends on the deliberate construction of a partnership between humans and algorithms that preserves the analytical capacity of AI and the contextual, ethical, and human judgment of the healthcare professional, with shared responsibilities among those who develop, regulate, train, and provide care.