Browsing by Subject "Healthcare"

Artificial Intelligence is gaining prominence across various sectors, including finance, healthcare, and social security. Notably, its potential in healthcare stands out, with Machine Learning and Deep Learning offering opportunities to enhance diagnosis and treatment decisions. By leveraging health data and genetic information, AI can effectively diagnose diseases and personalize treatments to suit individual patient needs. Recognizing this potential, substantial investments are being made in healthcare AI development. Trade secret protection emerges as a valuable tool to safeguard the intellectual assets underpinning the design, development, and implementation of AI-driven healthcare systems, aligning with the European Union Trade Secret Directive framework. However, alongside its benefits, AI introduces inherent risks, notably its opaque nature, often referred to as a "black box". The European Union's regulatory framework towards trustworthy AI, addresses this by emphasizing explicability as a fundamental ethical principle for trustworthy AI. In this thesis, explicability is framed as a principle aimed at enhancing stakeholders' understanding of how AI systems operate and who bears responsibility for their functioning. Transparency and accountability serve as foundational pillars, achieved through disclosure and effective communication. Consequently, the thesis examines the obligations of transparency and information provision outlined in the AI Act to analyze its stance on explicability. Given these contrasting concepts, it is crucial to examine the dynamic between trade secrecy and explicability from diverse stakeholder perspectives within healthcare AI systems. Consequently, the thesis delves into the viewpoints of public authorities, healthcare professionals, and patients. It concludes that while explicability is vital for fostering trustworthy and ethical healthcare AI, disclosure obligations should not affect trade secrets, thereby discouraging innovation. Moreover, overwhelming patients with excessive information could impair their ability to make informed decisions. To address this balance, the thesis briefly explores practical and theoretical approaches, including Explainable Artificial Intelligence techniques, human oversight, and communication by design and by default. These approaches aim to provide explanations and achieve explicability while safeguarding trade secrets and promoting innovation in healthcare AI.