Evaluating Agentic Artificial Intelligence in Healthcare Using a Perceive Reason Act Learn (PRAL) Framework: A Systematic Review

Abstract

Abstract Recent advances in artificial intelligence have led to increasing interest in agentic systems capable of goal-directed behavior and autonomous tool use. Such systems are proposed in healthcare as an extension beyond traditional generative models. Which primarily function through prompt-based interaction. However, existing studies are varying widely in how autonomy is defined, implemented and evaluated. Which makes it difficult to assess the actual maturity of agentic AI in clinical contexts. In this paper a systematic review of agentic AI applications in healthcare published between 2023 and 2025. The papers are evaluated using the Perceive Reason Act Learn (PRAL) framework. By analyzing systems according to their behavioral capabilities rather than outcome metrics alone. The review highlights substantial heterogeneity in reported autonomy levels. The literature review findings indicates the perception and action components are frequently implemented. Also the reasoning transparency and learning mechanisms are often limited or underreported. The paper is concludes by identifying methodological gaps and also by proposing directions for standardized evaluation of agentic AI in healthcare. .

Ähnliche Arbeiten