From Code to Action: A Systematic Review of Conceptualizations of Intelligence, Autonomy, and Decision-Making in Robotics Research

Abstract

Robotics research has increasingly focused on the interplay between intelligence, autonomy, and decision-making, yet the conceptualizations of these constructs remain fragmented across the literature. We systematically review and meta-analyze how these concepts are defined, operationalized, and measured in robotics, bridging the gap from algorithmic design to real-world action. The study synthesizes empirical evidence to quantify the relationships between theoretical frameworks and practical implementations, addressing inconsistencies in performance metrics, task completion, behavioral outcomes, and safety. Our analysis reveals a moderate overall effect size for performance metrics (, ), with stronger effects observed for task completion and planning (, ), while behavioral metrics show smaller but significant effects (, ). Safety and reliability metrics, however, exhibit negligible effects (, ), highlighting a critical gap in current research priorities. Methodologically, we employ a rigorous synthesis of quantitative and qualitative evidence, identifying trends in how intelligence is encoded, autonomy is constrained, and decisions are translated into actions. The findings underscore the need for standardized definitions and metrics to advance reproducible research in robotics. This work not only maps the current landscape but also provides a foundation for future studies aiming to align theoretical aspirations with empirical validation.