EARLY TARGET PREDICTION IN ACTION OBSERVATION

Previous research has established that observers can predict action targets through hand preshaping. However, two critical questions remain unexplored: how predictions adapt to the available kinematic information and evolve throughout the movement timeline. We address these fundamental gaps by combining kinematic analysis with machine-learning approaches that differentiate between motor and visual cues. Using motion capture technology, we recorded reach-to-grasp actions toward large and small objects and had participants predict target size from hand kinematics at varying time points. Our analysis revealed that prediction performance not only evolved with increasing kinematic information but, crucially, differed significantly between target size choices. To provide insight into the underlying processes, we developed a comparative framework using two distinct machine learning approaches: Support Vector Machines (SVM) modeling kinematic information and CNN-RNN networks extracting visual patterns. The stronger alignment between human performance and SVM predictions offers empirical evidence that kinematic cues, rather than visual patterns, mostly guide target prediction. These findings advance our understanding of action prediction and have significant implications for social cognition and human-machine interaction. PUBLIC SIGNIFICANCE STATEMENTUnderstanding others intentions by observing their hand movements is crucial for social interaction, from passing objects to coordinating complex tasks. This study reveals that people use different cues to predict whether someone is reaching for a large versus a small object from the earliest stages of hand movement. By comparing human performance with Artificial Intelligence models, we found that people primarily rely on motor cues to make these predictions. These insights could improve rehabilitation techniques for individuals with social interaction difficulties and enhance the design of intuitive robotic assistants.