Ning Tang¹ (), Enjie Xu¹, Siyi Gong², Jifan Zhou¹, Mowei Shen¹, Tao Gao²; ¹Zhejiang Univeristy, ²UCLA

In the real world, agents are not omnipotent. Their motions often deviate from intended goals due to physical constraints, as exemplified by a leashed dog moving erratically while chasing a cat. Humans typically perceive this animacy, yet empirical psychophysical studies contradict this observation, indicating that spatial deviations and line connections between objects can impair perceived animacy. In this study, we propose that these limitations are special cases within a more unified theory, where animacy is inferred by intelligently interpreting various forces imposed on the agent through the lens of intuitive physics. Previous failures are caused by physically arbitrary deviations and connections. We aim to investigate whether perceived animacy improves when deviations from goal-directed motion are explained away by a constraint imposed by a physical line. Moreover, we examined two intuitive physics models in perceived animacy: first, a realistic Newtonian Physics model implemented by a modern physics engine; second, a schematic force dynamics model from cognitive linguistics, where an agonist’s motion is determined by its intrinsic force and the force imposed by an antagonist. While the former excels in explaining perceptual physical judgments, the latter offers a linguistic framework of the conceptual structure underlying verb usage in language. Our findings reveal that perceived animacy, assessed through free reports and visual searches, significantly increases when deviations and connections are physically explainable. Importantly, the force dynamics model yielded higher perceived animacy than Newtonian Physics. Extensive testing of physical parameters confirmed that the weaker results of Newtonian physics were a generic phenomenon. Collectively, these results indicate that vision effectively perceives animacy by explaining spatial deviations and line connections through intuitive physics. The intuitive physics in perceived animacy is better modeled by force dynamics from linguistics, making it an interesting case study showing that there is a common conceptual structure underlying both vision and language.