Context-mediated behavior for intelligent agents
International Journal of Human-Computer Studies - Special issue: using context in applications
| Hi-index | 0.00 |
In this paper we discuss the role of spatial context in interpreting and understanding actions from the perception of static images. Our proposal is that people use spatial context, which is to say the relations among objects, to infer actions and make predictions about future states even when no physical events are directly perceived. We begin by presenting an overview of relevant theories for action understanding. We then discuss the role of spatial context as a prerequisite for acting as well as for action goal attribution, by presenting and discussing an experimental study. The addressed question is how representation of motion in static images maps onto perception of an action goal. We tested the extent to which action representations attributed to a static stimulus can be changed when the physical relational properties of objects are manipulated. Results show that the goal attribution, considered as a highly informative component for the interpretation of the perceived action, is carried out by a contextual categorization process which takes into account both the objects' physical and relational properties, and the semantic relationships between them.