What makes some patterns of motion seem "alive" or "intentional?"
Animate motion. Certain patterns of object motion can convey the impression that the object is alive—famously demonstrated in a short film by Heider & Simmel (click here for a modern paraphrase), in which simple geometric shapes convey goals, emotions, and even personalities solely by their patterns of motion. In our research we have tried to reduce this phenomenon down to its barest essentials. Our displays typically consist of a single simple object (a dot or rectangle) moving through the visual field. When the object suddenly increases speed and direction simultaneously, it tends to look more animate. The factors influencing the way this velocity change is interpreted tell us a lot about the underlying mechanisms of animacy attribution. For example If the object's body is aligned with the motion direction before the velocity change but misaligned afterwards (panel c in the figure), it destroys the animate percept—objects that do not appear to be in control of their own motion do not appear alive. Conversely if there is a visible object along the new motion direction—looking like a "goal"—then the object appears more animate (panel d), as if it decided to change direction. One of the conclusions from this work is that animacy judgments are reserved for objects that appear to have "minds of their own."
Formal models of the attribution of mental architecture. Computationally, how do you decide whether an object you're observing has a mind? In more recent work, we have modeled this process by assuming that observers attempt to attribute to the agent the simplest mental (computational) architecture that is sufficent to explain the actions that have been observed. If this attributed architecture has key mental properties, than the object is interpreted as a mental (and thus animate) agent. Our theory formalizes how computational architectures that have goals or intentions can be mathematically distinguished from those that don't.
Links:
