Abstract
An agent model that estimates advertisement performances is proposed. The agent notices and remembers frequently presented advertisements among many objects as the human brain subconsciously does. The mere exposure effect shows that these remembered objects are considered a person's profile indicating his or her preferences. If an advertisement is included in the profile, the agent supposes it's effective. In this model, independent component analysis is used to extract advertisement images from pictures that contain many objects. Computer experiments in augmented reality spaces confirmed that the advertisement performance depends on how properly a person perceives simultaneously presented objects.
