Almost Alive: Robots and Androids

Abstract

Life-likeness is a property that can be used both to deceive people that a robot is more intelligent than it is or to facilitate the natural communication with humans. Over the years, different criteria have guided the design of intelligent systems, ranging from attempts to produce human-like language to trying to make a robot look like an actual human. We outline some relevant historical developments that all rely on different forms of mimicry of human life or intelligence. Many such approaches have been to some extent successful. However, we want to argue that there are ways to exploit aspects of life-likeness without deception. A life-like robot has advantages in communicating with humans, not because we believe it to be alive, but rather because we react instinctively to certain aspects of life-like behavior as this can make a robot easier to understand and allows us to better predict its actions. Although there may be reasons for trying to design robots that look exactly like humans for specific research purposes, we argue that it is subtle behavioral cues that are important for understandable robots rather than life-likeness in itself. To this end, we are developing a humanoid robot that will be able to show human-like movements while still looking decidedly robotic, thus exploiting the our ability to understand the behaviors of other people based on their movements.