Frederic Kaplan

08.06.2005 Interview by Vanessa de Sa (Galileu, popular science magazine in Brazil)

"The robots of tomorrow don’t have to be like the one imagined by science-fiction writers of the last century."

What kind of study of do (or have done) to evaluate the acceptance of the robot by children and adults? What were the main results?

We have conducted a series of study with children of various ages to see how they were spontaneously reacting during their first encounter with the robot. Infants (10y old) have been regularly found to engage in some form of experimental test of the behavior of the robot (e.g. placing the ball near, then far from the robot to see its perceptual capabilities). On the contrary, most adults were less keen to spontaneously interact with the robot, skipping this experimental phase to directly make comments about their impressions regarding the machine. These investigations suggest that from an initial basis of natural expectancies, experience and culture are likely to change in an important manner our immediate reaction to robots.

Robots because they are autonomous, physical artifacts tend to spontaneously foster interaction patterns that are usually characteristic of our experience with living animals. A crucial design issue is whether life-like design produces higher immediate experienced value or on the contrary introduces the machine in a misleading way. Human perception of automata has been a subject of reflection long before the arrival of the first robots. Life-like behavior can trigger interest or fascination, but can also be, in some cases, the source of some ‘uncanny’ feelings. Typically, this happens when the behavior or the appearance of the machine becomes very life-like and therefore violate our expectations about perceptual features that distinguish machines and animals. Freud was certainly one of first to put a word on this feeling. He calls it unheimlich, literally was is not familiar. Paradoxically, it also means what was so intimate that it is now hidden and secret. This uncanny feeling may therefore result from the interplay between natural expectancies and experienced and cultural ones.

How do real animals respond to the robot?

In collaboration with a team of ethologists from the Eotvos Lorend University in Budapest, a set of experiments have been conducted to study how dogs react to a four-legged robot like AIBO. In one situation, we let the robot approach as the dog was eating a big piece of meat. In such a context, several dogs displayed forms of aggressive behavior towards the robot and sometimes attacked it, considering it could be a potential competitor for the meat. However, in general, dogs clearly made the difference between AIBO and a real puppy of the same size.

What can make entertainment robots interesting in the long run?

We explore this issue by considering robots capable of autonomous development and long-term learning. The richness of the behavior of such a robot increases with its developmental trajectory: what the robot has seen, what situations it has encountered, who it has interact with, etc. If previous interactions shape the robot’s behavior in a distinctive way, entrainment dynamics between the user and its machine emerge. In such situations, the more the user interacts with the robot, the more the robot’s behavior changes, leading through a positive feedback loop to continuously renewed forms of interactions with the machine. We believe understanding such self-reinforcing dynamics is the key for sustaining long-term intrinsically motivating interactions.

Can robots learn like children do?

Children are capable of acquiring new competencies in a continuous and open-ended manner and, as they do so, they develop increasingly complex forms of awareness about their environment. Today, no robots are capable of such form of continuous learning. With my colleague Pierre-Yves Oudeyer, we have designed a new software prototype that permits to create robots that show some form of curiosity. Such robots always want to learn more about their environment. They will not develop exactly like a child, but like children, they start to be capable have the ability to learn in an open manner.

In your opinion, will robots be able to be definilately inserted in human society?

The imminence of the arrival of robots has been announced so often, that one should be careful in making predictions in that respect. Such evolution may happen, but in a way different from the scenario classically described. In the 50s, in the 60s, in the 70s, many eloquent representations of our future life were showing a happy family of the XXIe century in an apartment literally full of robots: robot maids, robot companions, robot nanny, robot guards. The robots of tomorrow don’t have to be like the one imagined by science-fiction writers of the last century. We can be free to imagine other forms of future life in which we will cohabit with robots, or with entities that only vaguely resemble to what we used to call robots.