next up previous contents
Next: Child Development Up: Project Motivations Previous: Affective Computing

Related Work

There are several other projects in progress to develop affective toys. Rather than recognizing cues from the playmate, these `pet robots' respond to the playmate with an expression of an emotion. Some are scientific based, designed to realistically model the emotional states of a real animal. Others are more closely aligned with toys. These toy `pet robots' are twitchy, cuddly creatures, designed to entertain the playmate. A robot that emulates the behavior of a real animal is not a new concept. The Affective Tigger is not, however, an exercise in artificial intelligence. His motivations are not self guided, nor do they arise on their own without provocation. The Affective Tigger does not get hungry or thirsty. The reason for this is that he is designed to facilitate research into the emotional education of children, rather than to explore an experimental implementation of a theoretical emotional model.

There are four specific products on the market today that are related to the Affective Tigger. The first one on the scene was Microsoft with their 1997 release of a computer and video reactive Barney doll, and their 1998 release of a similarly designed, albeit slightly more sophisticated, Arthur doll. Both Barney and Arthur are facilitators toys. That is, they are responsive only to a predetermined set of cues from either the television or the computer. Their goal is to engage the child with that medium. The Affective Tigger, on the other hand, responds to the child directly, as a social partner, rather than simply facilitating a shared television or computer experience. Both Microsoft dolls have limited responsiveness to the child directly. For example, if you cover their eyes they comment on how dark it suddenly became. However neither one strives to assess the emotional mood of the play experience. Barney and Arthur are classified as educational toys, Barney teaches the preschooler rudimentary computer skills and keeps them company while they watch television, while Arthur also teaches the slightly older child how to tell time. The Affective Tigger on the other hand is instructing the child on something far less tangible. The Affective Tigger is a toy on which to practice emotional development skills and affective communication.

The rest of the similar products are not commercial special plush as much as they are robots. Omron, a Japanese company, has developed a robotic house cat named Tama [Tashiro Tashima, 1998], Sony has a robotic dog [Laboratory, 1999], and Honda Motor Corp has a bipedal humanoid named P-2. The focus of these groups is the implementation of a realistic emotion system. They are developing a behavior model for the cat and dog that incorporates emotions into the motivational system. The cat, for example, interacts with users and her surroundings, as well as having her own internal drives, sleep, curiosity, attention. The emotional expressions of the cat robot include satisfaction, anger, uneasiness, and disgust. Neither the cat nor the dog responds directly to the affect of the user. However, similar to the Affective Tigger, the robotic cat does reinforce `good' behaviors like petting and stroking, and she does become angry when the user repeatedly hits her. The Affective Tigger is an instructional toy. He responds directly to the child without ulterior motives such as hunger or sleep to confuse his behavioral expression. When a child makes the Affective Tigger unhappy, it is because of something that the child has done, and not that the Affective Tigger is hungry or sleepy or has some other unfulfilled need.

Another scientific robot being developed is Kismet, a robotic infant for social interaction. Cynthia Breazeal, working under professor Rodney Brooks, of the MIT Artificial Intelligence Lab, is studying the anthropomorphic phenomenon that occurs between mother and infant. Kismet has a range of facial expressions which he and his user both learn to associate with his different drives. The focus of this research is on the way humans impose meaning on every `coo' and `gurgle' an infant may make. Kismet has been developed to investigate the emotional behaviors of the robot as seen through physical interaction between human and robot. You can pat, stroke, hold or even beat any of these robots. The difference is that the Affective Tigger is not an implementation of an emotional model. His internal composition is very flat so the user can see the effects of her actions directly. All of these scientific robots have complex internal Artificial Life systems that are reflected in that they strive to satisfy their own needs, and they do not necessarily obey you. Rather than toys focused on the playmate, these robots are autonomous ego-centric entities with behaviors arising from various needs. The Affective Tigger on the other hand is user `centered'. Without the stimulus of a playmate the Affective Tigger sits lifeless on the shelf.

Finally, there is Furby, a toy made by TIGER toys and Hasbro. Furby is a non-descript animal with huge ears, moving eyes and a beak which makes it very different from a character like Tigger, who has a well developed personality and a story. Furby is reported to ``not only belch and pass gas but it `learns' as it matures, going from `Furbish' nonsense to such English phrases as `big light' when placed in the sun and `big noise' when subjected to traffic'' [Variety, 1999]. However, the idea that Furby is `learning' is questionable at best. Furby's only exhibition of an emotional state is when he begins to sing. However, what causes Furby to become happy enough to start singing is unclear, and probably random. The Affective Tigger, on the other hand, displays emotions in response to the playmate. His happy state is a form of reward to the playmate for gentle behavior. Furby's unresponsiveness to being smacked around, or turned upside down, make him seem less like a social friend and more like a toy. Cynthia Breazeal sums up Furby's attributes quite eloquently when she says:

I actually have 2 Furbys. I bought them to see what $34 of AI will get you these days. Turns out, it won't get you any AI. Not surprisingly, Furby is not a social creature. It doesn't have drives, goals, or any sort of internal agenda it has to satisfy. It doesn't perceive or really express any sort of emotions, feelings, affective states, etc. It doesn't learn, although that's what the Tiger folks would like kids to believe. However, they have hooked up some cute animated responses to various sensory triggers (switches, IR, inclination, simple sounds). They took the extra step of having some of these animated responses released as Furby gets `older' which is kind of cool. So, their behavior does change over time, although not because of learning. They can also respond to each other through IR, and their interactions get more sophisticated over time (again, not because of any learning). I think that's a nice extension of past toys. However, needless to say, my fears of having my Ph.D. embodied in a $34 toy have not come to pass (Excerpt from an email sent by Cynthia Breazeal (1999)).

In general, these affective toys and robots have begun to produce emotionally conscientious computer interfaces. The Affective Tigger is in the unique position of being an educational computer-toy. He is not only recognizing and synthesizing emotional information, but also using responsiveness to teach the playmate about actions and consequences.

The Affective Tigger's reaction is indirectly produced by the child's emotional behavior. Thus, the connection can be made that if the child is hurting him, the Affective Tigger becomes unhappy. And visa versa. ``We expect the child to discover the general rule from a set of experiences by means of induction. Script metaphor: to know the meaning of the word fear, to have the concept expressed by that word, is to know the sequence of experiences which occur in a fearful situation'' [Saarni and Harris, 1989]. Specifically, through normal play experience, the child learns what she can do to `make Tigger sad'. Once the child makes the connection between her behavior and his response, she has learned and demonstrated what I am referring to as emotional recognition. If the child can take the next developmental step and discover what she can do to `make him happy again', she has learned and demonstrated compassion and empathy. This is the beginning of emotional intelligence.

next up previous contents
Next: Child Development Up: Project Motivations Previous: Affective Computing

Dana L Kirsch
Tue May 25 08:59:22 EDT 1999