What is a social robot?
Because we are often asked what a social robot is and what a social robot should be able to do, we want to explain our understanding of social robots on this page.
Beginning of social robotics
Robots were originally developed to relieve humans of frequently recurring mechanical labor. For 20 years, the so-called “social robots” have also been researched, but they are supposed to take on different tasks.
Cynthia Breazeal was one of the first to develop “social robots” at the Massachusetts Institute of Technology (MIT) in the 1990s. In her book “Designing sociable robots” from 2002, she defines social robots through their abilities: They should be able to communicate and interact with people, build social relationships, adapt to their environment, learn throughout their lifetime and integrate new experiences into their understanding of the world and themselves. (see also https://en.wikipedia.org/wiki/Social_robot).
She goes on to write that their interaction must be designed to be intuitive for people. The behavior of the robot should be lifelike, so that it is easily “readable” and understandable for humans.
Intelligent products today
Breazell’s definition is now a few years old and in the meantime some of these criteria have been met by some “intelligent” products that are not social robots. Other criteria, on the other hand, do not seem likely to be fulfilled in the foreseeable future. Today, all interactive technical products already establish some sort of communication with humans. The so-called smartspeakers have established voice interfaces. For decades, the disciplines of usability engineering and user experience design have been concerned with making interaction with all kinds of devices intuitive. Software is made readable, for example, by displaying its status and possible interactions through graphical elements.
The vast majority of technical products also adapt to their environment by means of sensors which, for example, detect temperature, ambient lighting or radio signals. The requirements that social robots should learn for their whole lifetime and build an understanding of the world come from the domain of general artificial intelligence and are so demanding that these challenges may not be solved for decades. Consequently, this would mean that we would not be able to develop social robots that meet the above requirements for so long. It also seems questionable whether these capabilities are mandatory for social robots.
Social Intelligence as differentiating factor
We therefore argue that the definition of social robots should be based solely on the ability to actively build a social relationship with the user. This goal clearly differentiates social robots from any other product category. Most of the capabilities mentioned by Breazeal continue to serve as a basis, but the core capability of the social robot is its social intelligence, which enables it to purposefully and independently initiate a relationship-building process.
The criterion cannot simply be whether a relationship is established or not, but what degree of intensity and what qualitative level it is able to reach. This, in turn, leads to the question of what relationship quality must be achieved in order to be able to describe a machine as a social robot? How does one measure the quality of a relationship at all? Would it be sufficient to ask the human being about it or are there also objective standards? And what does a robot need to be able to do?
Social Resonance and Social Intelligence
You can find answers to these questions in the field of social sciences and psychology.
The sociologist Hartmut Rosa published the book “Resonanz” in 2016. In it Rosa defines the concept of resonance and presents it as a measure of successful relationships between subject and world in the sense of a “good life”. Its essence is based on the phenomenon of inner touch and being touched. In the individual an intensive perception takes place, followed by sustainable processing and then, as a reaction, a response, which in turn is perceived and processed by the counterpart. Life succeeds when we are in resonance with our environment and especially with our fellow human beings.
In order to establish a social resonant relationship, specific abilities of social intelligence are required. People who easily build relationships of high social quality are said to have a high social intelligence. That is why this form of intelligence should also be decisive for social robots.
The term social intelligence originated in the twenties, from the psychologist Edward Thorndike. For Thorndike, social intelligence means behavior that involves an intelligent way of dealing with and in our relationships: “acting wisely in relationships”. However, this descriptive definition does not help as an objective and measurable parameter of social intelligence.
The psychologist Daniel Goleman, known for his book “Emotional Intelligence”, has also written a book on “Social Intelligence” in which he divides the latter into concrete abilities. On the perceptual side, he distinguishes between affective and cognitive empathy, affinity and knowledge of social rules. On the active side there are the abilities of synchrony, self-expression, influence and care. The individual components are described specifically and are definitely measurable. In this definition, Goleman explicitly thought about the abilities of humans – not of animals or robots; abilities that allow humans to act together in a socially intelligent way.
Can robots also possess these abilities? Since Goleman does not proceed from sensations, but from observable characteristics, the argument that robots do not have real (empathic) sensations, but merely simulate them, becomes obsolete. From a behavioristic point of view, it does not matter whether the socially intelligent action arises from a “real” feeling or from a simulation. Thus, social robots should also be able to be assessed according to these criteria by observing behavior alone.
We have given our robot figure navel very expressive facial expressions with which it can show and reflect emotions. There are special three-dimensional lenses above the OLED displays so that the figure can make real eye contact. Both are essential basic requirements for social resonance.
How robots "perceive"
If Navel perceives how people are doing, processes this and communicates the processed result verbally and non-verbally, it should be possible to attribute a certain social intelligence to Navel. It goes without saying that the “perception” of the robot works differently from that of humans.
From a technical point of view, for example, the robot produces an image of the other person by using a camera. This is analyzed with various algorithms (e.g. neural networks) with regard to graphic elements in order to find the face and the facial features important for the emotion, such as eyebrows or lips, including their relative position. From this, a further process calculates which muscles are tense in humans. Since one knows, from extensive research done over the years, which muscles are active in which emotion, the robot is also likely to calculate the corresponding emotional state of its counterpart at this moment.
Since facial expressions alone do not always permit reliable statements, further algorithms also analyze the voice and movement patterns in a similar way. This means that the robot calculates the most probable emotion based on the same information that we or our mirror neurons use when we estimate the emotion of our counterpart. We also use neural networks, although these are biological and not comparable to those of current artificial intelligence.
Levels of social skills
With this ability, Navel possesses a specific form of empathy. Navel can also recognize similar principles when a person speaks and synchronizes his utterances with the utterances of the person. Navel’s perception and actions are founded on its character values and a user model that represents social rules. In addition, Navel can express its own state and take people and their needs into consideration. As a direct result, this will influence human actions and a continuous interaction loop is created. This means that social resonance is built.
As we all know, animals that live in social groups also have the ability to resonate with humans. This is the same reason why we can build such close social relationships with dogs. Research by neurologists has shown that dogs, just like humans, have mirror neurons that not only fire when they see their peers, but also when they observe humans.
Conversely, our mirror neurons also fire when we see dogs. A measurable goal for social robots should therefore be to stimulate these neurons in humans when interacting with social robots.
Social robots will not have the social intelligence of humans. Despite the clear definition and differentiation, the overall construction remains extremely complex and not easily controllable even with new methods of artificial intelligence. Reaching the social competence level of a pet such as a cat or a dog (in their relationship with humans, regardless of all the different facets) could, however, be a meaningful goal for social robots.
Definition of Social Robot
We therefore propose the following definition of social robots:
- A social robot should have social intelligence.
- It interacts with humans on both social and emotional levels.
- It builds social resonance and keeps people company.
It remains to be noted that social intelligence is a basic prerequisite for a social robot, but this alone does not guarantee a sympathetic lively character and thus a successful product. For this reason, social intelligence must be regarded as a minimum requirement to which Breazeal’s other described abilities need to be added, e.g. a lively character design as motivation factor.
In order to be perceived as alive, a robot must move and act independently and at the same time react continuously to its external environment. A lively behavior supports the readability and intuitiveness of the interaction and simultaneously strengthens confidence in the character.
For the character design, we orientate ourselves on characters from animated films that evoke a high degree of sympathy and expressiveness.
We believe we are able to create a kind of robotic being, which will have no consciousness similar to those of an animal or a human being, but which is more than just a dead product. It is an artefact comparable to the character of an animated film, but it is materialized and interactive. Through its social intelligence it can contribute for a pleasant society for the human being and provide well-being through resonance.