"Is it ever too young to start?" is the question the Volkswagen Foundation put to a panel of experts on the topic of how artificial intelligence can contribute towards the learning process. Will social robots soon be looking after our little ones in kindergartens and preparing them for the digitalized world of tomorrow? Will robots one day replace school teachers and give lessons independently? At least not in the foreseeable future: This was something quickly agreed on during the podium discussion moderated by NDR-television’s editor Ulrike Heckmann – and was perhaps one of the reasons why several questions remained unanswered.
AI is still a long way off from being established in the education system
Computer scientist and physicist Prof. Dr. Stefan Kopp, who heads the research group "Cognitive Systems and Social Interaction" at the University of Bielefeld, has been doing research on learning and teaching processes for many years. The 46-year-old explained that the research goal is to arrive at automated and at the same time personalized knowledge acquisition that leads to better learning outcomes. Such tailor-made learning could be expected to make access to education easier, more flexible and thus also more equitable. AI systems can already be used in giving single lessons and can enable children with the appropriate disposition to learn in a self-determined and anxiety-free way. So, will social robots ultimately be used as assistant teachers alongside their human counterparts?
That is still a very long way off, admitted Kopp self-critically. Although AI research has been going on for more than 30 years, as yet nowhere in the world has an AI system been firmly established as part of the normal education system. "Unfortunately, the outcome has been no more than isolated individual projects." Kopp did refer to a study from Iran, though, where a teacher runs a class together with a robot – but this remains an exception. On the whole, it can be said that nowhere can a system be found that exhibits robust results in everyday environments. "We must therefore still do a lot of research to make the systems work in everyday environments." Calling for patience in the development and implementation of AI, Kopp likened the situation to the very first flying machines, which were likewise not so good at flying.
Man and machine must understand each other
During the ensuing discussion, it became clear that technology is only a partial problem confronting the digital revolution in the classroom, and it is obviously not the biggest. Another unsolved challenge is how to teach robots social behavior so that they can interact with humans in a meaningful way? How do they learn to understand and react to their human counterparts? Before AI will be able to make a useful contribution in the classroom, there must be mutual understanding: Can humans understand the machine, and are machines capable of understanding humans? And beyond this: How can machines learn human behavior? According to Stefan Kopp, the answer is: "By first observing us humans and learning to understand very precisely how humans interact socially".
Pedagogical concepts are missing
There was broad agreement on the podium that there is still a lack of convincing pedagogical concepts to make robots fully-fledged learning partners. Dr. Katharina Rohlfing, Professor of Psycholinguistics at the University of Paderborn, as a language learning researcher, in her own words "meddled" in the discussion, because AI researchers, in her opinion, understand too little about learning as such. Scarlet Siebert, a doctoral student at the Cologne University of Technology, also criticized that current AI research focuses mainly on the further development of technology and less on pedagogical concepts. Stefan Kopp pointed out that he does seek to involve educators in his research. For example, he asks teachers and kindergarten specialists about how to detect when children are becoming inattentive.
Children are a challenge for AI
Katharina Rohlfing criticized that social robots are today often still mere suppliers of information with predetermined, manifested concepts. Learning content is distributed over the children according to the "watering can" principle without taking their individual needs into account. She urged that if AI is to work with children, then it must adapt to the children with a multimodal approach, i.e. in language and gestures. From the 48-year-old's point of view, the social robots’ inability to recognize children's speech patterns is one of the major hurdles to the use of AI in the classroom. Children communicate in a different way than adults. Such deviations are still very difficult for an artificial intelligence to cope with. From a technical point of view, it is no longer a problem to simulate a way of speaking which is almost indistinguishable from human language, replied Stefan Kopp. However, as the computer scientist was quick to admit, correct speech recognition does indeed rely on the speaker using correct language."And children do articulate themselves differently. Sometimes, they just nod their heads."
Robotics as an economic factor
The Robokind Foundation from Hanover represents a fundamental approach. Dr. Jasmin Grischke, a member of its Board of Trustees, unreservedly affirmed the question "A robot for every child?" Her foundation wants to raise the level of awareness of robotics in general and create a broad acceptance of technology. The 37-year-old emphasized that the foundation was not only concerned with social robots, but also with robotic tools. New technologies should ultimately be used profitably, and effective robotics contributes towards bringing outsourced value-added chains back to Germany. "Should teaching robots one day become reality, we will welcome it with open arms," said Grischke.
Whether social robots will actually ever find their way into everyday teaching was something the Herrenhausen Forum was unable to clarify on this occasion. Many questions remained unanswered, leaving the more than 160 guests of the forum a little perplexed. Can children learn better with a robot? Yes and no. Robots are well suited for mathematical concepts, learning vocabulary and motoric tasks, said Stefan Kopp. A robot is never distracted, never impatient – on such points it is superior to a human tutor. Katharina Rohlfing outlined areas of application for children with delayed language development and learning difficulties, who need a lot of routine. However, there were different views on the fundamental question of whether a robot can be the better teacher.
Learning about learning
Where will we come across social robots ten years from now? Difficult to say. Katharina Rohlfing would like to see them occupying a permanent place in kindergartens, where technology is on the agenda, where exploring technology has become part of the day-to-day routine in order to generate a critical understanding of technology and the associated vocabulary. Stefan Kopp's research aims to make AI systems smarter socially, so that children can have fun with their new interaction partner and learn more as a result. Kopp believes that even in ten years' time we will not find AI being applied everywhere in the field of education: But by then basic research will have taught us a lot more about the process of learning.
The experts agreed that it was not a matter of artificial intelligence competing with human didactics, but rather of its integration into school processes. Katharina Rohlfing predicted that even ten years from now social robots will still not have become fully-fledged teachers – but rather toys, pals, peers, assistants, and perhaps tutors. Which of these exactly, or all of them at the same time – ten years from now, we will all be smarter.
Author: Bruno Brauer