Mrs. Schmid, why do children need to know what AI is and how it works?
Nowadays, some children get hold of their parents' smartphone for the first time at the age of 2. And depending on which app they use, AI may also be involved. Thus, children can often operate these programs at an early age, but they don't understand the algorithms behind them. That is important, however, because understanding AI is part of what is known as "digital literacy" – depending on how well a person can handle digital media, they face significant advantages or disadvantages.
What are these advantages and disadvantages?
In certain countries, e.g., the U.S., there is already an elite that benefits from digitalization, while other parts of the population are manipulated by it – with regard to consumption, but also opinion-forming and all the way to voting behavior. Being able to see through phenomena such as algorithmic selection processes or information silos in social networks, somewhat protects a person against such manipulative influences. Teaching a basic understanding of how AI programs and algorithms work should start as early as kindergarten.
Some parents use digital media to keep their children quiet, others keep children away from them completely up to a certain age. In my opinion, neither of these approaches is very useful. Therefore, state education should systematically address the issue as early as possible.
Why so early?
For one, because children today, as "digital natives," are confronted with the digital world from birth. Computers and, increasingly, systems with AI components are therefore part of children's lives. But also, because this gives children from all social classes the chance to engage meaningfully with AI and digital media and become confident and creative users. Some parents use digital media to keep their children quiet, others keep children away from them completely up to a certain age. In my opinion, neither of these approaches is very useful. Therefore, state education should systematically address the issue as early as possible.
And how can we introduce preschool children to this topic in a meaningful way?
Before tackling the topic of AI, one should first teach them the essential concepts of digital representation and information processing with the help of algorithms. You can show children in a playful way how images or text are represented in the computer – for example, that an image consists of individual pixels with color information.
Similarly, one can use various examples to illustrate the principle according to which digital information processing works – input of information, processing by means of a computer program, and output of a result. The next step would then be: what if this program is an AI program, e.g., facial recognition on the smartphone, which is based on the analysis of such pixel images.
What is also important for children to understand is that all computer programs are invented and made by humans and did not just fall from the sky. Accordingly, the goal is not to discover what natural laws a particular phenomenon is based on, as in the natural sciences, but to understand how a program is built; to recognize that you could extend and improve it – and how.
... can help prevent dementia? Listen to our podcast about it!
... can drive your business forward? Data product expert David Hollembaek gives tips in this feature!
... helps healthcare professionals with the early initial diagnosis of acute myeloid leukemia? Read more in this article!
Can children at this age even understand something this abstract? And what concrete benefits does this knowledge bring them?
In early childhood pedagogy, the prevailing opinion is that, theoretically, you can explain anything to children – if you reduce it suitably in didactic terms. There are numerous successful examples of how basic mathematical concepts or initial scientific thinking can be taught in a way that is appropriate for children. In my opinion, conceptual knowledge is essential if children are to be able to use digital media creatively and confidently, and not just to use them as a kind of TV 2.0.
What tools are already available for teaching AI and machine learning at preschool age?
One example is the games in Haba Education's Digitial Starter series, including three AI games designed for preschool. In one game, there is a robot dog and gift packages with bones hidden underneath. The children are supposed to help the dog figure out which packages the bones are under. To do this, they check a few packages and infer the rule according to which the rest of the bones are hidden – under all the packages with blue dots on them, or only under those that don't have a bow? This approach – building more general assumptions from examples – is the basic principle on which machine learning is based.
In another game, children can train a neural network to distinguish dogs from cats. Since children at this age cannot yet do math, inputs are color-coded according to their importance. Calculation is also implemented using color codes. For example, a red value says that there is a cat in the picture and a green value says that it is another animal.
It is important that children learn that all computer programs are invented and made by humans and did not just fall from the sky. Accordingly, the goal is to understand how a program is built; to recognize that you could extend and improve it – and how.
To what extent does the knowledge conveyed in these games help children in their current and/or future everyday lives?
Children can build a bridge between such games and what happens in a digital device. And if children have more background knowledge about how AI systems work, they will, also in later life, be able to assess more realistically what AI systems can and cannot do.
The analog aspect of being able to pick up game pieces and turn over cards – would you say that this is also important when it comes to teaching something digital like AI?
Absolutely. There is a lot of talk about introducing tablet classes as early as elementary school, but findings from developmental psychology tend to argue against this. We humans are embedded in an analog world. The younger children are, the more important it is that complex cognitive concepts are grounded in their analog world of experience.
To ensure that children are not left alone in the digitalized world in which they are growing up, they need a targeted and pedagogically as well as didactically well thought-through introduction of concepts of computer science and AI as well as some use of computers. However, this should occur very moderately: After all, it would be absurd to let a child learn a foreign language before it has mastered its own native language – unless it grows up bilingual. In the same way, children should not be overburdened with complex concepts before they have the basics to grasp them.
"Art will never work without the genius of a human artist, whether it's AI art, carpet art or fresco art."
Click here to read the interview with designer Christoph Grünberger.
"When companies ask me what the workplace of the future will look like, I say that in ten years a company will be a kind of giant adventure playground."
Futurologist Lars Thomsen talks about AI and the working world of tomorrow in this interview.
"The industry still suffers from its dusty image of the developer sitting in the basement, shunning daylight and any interpersonal contact. This image has not corresponded to reality for a long time."
We talk to tech influencer Annahita Esmailzadeh about diversity in the tech industry – this way!
What should we generally keep in mind when we want to introduce children to AI?
I would always start with the basic concepts of digitalization and digital information processing before, on this foundation, introducing the first AI concepts. By and by, you can explain when AI methods are even needed – namely, when standard algorithms are not applicable. This is especially true for problems that cannot be completely described in formal terms. Such discussions, however, are certainly not for kindergarten, but rather for children in the 7th or 8th grade.
Many adults, also in the education system, do not know themselves what is behind the term AI. Wouldn’t they need some additional training first?
That's right. We definitely need proper training for educational professionals. Of course, we always have to take into account that they have studied neither mathematics nor physics or computer science. Accordingly, it is important that suitable material is available for all these topics, which can be used easily.
A good example are the Montessori materials: the educator Maria Montessori used them to bring mathematics into the kindergarten. Her materials for early number understanding are so well prepared that even people who have not studied this subject can use them to teach children mathematical principles. We need something similar for the early teaching of computer science skills. Unfortunately, it has long been neglected to include computer science in the early education of children and thus also in the training of educators. Until we have integrated it there, we will have to teach them by means of further education.
If children have more background knowledge about how AI systems work, they will, also in later life, be able to assess more realistically what AI systems can and cannot do.
Would you say introducing children to AI is the sole responsibility of the education system? What about, for example, companies that are driving the development of AI, such as Google, Amazon, etc.?
To my knowledge, most of these companies are pursuing approaches that go in the direction of AI education. Google, for example, offers a few interactive elements online that let you try out machine learning and train neural networks. However, while kids can play around with these programs and see what the network has learned and what it hasn't, the why – why did my network learn to tell dogs from cats but not VWs from Fords? – is not answered on these websites. Nor do they explain exactly how these learning approaches work in the first place. This, in turn, has to be explained differently to a 5-year-old child than to an 8- or 10-year-old, and so on. This component, for which one needs pedagogues and didacticians, is missing.
As we have already discussed, children come into contact with AI systems at an early age, and many seem to develop a negative image of them. For example, they assume that audio bots like Alexa are lying to them or trying to fool them.
This again shows how important it is to first teach essential basics to everyone. We humans readily attribute an intelligence similar to ours to a system that does something just a little bit like us. For example, we impute an intention to an Alexa – that she has the goal of lying to us. But an audio bot has no intentions. Alexa simply responds to speech patterns and seeks appropriate information.
People who don't understand what's behind AI technology tend either to expect far too much and then be disappointed, or to have fears that often miss the point.
Google offers interactive elements online that let you try out machine learning and train neural networks. However, the why – why did my network learn to tell dogs from cats but not VWs from Fords? – is not answered.
Speaking of fears: What's the right balance between pointing out "opportunities" and "risks" when talking to kids about AI?
I'm not sure how useful it is to talk about this at kindergarten age. But being open and yet critical of new technologies – this dichotomy is always needed.
A look into the future: What is your wish for the future teaching of AI to children?
The awareness that an early understanding of basic computer science concepts and AI methods is important is, I think, established. What I would like to see is targeted funding for projects in which computer scientists and educational scientists collaborate so that appropriate materials are available for training educational professionals as well as for imparting knowledge to children. In order to implement these solidly, however, we should rather give ourselves enough time than unleash unhelpful (because not well elaborated) material on our children.
Ute Schmid is head of the Department of Cognitive Systems at the University of Bamberg. The habilitated and doctoral computer science professor also holds a degree in psychology in addition to her diploma in computer science. Besides her work at the university, Schmid is involved in AI education. Among other things, she regularly gives workshops on AI and machine learning for teachers and students. In 2020, she received the Rainer Markgraf Prize for her educational work.