How does the impact of AI differ between children and adults? How does the ability to assess and evaluate AI differ between these age groups? It would also be interesting to see how this evaluation and assessment changes as children grow into teenagers.
Dear Connie Huo,
Thank you very much for your question.
First of all, I’d like to clarify a few terms: in most countries, we use the term “adults” to refer to people aged 18 and over until their death. This is a very wide age range and there are, of course, differences within it. My grandmother, for example, has never used an LLM. That is why I will limit my answer to people aged between 18 and 65.
Children and young people
We broadly define children as people aged between 6 and 13; young people are those aged between 13 and 18.
Simulation of human communication
The term ‘AI’ is also very broad. In my answer, I am referring to so-called Large Language Models (LLMs), which process text. These models are very interesting, but they have the disadvantage that they do not reveal the sources from which they gather their information (which is sometimes done in ways that raise copyright concerns). At the same time, they convey information in a way that is very similar to human communication. This is why users often attribute human characteristics to LLMs (the ELIZA effect). The problem with this is that, as a result, users often fail to question the reliability of the information, as Sunhyo Oh et al. 2025 have shown.
Assessment of AI responses
Children and adults differ in their cognitive development (e.g. risk assessment, abstract thinking, perspective-taking), but also in their knowledge. When it comes to AI, knowledge of the technology naturally plays a role. Because children have less knowledge of the technology behind LLMs, they are more likely to anthropomorphise them, as Jan Jakob Krüger et al. demonstrate in their 2025 study.
Meta-knowledge
Knowledge also plays a role in interpreting the answers. Those who have built up a great deal of knowledge on a topic are better able to contextualise an LLM’s response. Adults tend to have a greater prior knowledge base. Moreover, they possess more knowledge about knowledge itself. This so-called epistemological knowledge encompasses our understanding of how knowledge is created, how it is shaped by science, and how it can always contain errors. People usually only begin to develop this epistemological knowledge at secondary school level.
This is why children generally find it harder to evaluate the answers provided by LLMs than young people, and young people generally find it harder than adults. However, I have not yet found a suitable study on this.
Overall, however, it is important that we all learn to use LLMs effectively and empower people to think critically. This is what we are working on in learning research.
There are four further aspects that are important to me:
- Even within specific age groups, there are significant differences between individuals.
- LLMs are still fairly new, and we all need to keep learning about them (keyword: lifelong learning).
- Because this is such a new topic, there aren’t that many studies on it yet – particularly no longitudinal studies that track individuals over several years. This means that some findings may not be related to age at all, but rather to the cohort being surveyed.
- Many LLMs are only authorised for use from the age of 13. That’s why it’s difficult to research how children interact with LLMs.
I hope my answer has been helpful. I find the question fascinating, and if anything else occurs to me, I’ll write again.