For this promise to be fulfilled, the implementation of generative AI requires a critical understanding of both the benefits and drawbacks of personalized education.
"Personalized education" is a buzzword with many possible meanings: that students learn with 1:1 personal devices, or that the content they see on their reading apps adapts to their progress, or that they receive individualized feedback on essays.
For simplicity, one can think of personalized education as education that is uniquely tailored for one human student. With AI tutors, personalized education is both static and dynamic: Students are addressed by their first names, and they get content adapted to their progress. This creates a sense of uniqueness, which increases the student’s motivation and self-determination — psychological processes that are implicated in student engagement.
The benefits of personalized education are well-researched, particularly if students are given a choice in how their learning is personalized. That said, it would be wrong to assume that personalized education works for all types of learning. Nor is it a panacea for all the ills of public education. There are several limitations that educators, tech developers and decision-makers should pay attention to.
First, they should beware of claims that AI is positive because it offers personalized learning. Given the novelty of the technology, researchers do not have longitudinal or robust studies to confirm that. What they do have are some solid theories and previous research on personalized education, which they can use to extrapolate possible effects. A 2023 article in Educational Psychology Review says that for optimal learning, students need both personalized and non-personalized content. They need to be given some choices but not all the time. The careful balance of self-determined and automatically assigned content is essential for effective learning.
Smaller studies have also indicated possible effects with personalized AI. For example, a 2014 study in the Journal of Pragmatics found that the use of personalized books with young children is beneficial for increasing children’s talk during reading, but negative in encouraging self-centered talk. Other studies have shown that personalized education is not always the best approach for effective learning. Indeed, for holistic education, personalized learning needs to be balanced with what researchers call “pluralization” — shorthand for diversification and collective learning.
From a psychology perspective, personalized education pushes for individual rather than collective learning and collectivism. Individualism, the opposite of collectivism, is a focus on personal goals and is prevalent in Western and higher-income societies. Collectivism is the focus on groups and societies, particularly prevalent in Asia and Africa. The proponents of personalized AI tutors often forget that they push for a particular model of education associated with the Global North.
Then, consider this: Personalized education originated in private tutoring classes that could be afforded by selected members of society. For sure, AI tutors can increase access to such privileged personalized education. However, for the benefits of that access to be equitable, AI tutors need to be trained with diverse populations. Implementing them too early, with unknown or biased data sources, risks propagating social and economic biases from the past.
Then there is the issue of shared global citizenship of today’s generation. Although personalized education advances the learning of individuals, the challenges faced by children today, such as climate change and war conflict, will require collective action. In this respect, the current, individual-centered models of personalized learning fall short.
These issues are ethical but also moral: Current AI and children’s data protection frameworks are too broad and adult-centered to apply to personalized AI for children. For the technologies to respond without bias, they need to be trained on linguistically and culturally varied language models. Yet children’s rights and consent are not safeguarded with current legislation. Alarmingly, their data are used to improve the algorithms of a for-profit AI industry.
Families should play active roles in the development of such tools, but the technologies are being rolled out without analyzing the families’ perceptions. In one of my studies published in January by the peer-reviewed journal Learning, Media and Technology, parents expressed strong views against the presence of conversational AI agents such as Siri or Alexa in their children’s lives. They described the tools as propagating neoliberal values and damaging children’s social and emotional skills. Family concerns need to be listened to if AI technologies are to bridge, and not increase, the home-school learning gaps.
In sum, if AI is to transform education, it needs to be developed to maximize the benefits and minimize the risks of personalization. The questions that come with AI are questions that come on top of the social, ethical and moral questions of personalized education. These cannot be pushed aside if we are to truly advance learning for all children.
Natalia Ingebretsen Kucirkova is a professor of children’s development and author of The Future of the Self: Understanding Personalization in Childhood and Beyond.
