Prominent educational researchers and organizations agree that education must be learning-centered for every pupil to reach their potential (Bransford et al., 2000; Groff, 2017; O’Brien et al., 2009; OECD, 2006; Vygotsky & Cole, 1978). The learning-centered paradigm stems from changes in industry and society: The predominantly knowledge-based form of work requires an education system that values individual differences and promotes lifelong learning skills (Lee, 2014). Since early experiments demonstrated the benefits of one-on-one tutoring over group-level instruction (Bloom, 1984; Salvin & Karweit, 1985), researchers he sought technological solutions to achieve cost-effective personalized tutoring (Lee et al., 2018). Personalized learning (PL) aims to achieve this goal by tailoring instruction, pace, methods, and content to the interests, needs, and goals of individual learners (Beese, 2019; US Department of Education, 2010; Walkington & Bernacki, 2020). Breakthroughs in technology and artificial intelligence (AI) he led to a rapid increase in the applications of PL (Chen et al., 2020). AI-driven adaptive learning systems (Corbett et al., 1997; Maghsudi et al., 2021; VanLehn, 2011) he emerged to provide learners with individualized lesson sequences, content recommendations, tasks, and automated assessments (Zawacki-Richter et al., 2019). A prevailing view in the application of AI to Education (AIEd) literature is that personalized adaptive learning systems increase access to high-quality education and are contrasted with a “traditional,” one-size-fits-all approach (Tetzlaff et al., 2020; Maghsudi et al., 2021; St-Hilaire et al., 2022). However, previous reviews he pointed to limited analysis of the educational underpinnings of AIEd (Chen et al., 2020; Hinojo-Lucena et al., 2019; Zawacki-Richter et al., 2019).
Here, we analyze the PL approach from the perspective of the educational goals set by the OECD Learning Compass 2030 as a framework to support our arguments (OECD, 2019b). In addition to core knowledge and skills, the framework highlights three main goals of education: focus on general competencies, developing learners’ agency, and building on the Anticipation-Action-Reflection (AAR) cycle. First, the AAR cycle reflects the need for deeper cognitive engagement before and during learning (Bransford et al., 2000; Chi & Wylie, 2014; Dehaene, 2020), also known as active learning. Second, agency is a broad and interdisciplinary term that denotes the skills necessary for lifelong learning (Brod et al., 2023; Council of the EU, 2002; OECD, 2006). An essential part of agency is related to self-directed and self-regulated learning (SRL) skills that enable the learner to take the initiative in leading their learning, monitor their level of understanding, setting goals, planning tasks, and taking action toward them, and evaluate their performance (Knowles, 1975; Zimmerman, 1986, 2000; Saks & Leijen, 2014; Panadero, 2017). It has been consistently shown that without proper SRL skills, learners overwhelmingly choose the least effective strategies for learning (McCabe, 2011). Third, we define general competencies as personal attributes or behiors developed through action, experience, and reflection (UNESCO, 2017). Examples of these competencies include critical thinking, collaboration skills, creating new value, and complex problem-solving skills (OECD, 2019b; UNESCO, 2022). We will expand and use these overarching goals in our analysis of the challenges of personalized learning systems.
We recognize the value of PL technologies as supportive tools for teachers and effective knowledge acquisition, particularly when they promote self-regulated learning, enhance learner’s agency, and engage students cognitively (e.g., Azevedo et al., 2022; Biswas et al., 2016; du Boulay, 2019; Long & Aleven, 2017). The benefits of AI-enabled PL systems are manifold, including flexibility in time and location, timely feedback, and faster student progression (Moreno-Guerrero et al., 2020; Pliakos et al., 2019), offering a more dynamic approach to knowledge acquisition compared to traditional methods. However, PL is described as an AI-based individualized learning approach and is often used synonymously with “adaptive learning” (Bernacki et al., 2021; Maghsudi et al., 2021; Shemshack et al., 2021; Shemshack & Spector, 2020; Tetzlaff et al., 2021). Rather than being the antithesis of the traditional one-size-fits-all approach, PL, as currently implemented, is simply a more effective method of knowledge acquisition. In contrast to the PL as currently implemented, the “ideal” personalized learning adheres to all principles of modern learning and aspires to a broader change in the school, educational strategy, curricula, effective pedagogies, etc., in which technology plays only a supporting role (Hopkins, 2010; Lee, 2014; Lee et al., 2018; Miliband, 2006; Watters, 2023).
We explore this perspective by examining how the PL literature emphasizes certain aspects of modern education while neglecting others. AIEd is still in its infancy, and there is an opportunity to guide its development in line with the principles of contemporary education (Shemshack et al., 2021; Cukurova et al., 2023). In the concluding sections, we explore the potential of large language models (such as ChatGPT) and propose a hybrid framework that envisions a collaborative relationship between AI and teachers.