Abstract
Over the past decade, artificial intelligence (AI) has undergone a tremendous shift from a highly technical instrument to a household learning, problem-solving, and decision-making cognitive aid. In the context of development, children, adolescents, and adults become increasingly exposed to AI systems that guide or provide feedback and information in real-time. Nevertheless, developmental psychology has failed to provide a theoretical framework concerning the role of AI in cognition regulation across the lifespan. Our conceptual analysis proposes that AI may be conceptualized not only as an external tool but also as a cognitive partner involved in the co-regulation of thinking, learning, and self-control. Using the concepts of executive function, metacognition, distributed cognition, and sociocultural development, we describe a developmental paradigm of human–AI co-regulation, with AI systems serving as scaffolds, metacognitive support, and external memory systems. We also comment on the benefits of conceptualizing AI as a mental partner in the form of offloading, the possible risks of cognitive offloading, and education and research implications.
1 Introduction
Developmental psychology has made significant progress over the past few decades in terms of elucidating the way cognitive capabilities arise and evolve throughout the lifespan. Studies of executive functioning, metacognition, language, social cognition, and learning have shown that cognitive development is influenced by not just biological maturation, but also interaction with environment, cultural tools and social partners (). Historically, these environmental factors have encompassed caregivers, teachers, peers, and symbolic systems like language and writing which are scaffolds to help in the progressive acquisition of self-regulated thought (). But recent advances in technology have brought a new category of interaction partners, one that is fundamentally different than previous ones to be thought of in developmental theory, and that is artificial intelligence (AI) systems that can offer real-time feedback, create information, and direct problem solving ().
Developmental psychology has long explained cognitive growth through stage-based and sociocultural frameworks. The stage theory proposed by Jean Piaget describes cognitive development as progressing through distinct stages, from sensorimotor interactions in infancy to formal operational thinking in adolescence and adulthood (; ). Complementing this, the sociocultural theory of Lev Vygotsky emphasizes the role of social interaction, scaffolding, and guided participation in the development of higher mental functions (). These theories collectively highlight that cognitive regulation emerges through structured interactions with external agents and gradually becomes internalized.
However, these foundational perspectives were developed in contexts where the primary sources of external regulation were human agents or static tools. The emergence of adaptive artificial intelligence introduces a qualitatively different form of interaction, where the external partner is capable of real-time response generation, feedback, and decision support. This raises the need to revisit classical developmental frameworks to examine how AI may participate in cognitive regulation across different stages of development.
During the past decade, AI-based tools like smart tutoring systems, chatbots, and large language models have become accessible to children, adolescents, and adults. These systems find their way more frequently in the educational facilities, professional activity and in daily life to help to write, reason, make decisions and learn. In contrast to the previous digital devices, modern AI systems do not simply store or present information but are involved in the mental process by proposing solutions, pointing to mistakes, and organizing work (). This has led to the fact that people are no longer controlling their thoughts by themselves, but more cognition is usually coming as a result of the continuous dialogue between human users and intelligent systems. This change poses significant theoretical issues to developmental psychology (; ; ). Assuming that cognitive development may have been viewed historically as the progressive internalization of socially mediated regulation, what is the developmental conceptualization of development in the presence of non-human agents capable of adapting, responding and directing behavior?
Theories that are available capture the implications of AI-mediated cognition in partial measures. Sociocultural theory focuses on the fact that greater mental functions are formed in the process of interaction and tend to be internalized by the individual. Equally, studies on distributed cognition maintain that thinking may be distributed among people, technologies and spaces. Executive function and metacognition practice have driven the examination of self-regulation further by proving that self-regulation is slowly formed with the help of outside support, feedback, and practice. All these views bring out the fact that cognition is not limited to the individual mind, but is influenced by interactions with external structures (; ). However, most of these theories have been created prior to the advent of adaptive AI systems that are able to produce new responses and engage actively in solution of problems. Thus, developmental psychology does not have a unified framework to explain how development of cognitive control, reflection, and learning through developmental continuity depends on the constant interaction with AI (; ; ).
Growing role of AI in the learning process has already started to create controversy. Other researchers note that AI-assisted learning might be effective in increasing cognitive performance through the provision of individual feedback and alleviating cognitive load. There are also concerns by other individuals that overdependence on AI will cause over cognitive offloading, less effortful thought, and poor development of self-regulation (). There are still only limited empirical findings, which are frequently inconsistent; this is partially due to the fact that the existing research has no clear theoretical definitions of how concept of AI participation is supposed to be. In most of the research, AI is regarded merely as a utility that provides the information and not as an active partner of the cognitive regulation (). These theoretical uncertainties complicate the ability to come up with specific hypotheses, compare the results of various studies, or assess the long-term developmental impact (; ).
Our recommendation is that the development of this direction needs a change of perception. Developmental science cannot avoid taking into account the fact that AI is a cognitive partner that can be involved in the regulation of thought and action instead of considering AI as an external aid. In the day-to-day learning scenario, people often refer to AI systems so as to organize work, assess concepts, create descriptions, and make choices. Such interactions are like guided participation that has been traditionally related to teachers or peers and they are conducted using technological systems that can respond in real time and continuously. Based on this perspective, cognitive activity is neither totally internal nor entirely external, but it is co-regulated by human and artificial agents. This perspective is in line with the previous theories that focus on the social source of cognition but expands them to a scenario where the partner who regulates is not a human being but an intelligent technological system (; ).
The aim of the current conceptual analysis is to come up with a theoretical framework of how this new kind of cognition can be understood. Based on the studies on executive functions, metacognition, distributed cognition, and sociocultural development, we hypothesize that the interaction with AI can be viewed as a co-regulation process, where cognitive regulation is distributed between the learner and the external intelligent system. Our thesis is that the framework can be used to explain the ongoing discussions regarding the advantages and threats of AI-assisted learning and give a framework on which future empirical studies can be conducted. Furthermore, we analyze how the use of AI as a cognitive partner could vary with the developmental stages, i.e., when a person is a child and when he is an adult and what it implies to education and psychological theory.
In what follows, we first describe the growing role of AI in learning and cognitive activity and outline the conceptual challenges it poses for developmental science. We then review theoretical perspectives on regulation, scaffolding, and distributed cognition that provide a foundation for understanding human–AI interaction. Building on these ideas, we propose a developmental framework of co-regulated cognition and discuss its implications for research on learning, self-control, and cognitive development in the age of artificial intelligence.
Importantly, the role of AI as a cognitive partner is unlikely to be uniform across all developmental stages. While early childhood cognition is primarily grounded in sensory interaction and human-guided regulation, AI-supported co-regulation becomes more relevant in later stages, particularly during middle childhood, adolescence, and adulthood, where metacognitive and self-regulatory capacities are more developed.
2 The rise of AI-supported cognition in development: a conceptual gap in developmental research
The adoption of artificial intelligence (AI) in the increasingly fast aspects of daily learning, reasoning, and decision making is one of the most conspicuous developments in the cognitive landscape of the 21st century. Children, teenagers, and adults now actively communicate with intelligent systems where the system answers, gives recommendations, and feedbacks in real-time. Intelligent tutoring systems help students work through academic assignments, recommendation systems help define the exposure of information, and conversational AI systems help write, plan, and solve problems (). What these technologies have in common with other older digital tools is that they do not passively provide information but rather take a direct role in the processes of thinking because they engage in response generation that can affect the way users think and behave (). Consequently, numerous cognitive processes in the past that were being done independently are now done through constant engagement with AI systems ().
Recent advancements in artificial intelligence have led to the emergence of learner-oriented AI environments specifically designed to support structured learning and cognitive development. For instance, features such as ChatGPT Study Mode, NotebookLM Learning Mode, and Gemini Guided Learning provide guided explanations, step-by-step problem solving, and interactive feedback tailored to learners' needs. Unlike earlier AI applications that primarily delivered information, these systems are explicitly designed to scaffold learning, promote reflection, and support metacognitive engagement.
These developments further reinforce the argument that AI is increasingly functioning not merely as an informational tool but as an active participant in cognitive processes. By structuring learning pathways, prompting reasoning, and adapting to user responses, such systems exemplify how AI can operate as a co-regulatory partner in learning environments.
Although the phenomenon of AI-mediated cognition has become increasingly widespread, the issue of developmental psychology is only starting to respond to it. Majority of the studies on cognitive development has still been based on the assumption that regulation of thinking is an internal process through self-control and executive functioning or external process by interaction with other human beings like parents, teachers, and peers (). Although these assumptions make sense in the previous technological circumstances, they do not give all the reality of the circumstances in which development takes place. Cognitive regulation becomes shared between human and technological agents when learners turn to AI systems to come up with ideas, cheat or have their answers checked, and arrange things. Theories which fail to specifically explain this type of interaction run the danger of missing an even more significant aspect of the developmental environment.
The initial studies conducted on technology and cognition have been usually concerned with the impact of computers and internet on attention, memory and learning. Research on the application of digital media, such as, has been conducted on whether exposure to screens impacts executive functionality, academic achievement or interpersonal development. The phenomenon of cognitive offloading, where people use relying on external devices to store or restore the information rather than retain it in memory, has been examined by other works (). These results are indicative that availability of technological aid can modify the allocation of cognitive resources, and in some cases, effortful processing may be reduced and efficiency enhanced. But the bulk of this research is based on a passive usage of technology instead of the interactive partner that can also influence the organization of the cognitive activity itself.
The advent of the modern AI systems means a qualitatively different case. Unlike the conventional tools, AI systems are able to adjust to user input, produce new responses, and give advice that is reminiscent of human guidance (). Intelligent tutoring systems can also be applied in teaching to track performance and set various tasks according to the optimal level of difficulty. By conversing with agents, it is possible to explain concepts, offer suggestions, and call to reflection, thus affecting the metacognitive processes. In particular, the large language models can generate long descriptions and troubleshooting steps that may be directly integrated by the user into his or her own thoughts (). The interactions make it less clear where internal and external regulation is concerned, as the system is involved in the organization of thought as opposed to merely providing information.
The existing development theories are not very useful in explaining this shift. Sociocultural theories lay stress on the fact that cognitive development is achieved by means of a directed involvement and progressive internalization of socially mediated regulation. In this sense, development is supported by the tools and symbols, which may help to organize activity and by allowing the new types of reasoning. These theories however were developed in environment where the guiding partner was a fellow human being. AI systems vary in significant aspects: they are able to react immediately, run constantly, and offer help without the social limitations that usually govern human communication (). Because of this, the developmental outcomes of AI-mediated regulation might not be consistent with the results pertaining to the old and new forms of scaffolding ().
The studies on distributed cognition also emphasize the fact that one can share thinking with other people and objects. Notebooks, calculators and computers according to this perspective can be incorporated in the cognitive system since they assist in memory, calculation or problem solving. Although the framework proves very advantageous in comprehension of the role of external aids, it fails in situations where the external system plays an active role in creating strategies or judging decisions (). Once the AI proposes the way one should solve a problem, the user is no longer merely using a tool but instead is having certain kind of interaction that is almost a collaboration. This brings the question of whether AI is to be conceived to be a component of the environment or a component of the cognitive system or a partner in the regulation of thought.
This lack of a definite conceptual framework has resulted in erratic understandings of the impacts of AI on learning and development (). In other studies, it is stated that AI-assisted teaching enhances performance with the help of personalized instructions and less redundant thinking (). Other literature has indicated that the use of automated support can undermine the acquisition of autonomous problem-solving abilities especially when learners blindly accept the answers provided by AI (). Due to a frequent discussion of these findings without a theoretical model, one cannot easily decide whether these findings are indicative of actual developmental changes or variations in the structure of cognitive activity in AI-mediated situations.
Another problem is that the utilization of AI is different at the different developmental stages. Young children can be exposed to AI as a guide just like a teacher, adolescents can utilize AI to explore and make decisions, and adults could use AI to handle complicated tasks (). These variations imply that the role of AI in cognitive regulation can shift down the line but the existing studies seldom take developmental paths into account. In the absence of a framework that would incorporate AI into current theories of self-regulation, metacognition, and learning, the field could be branded with the creation of fragmented findings that could not be easily compared or interpreted ().
These are the reasons why the growing role of AI in daily cognition is not only a technological shift but also a conceptual developmental challenge to developmental psychology (). To gain an appreciation of the dynamics of development in contexts that involve shared thinking with intelligent systems, there is need to redesign the conventional beliefs on the limits of the cognitive system and regulations (). Next, in the section below, we overview the theoretical views of regulation, scaffolding, and distributed cognition that gives us an opportunity to solve this issue and to create a framework according to which AI can be discussed as a partner in the process of co-regulation of learning and thinking.
3 The theoretical foundations of co-regulation and distributed cognition
To understand the impact of artificial intelligence in cognitive development, we need a theory that will provide an explanation of how thinking can be controlled both internally and externally with the help of external agents and tools. Higher mental functions have always been known to be developed by processes of guidance, scaffolding, and social involvement, but seldom have been projected onto the interactions with intelligent technological systems (). To frame the human interaction with AI in a developmental framework, three similar traditions of psychological theory can be tapped, such as sociocultural explanations of regulation, published work on executive and metacognitive control, and distributed cognition theories (). Collectively, these points of view show that cognitive action may be inflated among individuals and objects and, thus, one can think of AI as a companion in the process of controlling thoughts instead of a tool.
3.1 Regulation and the social origins of cognition
One of the main concepts of developmental psychology is that cognitive control is formed in the interaction with other people. The sociocultural theory suggests the emergence of higher mental process in social activity and then their internalization into individual capabilities. Children at the inception are dependent on caregivers to organize tasks, provide feedback and hold attention, but over time they learn to exercise such functions independently. This shift of externally directed behavior to self-regulation has been called a process of internalization, where social ways of control are internalized to the individual cognitive system (). The studies on executive functioning and self-control development have continuously proved that external assistance that aids young children in goal maintenance, distraction inhibition, and action organization is helpful ().
A key implication of this approach is that regulation is not necessarily found inside of the individual. Control can be shared between the guiding partner and the learner during development and this can form a kind of co-regulation whereby the responsibility of sustaining the task is distributed among the participants (). An example is if a teacher reminds a child of instructions or assists in organizing a problem, temporarily, they are offering regulatory assistance that the child is not yet able to maintain on her own. Repeated engagement in such interactions over a period of time enables the child to have a higher level of control and therefore, the development of a self-regulated behavior can be realized. This framework has historically been used in interpersonal interactions but the same rationale can be applied to scenarios where instructions are given by the technological systems that are in a position to adjust to the behavior of the learner.
3.2 Scaffolding and guided participation in learning
Very similar to the notion of co-regulation, there is the idea of scaffolding, which is used to explain the process of having partners who are more knowledgeable help with learning, by organizing tasks in a way that makes them manageable. Scaffolding is the simplification of complex tasks, the emphasis of the information, and the feedback that can assist the learner to be focused on the purpose. The more competent the learner, the less support is provided and the learner takes more responsibility of the task. This is a process that has been researched extensively on the educational setting and is regarded as one of the key mechanisms in the development of cognitive skills ().
There is a growing use of technology based forms of scaffolding in modern learning environments as opposed to actual human interaction. Smart tutoring systems, such as, have the ability to set the difficulties, hints and track progress in similar manner as teachers would. Developmentally, these systems can serve as scaffolds which assist learners to perform at a higher level that they are not able to do independently (). Nevertheless, in contrast to the classic type of scaffolding, AI-based support can be provided on a round-the-clock basis and be able to respond to the learner in real-time. This brings the idea that there is a chance that the line between the external direction and the internal control can be blurred when the regulation is shared with intelligent systems.
3.3 Executive function and metacognitive regulation
The studies of executive function and metacognition give additional information about the functioning of regulation in the developmental context. The executive functions such as working memory, inhibitory control and cognitive flexibility are those that help individuals to adhere to goals, avoid distractions and adapt behavior according to evolving situations (). These skills are built throughout the early years to adolescence under the strong influence of the environmental support and practice (). On the same note, metacognition is the ability to observe and manage one own thinking and it makes people able to strategize, assess progress and change strategies when there is need to do so. Beyond working memory and long-term memory, cognitive processing also involves episodic memory, semantic memory, and procedural memory. AI systems may support these memory systems by enabling contextual recall (episodic), conceptual understanding (semantic), and guided practice (procedural), thereby extending their role in cognitive support and offloading.
Notably, the executive and metacognitive processes may be supplemented with external aids. Written notes, reminders and instructions may ease the working memory load whereas the teacher or peer feedback may assist individuals to judge their performance (). Over the recent years, digital technologies have entered into these functions to a greater extent. Search engines are instant sources of knowledge, calculators are used to solve complicated tasks, AI systems are created to offer clarification or propose solutions. Such means can improve performance by weighting less on the cognitive load, yet these tools can also change the formation of self-regulation when people have to turn on the outside assistance rather than train the internal control (; ). To know this balance, the framework of regulation needs to be considered as a dynamic process that is shared between environment and the individual.
In addition to executive function and metacognition, cognitive development also involves perceptual processing and emotional regulation. Perception shapes how individuals interpret environmental inputs, while emotional processes influence attention, motivation, and decision making. AI systems, through adaptive feedback and interactive interfaces, may influence not only cognitive strategies but also learners' emotional engagement and perceptual framing of tasks. Incorporating these dimensions provides a more comprehensive understanding of AI-mediated cognition.
From an information-processing perspective, cognitive regulation involves both bottom-up and top-down processes. Bottom-up processing refers to data-driven responses based on sensory input, whereas top-down processing involves goal-directed control guided by prior knowledge, expectations, and executive functions. AI systems may influence both processes by structuring incoming information (bottom-up) and guiding planning, monitoring, and decision-making (top-down), thereby reinforcing their role in co-regulating cognition.
3.4 Distributed cognition and extended mind perspectives
The notion that the information processing can be extrapolated beyond the individual has been elaborated most clearly in extended mind and distributed cognition theories. Based on these strategies, the cognitive processes can be shared among individuals, tools, and external display that create a functional system (). Investigatively, an example of some calculation that a navigator would have to use a map and instruments to do would be calculations that would otherwise be hard to perform without writing down. The thinking system in such occasions consists of the person and the tools that contribute to thinking ().
These viewpoints indicate that the technological devices may become the components of cognitive process itself and not only affect it externally (). The majority of the early research on distributed cognition, however, was on comparatively simple tools to store or display information. The difference with artificial intelligence systems is that they are capable of creating new responses, considering alternatives and in decision making. In the case of a learner asking an AI system to plan an essay, solve a problem, or verify an answer, the activity of both the human and the machine is merged in performance (). This kind of interaction is better termed as collaboration rather than tool use since both the partners work toward regulation of the activity.
3.5 Toward a framework of human–AI co-regulation
Combined, sociocultural theory, studies of executive and metacognitive development, as well as descriptions of distributed cognitive all point to the same conclusion that cognitive activity is normally regulated by communication with external partners and tools. The question that is not clear is how these concepts can be used to scenarios where the external partner is an adaptive AI system that can actively engage in problem solving (; ). When AI systems offer instructions, organize work and make judgments, it can be said that they might act as colleagues in controlling cognition and not as auxiliaries.
In the next part, we continue to develop upon these theoretical backgrounds to suggest a developmental framework of human–AI co-regulation where artificial intelligence systems are viewed as cognitive partners who are jointly responsible in planning, monitoring, and controlling behavior. This framework enables us to look at the possible advantages of AI-enhanced learning and the dangers of over-dependence on external regulation, and gives us a foundation on which to incorporate technological change into the existing cognitive-development theories.
The proposed framework integrates multiple theoretical perspectives, including sociocultural theory, executive function, metacognitive regulation, and distributed cognition. By combining these perspectives, the model extends traditional developmental theories to account for adaptive technological agents that actively participate in cognitive regulation.
4 Artificial intelligence as a cognitive partner in development
The above considered theoretical approaches indicate that the cognitive activity usually arises due to the interplay between individual and external regulation sources. Within the conventional developmental environments, these sources are: caregivers, teachers and cultural tools that assist in organizing behavior until the learner gains enough control to be able to act independently. The increased availability of artificial intelligence in the daily learning experience presents an alternative interaction whereby regulation can be distributed with the adaptive technology (). In contrast to traditional tools, AI systems can make suggestions, assess responses and direct decision making in a manner that is similar to how humans are instructed. This is why it is not enough to conceptualize AI as a mere aid. Otherwise, it is possible to more effectively think of AI as a thinking partner, which is involved in the regulation of thought, learning, and problem solving ().
In this part we suggest that the encounter with AI can be viewed as the experience of human–AI co-regulations where the burden of keeping the goals, structuring information and measuring performance is shared between the person and the smart system. Such co-regulation can happen in various forms, which are determined by the role the AI has in the cognitive activity. Four roles in which AI can be involved in regulation are described below, namely as a scaffold, as a metacognitive support, as an external memory system, and as a decision partner. Figure 1 presents the proposed co-regulation framework of humans and AI, showing how the cognitive activity is co-regulated between the human cognitive system and artificial intelligence at different developmental stages resulting in co-regulated cognition, temporary cognitive performance, and long-term developmental effects.
Figure 1
The model highlights four primary roles of AI in cognitive regulation: scaffolding, metacognitive support, memory augmentation, and decision assistance. These interactions operate through both top-down and bottom-up processes, resulting in co-regulated cognition. The model distinguishes between short-term outcomes (enhanced task performance) and long-term developmental effects (changes in self-regulation and cognitive strategies). The nature and intensity of AI involvement vary across developmental stages, becoming more prominent in later stages where metacognitive capacities are more developed.
4.1 AI as a scaffold for cognitive performance
Among the most typical examples of how AI systems can impact cognition, there is the scaffold, which enables people to work at a level that is not yet within their independent capabilities. Intelligent tutoring systems have been used in education contexts to display a problem in a systematic order, give hints when there are errors and re-bias the difficulty of the tasks given to learners based on their performance (). These kinds of support are similar to scaffolding that has always been practiced by teachers whereby support is given to sustain progress in the area of the learner in the zone of proximal development.
Once AI is used to provide scaffolding, however, various significant differences are to be noted. To begin with, the technological support may be 24-h long, and students will be able to seek help any time as opposed to seeking it when a human instructor is in the classroom (). Second, AI systems can give real-time feedback, and the time gap between action and correction can be decreased. Third, the degree of support may be dynamically adjusted which may sometimes be invisible to the learner. Such features can contribute to higher efficiency at the same time as altering the conditions in which self-regulation is formed. In case they should be helped at all times, learners might not have so many chances to work on goal maintenance and problem-solving on their own. Consequently, the developmental impact of AI scaffolding can be conditional upon the frequency of external regulation and the availability of the chances to control independently ().
4.2 AI as a metacognitive support system
The second place of AI in cognitive activity is in the ability to regulate the metacognitive level of thinking. Metacognition is the skill to keep a track of personal knowledge and analyze the strategies and the behavior can be changed through mistake or doubt. Metacognitive assistance in the traditional learning setting can be offered by a teacher through questioning, offering strategies, or reflecting. More and more AI systems are being used to carry out these functions, suggesting that users should rethink their answers, come up with alternative explanations, or validate the correctness of their thought processes ().
In this respect, conversational AI systems are especially applicable, as they have the ability to give long explanations and answer follow-up questions (). The interaction between the learner and the AI system when the learner requests the AI system to describe something to the learner can lead to thoughts on the part of the learner, which are similar to those of a conversation with a knowing interlocutor (). In these situations, there is no complete internal regulation of the task, since the system plays a role in planning and assessment (). This type of interaction is capable of improving the learning process because it makes strategies explicit, but it may also decrease the necessity of the learner creating the strategies themselves. In case, the metacognitive control often gets transferred to some external system, the self-monitoring skills might be changed accordingly, particularly in the stages, when such skills are still being developed.
4.3 AI as an external memory and cognitive offloading system
The third role of AI in the cognitive activity is the storage and retrieval of information. Studies on cognitive offloading have indicated that in most cases people tend to use external devices in an effort to lessen the load imposed on the working memory. Cognitive resources, which are normally needed to recall or calculate, can be occupied by writing notes, calculating with calculators or searching information on the Internet (). AI systems take this process a step further because they do not only give the stored information but they also give the generated responses, which may replace the need to take an internal search.
As an illustration, when learners get their learners to create summaries, solve equations or create written text using AI, the same is an external part of the cognitive work (). According to distributed cognition view, the performance that ensues is an aggregate of the actions of the individual and the technological system. Although such cooperation may make a person more effective, it also creates a problem of how inner cognitive skills may be formed when the outside aid turns out that much dependable (). When memory, planning, and reasoning are often transferred to AI, the person can have less time to become good, practicing in the areas. The balance between the internal processing and external assistance may therefore determine its developmental outcomes.
4.4 AI as a decision partner in problem solving
In addition to scaffolding, metacognition, and memory support, AI systems may also be decision making partners. Most contemporary applications of AI will provide recommendations on what to do, what is the best option, or how to resolve an issue (). Recommendation systems manipulate decisions regarding information, entertainment, and educational content in addition to a conversational system suggesting answers to a school or even profession-related task. In such cases, the person is not simply making use of a tool but is involved in a sort of partnership where the AI is involved in the management of action.
Decision making has been more or less considered as an internal process which is controlled by the executive and experience. Nevertheless, some of the regulation processes become external to an individual when AI systems offer recommendations that are not widely assessed. This change can be positive and negative. On the one hand, it is possible to enhance performance and minimize mistakes with the help of correct guidance (). Conversely, overdependence on external advice can lead to the elimination of autonomy of judgment, especially when forming individuals who are yet to master the skill of appraising alternatives. To feel the impact of all these, a framework is needed where decision making will be perceived as a collective process and not an internal process ().
Combined, these roles demonstrate how AI can contribute to cognitive activity in various ways, each of which implies a co-regulation between the person and the technological system of a different nature (). By seeing these interactions as partnership and not the use of tools, developmental psychology can answer the questions that could not be answered in the framework of the traditional concept of self-regulated cognition. We elaborate this analysis in the following section and in an attempt to propose a developmental framework of how the human–AI co-regulation can evolve in childhood, adolescence and adulthood and explore its implications to learning and education.
5 Toward a stage-sensitive framework of human–AI co-regulation
For clarity, the developmental stages discussed in this framework are defined as follows: early childhood (approximately 3–6 years), middle childhood (7–11 years), adolescence (12–18 years), and adulthood (18 years and above). While exposure to AI may begin in early childhood, meaningful engagement with AI as a co-regulatory partner is more likely to emerge from middle childhood onwards, when cognitive and metacognitive capacities are sufficiently developed. Accordingly, the present framework focuses primarily on middle childhood, adolescence, and adulthood rather than attempting to model cognitive development across the entire lifespan.
It is important to note that the proposed framework is more directly applicable to later developmental stages, where individuals possess sufficient cognitive and metacognitive capacity to engage meaningfully with AI systems. In early childhood, cognitive development remains predominantly dependent on embodied interaction and human guidance, limiting the role of AI as a co-regulatory partner. Therefore, the present framework primarily conceptualizes AI as a cognitive partner in late childhood, adolescence, and adulthood.
Artificial intelligence can be involved in cognitive processes by scaffolding, metacognition, information storage, and decision making. These functions refer to the fact that such interaction with AI may be viewed as a kind of co-regulation whereby the power to think and act is divided between the person and an external intelligent system. Nevertheless, such interaction is unlikely to have similar developmental implications throughout the lifespan (). External regulation role varies with maturity of cognitive abilities, and hence the impacts of AI participation might vary with the level of development of a learner. In this part, we suggest a developmental model according to which human-intelligent co-regulation can be applied in childhood, adolescence, and adulthood, and we take into account both the potential advantages and the dangers of growing more dependent on intelligent systems.
5.1 Early childhood: AI as external regulator of behavior
Cognitive control is in its early developments, and in the case of a young learner, external directions may be relied upon to keep attention and follow directions and tasks. Studies of executive functioning have revealed that behaviors like working memory, inhibitory control, and cognitive flexibility are acquired over time and in preschool and early school and highly affected by environmental support in terms of development. The usual sources of this support are caregivers and teachers who organize tasks, remind children of their goals, and correct their mistakes thus allowing children to perform to the levels that they would not have been able to achieve on their own.
By introducing AI systems to this context of development, they can act as other external forms of regulation. Children-targeted educational applications can include stepwise instructions, instant feedback and simplified instructions to keep the course of the children (). AI may have a limited and emerging role in early childhood, primarily as structured external support rather than a fully developed co-regulatory partner. This kind of support can be useful as long as it enables children to enjoy success in difficult activities. Nonetheless, over-dependent on external advice can also restrict the possibilities to use self-regulation. When intelligent system always organizes things on behalf of the children, they would have less opportunities to learn how to sustain goals on their own (). Developmentally, AI can thus be more useful in early childhood when it has decreased support over the competence gain process rather than being a constant regulator.
5.2 Middle childhood and adolescence: AI as metacognitive partner
Middle childhood and adolescence represent the period of increasing ability of people to self-observation of their thoughts, and adjustment of strategy in the case of troubles. The period is characterized by a large scale enhancement of metacognition, planning as well as reflective reasoning (). Learners start assessing their self-performance, think of other solutions, and control effort according to the task requirements. Since these skills are still immature, mentorship by outside sources still remains significant.
AI systems employed at this phase are usually helpful on a metacognitive level. As an example, intelligent tutoring systems can provide strategies, point out mistakes or prompt the learner to rethink an answer. Conversationally-based AI systems have the ability to produce explanations, suggest outlines, or provide feedback on written literature, thus shaping the way learners think. The interactions can facilitate development by rendering cognitive strategies more transparent and by promoting thinking (). Meanwhile, the presence of ongoing support can decrease the necessity of the learners to invent strategies by their own approach. Teenagers using AI regularly to schedule or assess their tasks might be efficient in accomplishing them but might not train to the full extent to control cognition without outside assistance. This fact highlights the significance of the perception of AI as an instrument of enhancing performance and as an ally that defines the formation of self-control ().
5.3 Adulthood: AI as collaborator in complex cognition
During the adult stage, cognitive control is usually established, but this does not mean that people are always ready to cope with tasks that are beyond the scopes of internal processing (). Working in the field, conducting academic research or making simple choices often implies a significant amount of information that needs to be organized, assessed and utilized. In this case, cooperation with the external systems may increase the performance through the enlargement of the capacity of the individual (). The distributed cognition theories propose that tools and representations may be incorporated into the cognitive system where they can enable the reasoning process to be reliable and coordinated.
This is becoming a common use of AI systems, and they will act as partners that will aid in planning, writing, analyzing, and decision making. AI can give suggestions which alter the flow of thought, unlike in the previous tools, and the interaction between the two may seem as a cooperation instead of using a tool. In adults, this type of co-regulation can enhance efficiency without always decreasing internal capacities due to the fact that core executive and metacognitive skills are already in place. However, even the long-term reliance on automated assistance might still change the allocation of cognitive effort (). With the common practice of outsourcing complex reasoning to AI, people can get used to the idea of not having to maintain attention or tax their alternatives on their own. These effects can only be understood in a developmental framework that examines not just performance, but also involvement in processes by which cognition is controlled.
5.4 Risks of excessive cognitive offloading
Although AI-assisted co-regulation may lead to an increase in performance, it also creates the risk of over-offloading science. Cognitive offloading happens whereby people depend on outside tools to execute tasks that would have been executed internally. This has been happening with external memory aids, calculators and digital reminders but AI systems take it a step further, offering solutions to the problem instead of merely storing information (). When students blindly trust answers provided by AI, they deprive a cognitive process of its role in an individual.
Developmentally, offloading will affect the outcome of the offloading process based on the time and frequency of use. Forming stable knowledge and skills during early years of learning is frequently accompanied by effortful processing. When this effort is substituted with the outside aid, the development can be stifled (). However, in later stages, offloading can enable people to employ higher level thinking because it eliminates the burden of routine work. It is thus not the problem of developmental theory to ascertain whether AI is good or bad overall, but that of how external regulation promotes growth and when it takes its place.
5.5 Implications for education and developmental research
The fact that AI can be listed as a collaborator in cognitive managing has significant implications concerning education and research. The educational practices tend to take the idea that learning takes place mostly inside the individual and the tools are used merely to provide information. In case cognition is co-regulated with intelligent systems, then instruction techniques should be based on the interaction of external assistance and the establishment of self-control. To ensure the successful use of AI in education, it can be necessary to design systems that are supportive enough and allow the process of independent problem solving. As an example, smart tutors can provide less and less hints as the performance increases, or ask the learners to justify their answers before giving them.
In the case of developmental research, the idea of human–AI co-regulation raises the necessity of developing new ways of cognition study. Conventional experiments only tend to assess performance under conditions where the individuals are acting individually but not the case in real life scenarios of learning. Future studies are required to investigate the dynamics of cognitive processes in the presence of AI, and how they evolve as time goes by. Longitudinal studies can be especially relevant in finding out whether the early dependence on intelligent systems can determine the further self-regulation.
Taken together, these considerations suggest that AI does not simply add another tool to the developmental environment but alters the structure of cognitive activity itself. A developmental framework of human–AI co-regulation makes it possible to analyze these changes systematically and to evaluate both the opportunities and the challenges created by the increasing presence of intelligent technology. In the next section, we consider how adopting this perspective may contribute to a more coherent and theoretically grounded approach to studying cognition in the age of artificial intelligence.
A critical implication of this framework is the need to distinguish between domains in which AI can productively support cognition and those in which human autonomy must be preserved. AI systems are particularly effective in domains involving information processing, feedback provision, scaffolding, and routine problem-solving, where they can enhance efficiency and reduce cognitive load. However, domains such as critical thinking, ethical reasoning, decision-making under uncertainty, and the development of self-regulation require sustained human engagement.
From a developmental perspective, over-reliance on AI in these latter domains may limit opportunities for effortful processing and independent reasoning, which are essential for long-term cognitive growth. Therefore, the goal of AI integration should not be to replace human cognition but to complement it, ensuring that external support enhances rather than substitutes the development of core cognitive competencies.
6 Toward a more coherent science of development in the age of artificial intelligence
The framework put forward in this article implies that the increasing role of artificial intelligence in the day-to-day lives demands developmental psychology to rethink some of its fundamental assumptions regarding the manner in which cognition is controlled (). The conventional theories have tended to assume cognitive development as the process where the control is transferred by way of external control to self-management. Although the validity of this perspective has been demonstrated through numerous studies, it has been created in situations where the chief sources of external control were other individuals or comparatively easy tools. The creation of adaptive AI systems is placing a different scenario whereby regulation can be continued to be shared with external agents even after the self control has grown up. In case this possibility is not included in the theoretical models, development studies will risk missing a significant aspect of the contemporary cognitive environment ().
A difficulty to the field is that research on learning and cognition has continued to presuppose that people carry out tasks under conditions that no longer correspond to daily experience. People have a habit of using intelligent systems in most learning and working environments to seek advice, criticism, and knowledge. Consequently, there is a possibility that performance on tasks that are measured in the lab and in which external aid is forbidden, does not reflect the manner in which cognition will be applied in the real world. It does not imply that the use of traditional methods should be dropped but says that more methods should be used to analyze how the process of cognition will work when regulation is distributed to AI systems. In the absence of such research, it will be hard to understand whether the change in performance is a reflective indicator of the actual difference in ability or the difference in the structure of tasks.
The second problem is connected with the necessity of more conceptual definitions. As has been mentioned above, the current literature tends to view AI as a tool without defining its role in the cognitive activity (). This ambiguity renders comparisons between studies challenging; it is also hard to draw specific hypotheses on developmental outcomes. To give an example, one study interprets the higher performance with the help of AI as the indication of better learning, whereas another study interprets the same finding as the indicator of the less independent thought. Such interpretations need not be inconsistent in the case of performance as the result of co- regulation and not a product of the individual. It will thus be necessary to develop common definitions of such concepts as external regulation, cognitive offloading, and human–AI interaction to advance in this field ().
The third problem is the learning environment design. Educational technologies should endure evaluation based mainly on the efficiency, accuracy, or user satisfaction, and these are not necessarily the aspects, which measure long-term developmental impacts. Systems that offer ongoing support can lead to immediate performance improvements and fewer chances of self-regulation by the learners. On the contrary, systems that demand arduous interaction might seem less effective, yet might lead to better growth of cognitive control (). One developmental approach implies that the efficacy of the AI-based instruction should not be evaluated based on the short-term outcomes only. Rather, the impact of alternative types of assistance on the progressive process of the externally directed behavior into the autonomy of regulation is to be plotted.
Interaction between humans and AI during various time scales should therefore be studied in the future. The short-term research can determine the impact of AI on the performance when performing certain tasks, and longitudinal research is required to comprehend the impact of repeated interaction on the formation of executive functions, metacognition, and strategies of learning. This type of research can potentially show that the place of AI varies at different points in time, acting as a scaffold at the beginning of the learning process, as a partner in the middle, and a partner at a later stage of learning. To test these possibilities, it will be necessary to have methods in which the participants can make use of intelligent systems in a manner that is similar to real-world, as opposed to limiting the interaction to simplified experimental tasks.
The other key way forward as regards future work is on individual differences. AI is not used by all learners, and the developmental outcomes of external regulating can be dependent on age, background knowledge, motivation, and learning environment. There are individuals who might be satisfied with a lot of guidance and others who might depend on it to the level of undermining self-solving of problems. These differences will necessitate an integration of the developmental theory and the studies of motivation, self-regulated learning, and educational practice. Integration of this nature can assist in determining instances where AI assists in development and instances where it interferes with development.
Lastly, the growing presence of AI in cognition casts more general concerns on what intellectual development is. In case thinking can often be done in cooperation with smart systems, it can be no longer relevant to introduce cognitive ability as the one that can be done by people without help. Rather, development can be required to be perceived as the capability of managing internal activities to match external assistance in adaptable and productive means. Within this varied context, education does not merely aim at reinforcing independent reasoning, it also aims at enabling learners to utilize external resources in a dependable and critical manner. The theory of development that involves the co-regulation of human and AI can answer these questions without leaving the insights that previous studies have given.
Embracing such a framework does not imply that the old school of thought on development is wrong, but instead, it has to be expanded to explain fresh forms of interaction that the traditional theories did not consider at the time of their creation. The acceptance of artificial intelligence as a part of the regulatory process of cognition will enable developmental psychology to build more correct models of the process of learning and thinking in the modern context. This change can assist the discipline to have theoretical consistency and at the same time respond to the radical changes that were introduced by intelligent technology.
7 Conclusion
The traditional explanation of cognitive development in developmental psychology has been that cognitive development is a process of transition to the internally regulated thought form through externally determined behavior. The growing role of artificial intelligence in learning and daily problem solving is increasing this assumption by a scenario where regulation can be shared with external systems in spite of self-control developing. Cognitive activity is not the outcome of the individual but the result of interacting with the artificial and human agents as intelligent technologies can lead the focus, propose strategies, and create information.
Our conceptual analysis suggested in this study that AI can be understood as a cognitive partner that participates in the co-regulation of thinking and learning. Based on the sociocultural theory, studies of executive and metacognitive development, and descriptions of distributed cognition, we provided a framework in which AI can serve as a scaffold, a metacognitive support, an external memory system, and a decision partner. Such roles can help in improving performance but also affect self-regulations during childhood, adolescence, and adulthood.
By appreciating the human–AI interaction as being a co-regulated cognition, developmental psychology is able to cope with the changing circumstances in which learning is currently taking place. It might also be of some use in extending current theories to incorporate intelligent technological collaborators so that the field can continue to be theoretically clear and flexible to the conditions of the contemporary cognitive environment.
Statements
Author contributions
SP: Conceptualization, Writing – original draft, Writing – review & editing. JoJ: Conceptualization, Writing – original draft, Writing – review & editing. MJ: Conceptualization, Writing – original draft, Writing – review & editing. SA: Conceptualization, Writing – original draft, Writing – review & editing. NR: Conceptualization, Writing – original draft, Writing – review & editing. JeJ: Conceptualization, Writing – original draft, Writing – review & editing.
Funding
The author(s) declared that financial support was not received for this work and/or its publication.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was not used in the creation of this manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1
AbishevN.RamazanovR.AbaideldanovaM.ChesnokovaK.BaizhumayevaA. (2025). Artificial intelligence model in the cognitive and learning activities of university subjects. Front. Educ.10:1623170. doi: 10.3389/feduc.2025.1623170
2
AlmullaA. A.AlhaznawiA. A. (2025). The role of executive functioning interventions in supporting students with learning disabilities. J. Posthumanism, 5, 3991–4007. doi: 10.63332/joph.v5i5.1842
3
ArmitageK. L.GilbertS. J. (2025). The nature and development of cognitive offloading in children. Child Dev. Perspect.19, 108–115. doi: 10.1111/cdep.12532
4
AtchleyP.PannellH.WoffordK.HopkinsM.AtchleyR. A. (2024). Human and AI collaboration in the higher education environment: opportunities and concerns. Cogn. Res. Princ. Implic.9:20. doi: 10.1186/s41235-024-00547-9
5
AxelssonA.AnderssonR.GulzA. (2016). Scaffolding executive function capabilities via play-&-learn software for preschoolers. J. Educ. Psychol.108, 969–981. doi: 10.1037/edu0000099
6
AzevedoR.BouchetF.DuffyM.HarleyJ.TaubM.TrevorsG.et al. (2022). Lessons learned and future directions of metatutor: leveraging multichannel data to scaffold self-regulated learning with an intelligent tutoring system. Front. Psychol.13:813632. doi: 10.3389/fpsyg.2022.813632
7
BanihashemS. K.BondM.BergdahlN.KhosraviH.NorooziO. (2025). A systematic mapping review at the intersection of artificial intelligence and self-regulated learning. Int. J. Educ. Technol. High. Educ.22:50. doi: 10.1186/s41239-025-00548-8
8
ÇelaE.FonkamM. M.PotluriR. M. (2024). Risks of AI-assisted learning on student critical thinking: a case study of Albania. Int. J. Risk Conting. Manag.12, 1–19. doi: 10.4018/IJRCM.350185
9
ChardonnensS. (2025). Adapting educational practices for Generation Z: integrating metacognitive strategies and artificial intelligence. Front. Educ.10:1504726. doi: 10.3389/feduc.2025.1504726
10
ChenX.XieH.QinS. J.WangF. L.HouY. (2025). Artificial intelligence-supported student engagement research: text mining and systematic analysis. Eur. J. Educ.60:e70008. doi: 10.1111/ejed.70008
11
DeverD. A.WiedbuschM. D.RomeroS. M.AzevedoR. (2024). Investigating pedagogical agents' scaffolding of self-regulated learning in relation to learners' subgoals. Br. J. Educ. Technol.55, 1290–1308. doi: 10.1111/bjet.13432
12
DrigasA.MitseaE.SkianisC. (2023). Meta-learning: a nine-layer model based on metacognition and smart technologies. Sustainability15:1668. doi: 10.3390/su15021668
13
Elbadiansyah SainZ. H.Shehu LawalU.Chansa ThelmaC.Luqman AzizA. (2024). Exploring the role of artificial intelligence in enhancing student motivation and cognitive development in higher education. TechComp Innov. J. Comput. Sci. Technol.1, 59–67. doi: 10.70063/techcompinnovations.v1i2.47
14
ElshallA. S.BadirA. (2025). Balancing AI-assisted learning and traditional assessment: the FACT assessment in environmental data science education. Front. Educ.10:1596462. doi: 10.3389/feduc.2025.1596462
15
FollmerD. J.SperlingR. A. (2016). The mediating role of metacognition in the relationship between executive function and self-regulated learning. Br. J. Educ. Psychol.86, 559–575. doi: 10.1111/bjep.12123
16
GrinschglS.NeubauerA. C. (2022). Supporting Cognition with modern technology: distributed cognition today and in an ai-enhanced future. Front. Artif. Intell.5:908261. doi: 10.3389/frai.2022.908261
17
HeroldA.SeufertT. (2025). Who regulates? Effects of scaffolding in system- and self-regulated learning. Front. Psychol.16:1543024. doi: 10.3389/fpsyg.2025.1543024
18
HuX.LuoL.FlemingS. M. (2019). A role for metamemory in cognitive offloading. Cognition193:104012. doi: 10.1016/j.cognition.2019.104012
19
IqbalJ.HashmiZ. F.AsgharM. Z.AbidM. N. (2025). Generative AI tool use enhances academic achievement in sustainable education through shared metacognition and cognitive offloading among preservice teachers. Sci. Rep.15:16610. doi: 10.1038/s41598-025-01676-x
20
Koć-JanuchtaM. M.SchönbornK. J.RoehrigC.ChaudhriV. K.TibellL. A. E.HellerH. C. (2022). “Connecting concepts helps put main ideas together”: cognitive load and usability in learning biology with an AI-enriched textbook. Int. J. Educ. Technol. High. Educ.19:11. doi: 10.1186/s41239-021-00317-3
21
LeiY.LiuJ.FuX.ZhaoJ.YiB. (2025). The effects of a generative AI-enabled CDIO teaching model on undergraduates' computational thinking and individual psychological constructs. Comput. Appl. Eng. Educ.33:e70075. doi: 10.1002/cae.70075
22
LiM.WilsonJ. (2025). AI-integrated scaffolding to enhance agency and creativity in K-12 english language learners: a systematic review. Information16:519. doi: 10.3390/info16070519
23
LiuY.QiY.SunY. (2024). “AI learning system research on cognitive ability and academic achievement,” in Proceedings of the 2024 3rd International Conference on Artificial Intelligence and Education (Wuhan), 663–669. doi: 10.1145/3722237.3722353
24
LuB.WeiZ.LiX.YinY.LinghuJ.WangY.et al. (2025). Progress of a novel dentistry teaching model based on the combination of virtual reality and artificial intelligence technologies in optimizing cognitive load: a systematic review. J. Dent. Educ. 70025. doi: 10.1002/jdd.70025. [Epub ahead of print].
25
LubbeA.MaraisE.KrugerD. (2025). Cultivating independent thinkers: the triad of artificial intelligence, Bloom's taxonomy and critical thinking in assessment pedagogy. Educ. Inf. Technol.30, 17589–17622. doi: 10.1007/s10639-025-13476-x
26
MaY.ChenM. (2025). The human touch in AI: optimizing language learning through self-determination theory and teacher scaffolding. Front. Psychol.16:1568239. doi: 10.3389/fpsyg.2025.1568239
27
NeshaeiS. P.Mejia-DomenzainP.DavisR. L.KäserT. (2025). Metacognition meets AI: empowering reflective writing with large language models. Br. J. Educ. Technol.56, 1864–1896. doi: 10.1111/bjet.13601
28
NücklesM.RoelleJ.Glogger-FreyI.WaldeyerJ.RenklA. (2020). The self-regulation-view in writing-to-learn: using journal writing to optimize cognitive load in self-regulated learning. Educ. Psychol. Rev.32, 1089–1126. doi: 10.1007/s10648-020-09541-1
29
Pakdaman-SavojiA.NesbitJ.GajdamaschkoN. (2019). The conceptualisation of cognitive tools in learning and technology: a review. Australas. J. Educ. Technol.35, 1–24. doi: 10.14742/ajet.4704
30
PiagetJ.InhelderB. (2008). The Psychology of the Child. New York, NY: Basic books.
31
PozuelosJ. P.CombitaL. M.AbundisA.Paz-AlonsoP. M.ConejeroÁ.GuerraS.et al. (2019). Metacognitive scaffolding boosts cognitive and neural benefits following executive attention training in children. Dev. Sci.22:e12756. doi: 10.1111/desc.12756
32
ReubenJ. S.MeiriH.Arien-ZakayH. (2024). AI's pivotal impact on redefining stakeholder roles and their interactions in medical education and health care. Front. Digit. Health6:1458811. doi: 10.3389/fdgth.2024.1458811
33
RomeoG.ContiD. (2026). Exploring automation bias in human–AI collaboration: a review and implications for explainable AI. AI Soc.41, 259–278. doi: 10.1007/s00146-025-02422-7
34
RovetaA.CastelloL. M.MassarinoC.FranceseA.UgoF.MaconiA. (2025). Artificial intelligence in medical education: a narrative review on implementation, evaluation, and methodological challenges. AI6:227. doi: 10.3390/ai6090227
35
SidiropoulouE.MetallidouP. (2023). Metacognitive phenomena during human–Internet interactions. Psychology28, 69–82. doi: 10.12681/psy_hps.36228
36
SieglerR. S.EllisS. (1996). Piaget on childhood. Psychol. Sci.7, 211–215. doi: 10.1111/j.1467-9280.1996.tb00361.x
37
SongC. (2025). Harmonising minds: how AI-powered learning tools shape music education students' cognitive load, well-being and academic success. Eur. J. Educ.60:e70122. doi: 10.1111/ejed.70122
38
SteyversM.KumarA. (2024). Three challenges for AI-assisted decision-making. Perspect. Psychol. Sci.19, 722–734. doi: 10.1177/17456916231181102
39
SuJ. (2020). A rule-based self-regulated learning assistance scheme to facilitate personalized learning with adaptive scaffoldings: a case study for learning computer software. Comput. Appl. Eng. Educ.28, 536–555. doi: 10.1002/cae.22222
40
SunJ.WangY.HuZ.LiuY.SunY. (2025). Cognitive offloading bias in primary and secondary school students and its relationship with metacognitive monitoring. Br. J. Dev. Psychol.43, 645–659. doi: 10.1111/bjdp.12540
41
TaranikantiV.DavidsonC. J. (2023). Metacognition through an iterative anatomy AI chatbot: an innovative playing field for educating the future generation of medical students. Anatomia2, 271–281. doi: 10.3390/anatomia2030025
42
TwabuK. (2025). Enhancing the cognitive load theory and multimedia learning framework with AI insight. Discov. Educ.4:160. doi: 10.1007/s44217-025-00592-6
43
VygotskyL. S.ColeM. (1978). Mind in Society: Development of Higher Psychological Processes. Cambridge, MA: Harvard University Press.
44
WangB.GongJ. (2025). “Educational reform in the era of artificial intelligence: from autonomous learning to competition in calculation power,” in Proceedings of the 2nd Guangdong-Hong Kong-Macao Greater Bay Area International Conference on Digital Economy and Artificial Intelligence (Guangzhou), 1355–1360. doi: 10.1145/3745238.3745449
45
WangP.LiuT.YangY.XiangX. (2025). Optimizing self-regulated learning: a mixed-methods study on GAI's impact on undergraduate task strategies and metacognition. Br. J. Educ. Technol. 70018. doi: 10.1111/bjet.70018. [Epub ahead of print].
46
YanL.Pammer-SchindlerV.MillsC.NguyenA.GaševićD. (2025a). Beyond efficiency: empirical insights on generative AI's impact on cognition, metacognition and epistemic agency in learning. Br. J. Educ. Technol.56, 1675–1685. doi: 10.1111/bjet.70000
47
YanL.Suleman Abdullah AlwabelA.MohamadU. H. (2025b). AI-powered education: transforming teacher-student interactions and advancing sustainable learning practices. Eur. J. Educ.60:e70351. doi: 10.1111/ejed.70351
48
ZhangY.LucasM.Bem-hajaP.PedroL. (2025). AI versus human-generated voices and avatars: rethinking user engagement and cognitive load. Educ. Inf. Technol.30, 22547–22566. doi: 10.1007/s10639-025-13654-x
Summary
Keywords
AI-supported learning, artificial intelligence, cognitive development, co-regulation, distributed cognition, executive function, human–AI interaction, metacognition
Citation
S P, Joseph J, Jose M, S. M A, N R and Joseph J (2026) Artificial intelligence as a cognitive partner: a developmental framework for human–AI co-regulation in learning. Front. Dev. Psychol. 4:1835258. doi: 10.3389/fdpys.2026.1835258
Received
20 March 2026
Revised
15 April 2026
Accepted
15 April 2026
Published
29 April 2026
Volume
4 - 2026
Edited by
Stephanie M. Carlson, University of Minnesota Twin Cities, United States
Reviewed by
Poonam Sharma, Vellore Institute of Technology, India
Varghese Panthalookaran, Rajagiri School of Engineering and Technology, India
Updates
Copyright
© 2026 S, Joseph, Jose, S. M, N and Joseph.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Jeena Joseph, jeenajoseph005@gmail.com
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.