PUBLICATION 1

The rapid emergence of Generative Artificial Intelligence (GenAI) has created new possibilities for teaching and learning in higher education while simultaneously raising questions about student autonomy, motivation, cognitive engagement, academic integrity, and the development of self-regulated learning capabilities. Although a growing body of research indicates that GenAI can support students learning, existing evidence remains fragmented regarding the psychological and behavioural mechanisms through which AI-assisted learning translates into meaningful academic outcomes. This study proposes and empirically tests a structural model linking Generative Artificial Intelligence use, academic self-regulation, learning motivation, cognitive engagement, and academic outcomes among university students. Drawing primarily on Self-Regulated Learning Theory and Self-Determination Theory, the study conceptualizes GenAI-assisted learning not merely as a technological behaviour but as an educational process whose effectiveness depends on how students regulate, motivate, and cognitively engage with AI-supported learning activities. The proposed model hypothesizes that effective GenAI-assisted learning positively predicts academic self-regulation, learning motivation, and cognitive engagement; that academic self-regulation and learning motivation positively predict cognitive engagement; and that cognitive engagement contributes to academic outcomes. Mediation relationships are also proposed to determine whether self-regulation, motivation, and cognitive engagement explain the relationship between GenAI-assisted learning and academic outcomes. The study adopts a quantitative, cross-sectional design and proposes Structural Equation Modelling (SEM) for measurement and structural model evaluation. The manuscript contributes to the emerging literature by integrating technological, motivational, metacognitive, and learning-outcome perspectives into a single explanatory model. It also provides a framework for universities seeking to move from unrestricted AI adoption toward pedagogically responsible, human-centred AI-assisted learning.

Keywords: Generative Artificial Intelligence; GenAI; AI-assisted learning; academic self-regulation; learning motivation; cognitive engagement; academic outcomes; higher education; Structural Equation Modelling; university students.

INTRODUCTION

Generative Artificial Intelligence (GenAI) has rapidly moved from an emerging technological development to a significant component of contemporary educational practice. Large language models and related generative systems can produce explanations, summaries, examples, feedback, questions, outlines, translations, and other forms of learning support. Consequently, students can increasingly interact with AI systems as on-demand learning assistants rather than relying exclusively on conventional instructional resources.The educational implications of this development are substantial. Recent reviews indicate that GenAI can produce positive effects on several university learning outcomes, including academic achievement, language skills, affective motivational outcomes, and higher-order thinking. However, the evidence is not uniformly positive. For example, recent meta-analytic evidence indicates positive effects of GenAI on several learning outcomes while finding considerably weaker evidence regarding metacognition. (ScienceDirect 1) This distinction is important because educational value cannot be reduced to whether AI helps students complete tasks. A student may complete an assignment more efficiently with AI while simultaneously becoming less engaged in planning, monitoring, evaluating, or independently constructing knowledge.The emergence of GenAI therefore creates a fundamental question for higher education: Does AI-assisted learning improve students' learning because the technology itself is effective, or because it changes the way students regulate, motivate, and cognitively engage with learning?This question is particularly important because university learning requires substantial learner autonomy. Unlike highly structured school environments, higher education requires students to organize study time, establish goals, select learning strategies, monitor understanding, seek assistance, evaluate information, and persist when learning becomes difficult. Self-regulated learning (SRL) is consequently regarded as a central mechanism underlying successful university learning. Zimmerman conceptualizes self-regulated learners as active participants who regulate cognitive, motivational, and behavioural processes in pursuit of learning goals. (DOI 2)The relationship between self-regulation and academic achievement is well established. A meta-analysis involving 49 studies and 5,786 university students found that SRL interventions improved academic performance, SRL strategies, and motivational outcomes. (ScienceDirect 3) Similarly, meta-analytic evidence suggests that SRL activity partially mediates the relationship between SRL interventions and academic achievement, highlighting the importance of understanding the mechanisms through which self-regulation influences learning. (ScienceDirect 4)At the same time, motivation plays a fundamental role in determining whether students initiate, sustain, and invest effort in learning. Self-Determination Theory (SDT) proposes that motivation is influenced by the satisfaction of the psychological needs for autonomy, competence, and relatedness. (DOI 5) From this perspective, GenAI could potentially strengthen motivation by providing immediate feedback, personalized assistance, and opportunities for autonomous learning. Conversely, excessive dependence on AI could undermine students sense of competence or autonomy if learners become passive recipients of AI generated answers.Recent research reinforces the importance of examining this relationship. A 2025 systematic review of AI-supported self-regulated learning in higher education identified evidence that AI applications can support student autonomy and several SRL processes, but also emphasized the emerging and heterogeneous nature of the literature. (Nature 6) Another systematic review specifically examining GenAI and SRL identified affordances involving personalized learning objectives, resource searching, analysis, integration, problem solving, and feedback. (ScienceDirect 7)The present study therefore proposes an integrated structural model in which GenAI-assisted learning is connected to academic self-regulation, learning motivation, cognitive engagement, and academic outcomes. Rather than treating AI use as a simple predictor of academic achievement, the study conceptualizes AI-assisted learning as a multidimensional educational process whose consequences depend upon students' psychological and behavioural engagement with the technology.

BACKGROUND OF STUDY

The increasing availability of tools such as "Google Gemini" and other large language models has fundamentally altered the information environment in which university students learn. AI systems can provide immediate explanations and feedback, generate examples, simplify complex concepts, assist with brainstorming, and support writing and problem solving.The opportunities are substantial. Kasneci et al. (2023), for example, identified both opportunities and challenges associated with large language models in education, including personalized support, innovative educational applications, accessibility, accuracy concerns, ethical issues, and equitable access. (ScienceDirect)At the institutional level, UNESCO has similarly advocated a human-centred approach to GenAI in education, emphasizing ethical, safe, equitable, meaningful, and pedagogically appropriate use. (UNESCO)The central educational challenge is therefore not whether universities should acknowledge GenAI, but how students should use it in ways that preserve meaningful learning.AI-assisted learning can theoretically perform two contrasting functions.First, it can function as a Cognitive Scaffold. Students may use AI to obtain explanations, generate alternative perspectives, identify weaknesses in their understanding, obtain feedback, and formulate questions. In this scenario, GenAI supports rather than replaces student cognition.Second, it can function as a Cognitive Substitute. Students may use AI to produce answers without engaging with the underlying reasoning, resulting in task completion without corresponding knowledge development.These two possibilities suggest that the effect of GenAI cannot adequately be understood by measuring frequency of use alone. The quality, purpose, and regulatory context of use are likely to be more educationally meaningful than simple usage frequency.

PROBLEM STATEMENT

The rapid adoption of GenAI in higher education has outpaced the development of empirical models explaining how students' interactions with these technologies influence learning processes.Existing studies have frequently examined relationships between AI use and individual outcomes such as academic performance, engagement, critical thinking, motivation, or satisfaction. For example, a survey-based study of 353 university students using PLS-SEM reported positive associations between GEN-AI use and student engagement. (ScienceDirect) More broadly, recent experimental meta-analyses indicate that ChatGPT and GenAI interventions can improve academic performance, affective-motivational outcomes, and higher-order thinking. (ScienceDirect)However, three important problems remain.First, technology-use measures are often insufficiently differentiated. Knowing that a student uses AI does not reveal whether the student uses it to understand difficult concepts, generate ideas, obtain feedback, verify information, or simply obtain completed answers.Second, the psychological mechanisms underlying AI-assisted learning remain insufficiently integrated. Research on AI, motivation, self-regulation, engagement, and achievement has developed along partly separate lines. Recent systematic reviews have begun connecting AI with SRL, but the field remains relatively fragmented. (Nature)Third, the possibility that GenAI may produce learning gains without strengthening metacognitive processes requires further investigation. Recent meta-analytic evidence suggests positive effects of GenAI on several outcomes but no statistically significant effect on metacognition in the synthesized evidence. (ScienceDirect) This raises an important question: if AI makes learning easier, does it also make students better at regulating their own learning?The present study addresses this gap by proposing a structural model that examines how GenAI-assisted learning relates to academic self-regulation, learning motivation, cognitive engagement, and academic outcomes.

PURPOSE OF THE STUDY

The primary purpose of this study is to develop and empirically evaluate a structural equation model explaining the relationship between Generative Artificial Intelligence-assisted learning and university students' academic outcomes.

Specifically, the study seeks to determine: 1. Whether GenAI-assisted learning predicts academic self-regulation.2. Whether GenAI-assisted learning predicts learning motivation.3. Whether GenAI-assisted learning predicts cognitive engagement.4. Whether academic self-regulation predicts cognitive engagement.5. Whether learning motivation predicts cognitive engagement.6. Whether cognitive engagement predicts academic outcomes.7. Whether academic self-regulation mediates the relationship between GenAI-assisted learning and cognitive engagement.8. Whether learning motivation mediates the relationship between GenAI-assisted learning and cognitive engagement.9. Whether cognitive engagement mediates the relationship between GenAI-assisted learning and academic outcomes.10. Whether the combined pathway from GenAI-assisted learning through self-regulation, motivation, and cognitive engagement explains variation in academic outcomes.

SELF-REGULATED LEARNING THOERY

Self-Regulated Learning Theory provides the principal theoretical foundation for the study.Zimmerman describes self-regulated learning as a process in which learners actively regulate their cognition, motivation, behaviour, and environment in pursuit of learning goals. (DOI)

The model is commonly understood in terms of three broad processes:

1. Forethought – goal setting, planning, strategic decisions and motivational beliefs.2. Performance – implementation of strategies, monitoring, attention and effort regulation.3. Self-reflection – self-evaluation, attribution and adaptation.

GenAI can potentially interact with each phase. During forethought, students may use AI to clarify learning objectives, develop study plans, or identify prerequisite knowledge. During performance, students may use AI to obtain explanations, examples, feedback, questions, or alternative solutions. During self-reflection, students may use AI to review their understanding, identify errors, generate practice questions, or compare alternative explanations. However, these affordances only become educationally beneficial when students remain cognitively and metacognitively active.

SELF-DETERMINATION THOERY

Self-Determination Theory provides the motivational foundation of the model.SDT proposes that autonomous motivation is supported when individuals experience autonomy, competence, and relatedness. (DOI)In AI-assisted learning environments, GenAI may influence perceived autonomy because students can determine when, how, and for what purpose they seek assistance.AI may also support competence by providing explanations and feedback that enable students to overcome learning difficulties.Nevertheless, the relationship is not inherently positive. If students become dependent on AI-generated answers, the technology could reduce opportunities for independent problem solving and weaken perceived competence.Accordingly, the present study does not assume that GenAI automatically produces motivation. Instead, it proposes that the quality of AI-assisted learning is associated with learning motivation.

TECHNOLOGY-USE PERSPECTIVE

Technology acceptance research provides a complementary perspective for understanding why students adopt AI systems. The Unified Theory of Acceptance and Use of Technology identifies performance expectancy, effort expectancy, social influence, and facilitating conditions among the determinants of technology-use intention and behaviour. (AIS eLibrary)The present research does not position technology acceptance as the central theory because adoption alone does not constitute learning. Instead, technology-use considerations are incorporated into the definition of GenAI-assisted learning.

This distinction is important: Technology adoption asks whether students use AI; the present study asks what happens educationally when they use AI.

CONCEPTUAL FRAMEWORK

The proposed conceptual framework consists of five principal constructs: Generative AI-Assisted Learning (GAAL), Academic Self-Regulation (ASR), Cognitive Engagement (CE), and Academic Outcomes (AO)

At the same time: GAAL → Learning Motivation (LM) → CE