Sage Journals: Discover world-class research

Abstract

While Generative Artificial Intelligence (GenAI) technologies like ChatGPT are revolutionizing education by offering unique interaction opportunities and prompting legislative shifts toward AI literacy, there remains a significant gap in understanding the factors that influence their acceptance and effective use in educational settings. In particular, the impact of students’ digital competencies and motivational factors on their acceptance and utilization of GenAI tools is not well understood. This study investigates how these variables influence undergraduate students’ acceptance and use of ChatGPT as an informal learning tool, aiming to advance understanding of GenAI adoption in higher education. By integrating the Digital Competence Framework (DigComp) with the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2), this research provides a comprehensive analysis of factors affecting students’ intention to use and actual use of ChatGPT. Data were collected from 544 undergraduate students using adapted UTAUT2 scales and digital competence measures aligned with DigComp. Confirmatory Factor Analysis (n = 140) validated the adapted UTAUT2, and Structural Equation Modeling (n = 404) explored relationships between variables. The findings reveal that habit, hedonic motivation, performance expectancy, and facilitating conditions significantly predict students’ intention to use ChatGPT, while behavioral intention, problem-solving skills, and ethical considerations positively influence actual use. Notably, the study highlights the critical role of problem-solving and ethical awareness—in the adoption of GenAI tools. These results suggest that existing digital competence frameworks may need updating to include GenAI-specific competencies such as prompt-writing skills, managing ongoing dialogs with GenAI tools, and critically evaluating AI-generated content.

Plain Language Summary

How digital skills and motivation affect students’ use of AI for learning

This study looks at how university students decide to use ChatGPT, a type of artificial intelligence (AI), to help with learning. It aims to understand what skills and motivations make students more likely to use this tool. They surveyed 544 students and found that students are more likely to use ChatGPT if they find it enjoyable, believe it will help them, and if they are used to using it regularly. Students with good problem-solving skills and a sense of what’s right and wrong are more likely to actually use the tool. The study also suggests that the skills needed to work with AI, like writing good prompts and judging the quality of AI-generated information, should be taught more. This research helps educators and policymakers understand how to better prepare students to use AI in education.

Keywords

generative AI ChatGPT digital competence UTAUT2 AI in education

Introduction

Digital technology integration has driven significant shifts in higher education, enhancing both teaching and learning processes (Alenezi, 2023; Bond et al., 2020; Shrivastava & Shrivastava, 2022). These technologies facilitate personalized learning, flexibility, and improved outcomes, fostering autonomy and engagement that contribute to positive educational impacts (Pinto & Leite, 2020; Rybakova et al., 2021; Wekerle et al., 2020). However, in the effective use of these tools, students’ digital competencies and their willingness to adopt new technologies play significant role (Artacho et al., 2020; Bergdahl et al., 2020; Venkatesh et al., 2012; Vuorikari et al., 2022; Zhao et al., 2021).

Advancements in technology during the digital age have profoundly reshaped learning processes. The emergence of Generative Artificial Intelligence (GenAI) has accelerated this transformation by enabling the creation of high-quality instructional materials and providing personalized support, instant feedback, and adaptive learning experiences tailored to individual learner profiles and habits (Alasadi & Baiz, 2023; Boscardin et al., 2023; Mittal et al., 2024; Ouyang et al., 2023). Nielsen (2023) describes the interaction between humans and GenAI as the third paradigm in computing user interfaces, marking a shift where users specify what they want rather than how to achieve it, effectively reversing the locus of control. In this context, GenAI tools like ChatGPT have the potential to revolutionize education by transforming how individuals interact with information, acquire knowledge, and develop skills (Alasadi & Baiz, 2023).

Recognizing the critical importance of AI literacy, educational policies are increasingly advocating for the integration of AI education into curricula. Legislative initiatives across various regions are mandating AI literacy in schools, emphasizing the necessity for students to understand AI’s principles, applications, limitations, ethical considerations, and real-world impacts (GovTech, 2024; Kean, 2024; Lieberman, 2024). For instance, the California Chamber of Commerce has called for the state to teach students how to use GenAI tools effectively (GovTech, 2024). These developments underscore the growing recognition of AI’s pervasive role in society and the urgency for educational systems to prepare students to navigate and utilize AI technologies effectively.

GenAI represents a cutting-edge advancement that is significantly altering educational methods by simulating human-like creation and ideation (Alasadi & Baiz, 2023; Pavlik, 2023; Xu & Ouyang, 2022). Leveraging natural language processing, ChatGPT provides immersive, customized learning experiences with immediate feedback, thereby enriching engagement and comprehension (Kohnke et al., 2023; Maheshwari, 2024). ChatGPT exemplifies AI’s transformative impact, capable of supporting curriculum development and offering real-time feedback, fostering a more creative, and adaptable approach to learning (Bai et al., 2023; Jonsson & Tholander, 2022; OpenAI, 2024; Yan et al., 2024). However, as with other digital tools in education, recent studies suggest that adopting GenAI technologies requires advanced digital skills and a proactive disposition toward AI engagement (Kreinsen & Schulz, 2023; Yilmaz et al., 2023). Although, Artificial Intelligence (AI) technologies, particularly GenAI tools like ChatGPT, are reshaping higher education by enabling personalized and interactive learning experiences, successful adoption of these technologies requires a range of digital skills and motivational factors since students must be equipped to use GenAI effectively (Bergdahl et al., 2020; Venkatesh et al., 2012; Vuorikari et al., 2022). Without these skills and motivations, students may struggle to fully leverage GenAI’s potential, risking underutilization or even misuse (Abdelghani et al., 2023; Dickey et al., 2023).

In addition to structured, formal education, informal learning is becoming increasingly crucial for skill development in today’s fast-evolving technological landscape. GenAI tools, in particular, offer immense potential to transform informal learning by enabling personalized, self-directed educational experiences that go beyond traditional classroom boundaries (Touvron et al., 2023). By providing flexible, tailored access to knowledge, GenAI tools empower individuals to learn at their own pace and on their own schedule, adapting to unique learning goals and needs. This accessibility supports lifelong learning, making skill development achievable and sustainable for diverse learners (Laato et al., 2023; Peters & Romero, 2019). As technology rapidly advances, the need for continuous skill updating grows to meet societal and professional expectations—demands that formal education alone may not fully address due to its inherent rigidity (Nygren et al., 2019). GenAI thus opens new avenues for informal learning, offering learners adaptable resources to stay relevant and competent in an increasingly AI-driven world.

This study seeks to address the knowledge gap in understanding how digital competencies and motivational factors together influence the acceptance and use of GenAI in educational contexts, with a specific focus on informal learning through ChatGPT.

Theoretical Framework

The Digital Competence (DigComp) framework provides a comprehensive structure for understanding and developing digital competencies essential for participation in the digital world (Ferrari, 2013; Vuorikari et al., 2022). DigComp outlines five key areas of digital competence: Information and Data Literacy (IDL), Communication and Collaboration (CC), Digital Content Creation (DCC), Safety (S), and Problem-Solving (PS; Carretero et al., 2017; Vuorikari et al., 2022). These competencies are crucial for effectively engaging with digital technologies (Carretero et al., 2017; Vuorikari et al., 2016, 2022). In the realm of higher education, application of DigComp framework extends across various facets, from evaluating student capabilities to guiding curriculum development. For instance, it has been effectively used to create self-assessment tools tailored for university students, enabling them to gage and improve their digital skills in a structured manner (Liu, 2023). Furthermore, the framework has facilitated an understanding of the interplay between digital competence and socioeconomic factors in higher education settings, as explored by Evangelinos and Holley (2016). DigComp framework has inspired the integration of information and communication technologies and digital skills into pedagogical models, as evidenced by the work of Evangelinos and Holley (2016). This integration is key to preparing students for a digital world, ensuring that higher education remains relevant and responsive to the evolving demands of the digital age. While DigComp has been instrumental in higher education for assessing digital competence and guiding curriculum development (Evangelinos & Holley, 2016; Liu, 2023), it does not explicitly address the competencies required for GenAI tools like ChatGPT. This gap underscores the necessity to investigate how these competencies influence the acceptance and utilization of GenAI technologies, particularly in informal learning contexts—a key focus of this study.

In addition to digital competencies, motivational factors are critical to technology acceptance. As GenAI technologies like ChatGPT become more embedded in educational environments, beside understanding digital competencies, how motivational factors predict their acceptance, and utilization by students in learning processes becomes important. The Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) provides a comprehensive framework for understanding the factors that drive technology adoption by integrating various motivational theories and earlier adoption models. Built upon the Expectancy-Value Theory (Fishbein & Ajzen, 1977) and Theory of Reasoned Action (Ajzen & Fishbein, 1988), UTAUT2 extends these foundational models by incorporating social and behavioral factors that address users’ motivations to adopt new technologies (Venkatesh et al., 2012). Key constructs in UTAUT2, such as Performance Expectancy (PE) and Effort Expectancy (EE), directly correlate with perceived usefulness and ease of use, core elements in motivational theories, where perceived benefits, and manageable effort drive intention to adopt (Davis, 1985; Venkatesh & Davis, 2000). UTAUT2 introduces additional constructs, including Social Influence (SI) and Facilitating Conditions (FC), which recognize the impact of social pressures, normative beliefs, and supportive resources on technology acceptance, responding to critiques that previous models lacked explanatory power for social and environmental contexts (Taylor & Todd, 1995; Venkatesh et al., 2003). By incorporating Hedonic Motivation (HM) and Habit (HT), UTAUT2 also accounts for intrinsic motivations such as enjoyment and familiarity with technology, reflecting elements of Self-Determination Theory and Flow Theory, which highlight the role of pleasure, and routine in sustaining user engagement (Ryan & Deci, 2000). The model’s adaptability to diverse settings and its inclusion of intrinsic, extrinsic, and social motivations make it particularly well-suited for analyzing adoption patterns of GenAI tools in educational contexts. By applying UTAUT2 in this study, the aim is to capture a holistic understanding of both the digital competencies and the motivational factors that influence students’ acceptance and use of ChatGPT as an informal learning tool. This integrated approach allows to examine not only the skills students possess but also the motivational drivers that encourage them to adopt and utilize GenAI technologies. Table 1 provides a summary of literature review aligned with research gap and study aim.

Table 1.

Summary of Literature Review Aligned with Research Gap and Study Aim.

Thematic category	Key references	Key findings and relevance
1. Digital Technology Integrationin Higher Education	Alenezi (2023), Bond et al. (2020), Shrivastava and Shrivastava (2022), Rybakova et al. (2021), Pinto and Leite (2020), Wekerle et al. (2020)	- Digital technologies enhance educational practices and reshape higher education institutions.
		- Offer personalized learning experiences, autonomy, flexibility, and improved learning outcomes.
		- Enrich literacies and stimulate student engagement, leading to positive educational outcomes.
2. Importance of Digital Competenciesand Motivational Factors in Technology Adoption	Bergdahl et al. (2020), Vuorikari et al. (2022), Venkatesh et al. (2012), Artacho et al. (2020), Zhao et al. (2021)	- Digital competence is crucial for academic success and career development.
		- Students need sufficient digital skills to effectively interact with new technologies.
		- Users must be motivated to embrace new technologies for effective adoption.
		- Effective integration hinges on digital competencies and willingness to embrace new tools.
3. Importance of Informal Learning in the Contemporary World	Cross (2007), Livingstone (2001), Watkins et al. (2018)	- Informal learning offers accessible and flexible solutions tailored to individual needs and schedules.
		- Constraints of formal education necessitate alternative learning avenues.
		- Essential for lifelong learning and adapting to technological advancements and societal shifts.
4. Personalized and Flexible LearningEnabled by GenAI	Alasadi and Baiz (2023), Pavlik (2023), Xu and Ouyang (2022),Kohnke et al. (2023), Maheshwari (2024), OpenAI (2024),Jonsson and Tholander (2022), Bai et al. (2023)	- GenAI tools like ChatGPT provide interactive and customized learning experiences, offering immediate feedback and support.
		- Foster immersive and interactive learning environments with AI-driven personalized content, enhancing engagement and comprehension.
		- Enable education to be more adaptable and personalized, accommodating diverse needs and schedules of learners.
		- Support informal learning by integrating seamlessly into daily life and providing self-directed learning opportunities.
5. Need for Advanced Digital Skills and Positive Disposition toward GenAI	Kreinsen and Schulz (2023), Yilmaz et al. (2023), Kohnke et al. (2023)	- Adoption of GenAI in higher education requires advanced digital skills and willingness to engage with AI technologies.
		- Effectively interacting with GenAI tools has become paramount, emphasizing the necessity for advanced digital competencies.
		- Highlights the intertwined roles of skill and motivation in technology adoption.
6. Evolution and Integration of Digital Competence in the AI Era	Korzynski et al. (2023), Touvron et al. (2023), Martzoukou et al. (2021), Suwanroj et al. (2019), Kreinsen and Schulz (2023), Hockly (2023)	- Digital competence now includes understanding AI mechanisms and applications.
		- Shift from traditional interactions to complex engagements requiring strategic thinking for AI tools.
		- Integration of AI literacy into educational frameworks is essential.
		- Supports expanding digital competence frameworks to include AI-specific competencies.
		- Emphasizes the necessity of updating educational curricula to prepare students for an AI-pervasive future.
7. AI in Higher Education	Akgun and Greenhow (2021), Chen et al. (2020), Ifenthaler andSchumacher (2023)	- Educators leverage AI to tailor educational content, enhance interactive learning experiences, and provide data-driven insights.
		- Highlights the importance of understanding AI adoption from both educator and student perspectives.
		- Underlines the need for students to be prepared to effectively engage with AI tools in educational settings.
8. Research Gap and Aim of the Study	Current Study	- Limited research on the acceptance and utilization of GenAI tools like ChatGPT in informal learning settings.
8. Research Gap and Aim of the Study	Current Study	- Aim to investigate how digital competencies and motivational factors influence the acceptance and use of ChatGPT among higher education students.

Model Generation

In formulating the model for this study, the DigComp and UTAUT2 were integrated to provide a comprehensive understanding of the factors influencing the acceptance and use of GenAI technology ChatGPT as an informal learning tool among higher education students. This integration allowed examining not only the role of digital competencies, as outlined in DigComp, in facilitating effective engagement with ChatGPT but also the motivational factors influencing behavioral intention, and actual use, as captured by UTAUT2.

The DigComp framework provides a structured approach to understanding digital competencies essential for effective technology use (Carretero et al., 2017; Vuorikari et al., 2022). This study examines specific DigComp competencies—Information and Data Literacy (IDL), Communication and Collaboration (CC), Digital Content Creation (DCC), Safety (S), and Problem-Solving (PS)—as predictors of Actual Use (AU) of ChatGPT. Recognizing the importance of ethical considerations in digital interactions, Ethics (E) is included as an additional factor (Gümüş & Kukul, 2023). By focusing on these competencies, the study assesses how each area, including Ethics, impacts students’ effective use of ChatGPT in informal learning, directly informing the hypotheses.

Information and Data Literacy (IDL)

The ability to collect, manage, analyze, and interpret data (Vuorikari et al., 2022). This could predict the effective use of ChatGPT for data-driven tasks and inquiries. Students proficient in searching, evaluating, and critically analyzing information can leverage ChatGPT more effectively for research and learning. This component of digital competence ensures that students can ask relevant questions, interpret ChatGPT’s responses accurately, and integrate this information into their learning processes.

Communication and Collaboration (CC)

The capability to communicate, collaborate, and participate in digital networks (Vuorikari et al., 2022). ChatGPT can facilitate peer discussions, group projects, and interaction with instructors by providing a platform for instant information exchange, feedback, and support. Students with higher levels of digital competence are likely to use ChatGPT more effectively for collaborative learning activities.

Digital Content Creation (DCC)

The skills involved in creating digital content (Vuorikari et al., 2022). Users’ adept in this area may be more inclined to use ChatGPT for generating and enhancing educational materials. The ability to create digital content is a vital component of digital competence. Students can use ChatGPT to assist in drafting essays, reports, and presentations, thereby enhancing their digital content creation skills. Familiarity with digital tools and platforms enables students to integrate ChatGPT’s outputs creatively and responsibly into their work.

Problem-Solving (PS)

The competence to identify digital problems and creatively solve them (Vuorikari et al., 2022). This is likely to influence how students and educators leverage ChatGPT to tackle complex learning challenges. Digital competence includes the ability to engage in problem-solving and critical thinking in digital environments. Higher education students can use ChatGPT to explore solutions to complex problems, simulate scenarios, and generate ideas, all while critically evaluating the information and suggestions provided by the AI.

Safety (S)

The confidence and capability to protect oneself and others in digital environments (Vuorikari et al., 2022). It may influence the propensity to engage with ChatGPT if users feel secure in their interactions with AI. Understanding the importance of online safety, data protection, and privacy is part of being digitally competent. Students knowledgeable about these aspects are more likely to use ChatGPT and other digital tools while maintaining ethical standards and safeguarding personal and academic integrity.

Ethics (E)

This study acknowledges the intricate layers within the DigComp framework, especially the nuanced delineation between safety and ethics. Traditionally enveloped within the broader scope of safety, ethical considerations in digital environments encompass a wide array of moral decisions individuals face regarding content use, creation, and communication (Gümüş & Kukul, 2023). These considerations are paramount in fostering a digital culture that is not only secure but also respectful and equitable. Therefore, this study proposes a deliberate separation of the safety and ethics dimensions to scrutinize how higher education students navigate these digital terrains. Safety, as defined within the DigComp framework, focuses on protective measures against digital risks, emphasizing knowledge, and skills for risk management and privacy maintenance (Vuorikari et al., 2022). Conversely, ethics in digital interactions concern adhering to moral principles that govern behavior in digital spaces—emphasizing respect for copyright, privacy, and digital etiquette across diverse cultures and generations (K. Lin, 2016). This distinction is pivotal in our study, allowing for a focused investigation into the specific competencies critical for responsible digital citizenship in higher education. By examining these competencies, this study develops hypotheses on their direct impact on students’ actual use of ChatGPT, aiming to identify which are most critical for effective utilization in informal learning contexts.

UTAUT2 is utilized within this study for its comprehensive approach to understanding the factors influencing the acceptance and utilization of technological innovations (Venkatesh et al., 2012). UTAUT2 synthesizes elements from various established theories, providing a robust framework to assess the interplay of PE, EE, SI, FC, HM, and HT on behavioral intention (BI) and actual use (AU). However, the construct of Price Value (PV) have been excluded from the model. The exclusion of PV is a strategic decision that aligns with the nature of ChatGPT’s deployment in higher education, where the technology offers both free and paid versions. The complexity of assessing PV arises because the two versions cater to different user needs and financial thresholds, potentially skewing perceptions among users (OpenAI, 2024). Moreover, in the context of educational use, the value derived from ChatGPT is not solely based on financial cost but on the qualitative enhancement of the learning experience. Therefore, including PV might not accurately reflect the factors influencing students’ acceptance and use of ChatGPT in an informal learning context.

UTAUT2 framework is particularly appropriate for academic settings, as it has been empirically validated across contexts and is adaptable to the nuanced dynamics of higher education technology use (Al Farsi, 2023; Strzelecki, 2024).

The constructs of UTAUT2 are (Venkatesh et al., 2012):

Performance Expectancy (PE)

The degree to which individuals believe that using the technology will help them attain gains in performance. In the context of ChatGPT in higher education, this translates to the perception that using ChatGPT can enhance learning outcomes and educational productivity. This study posit that if students perceive ChatGPT as beneficial for their academic performance, they are more likely to intend to use it.

Effort Expectancy (EE)

This refers to the ease of use of the technology. For ChatGPT, it would be the user’s belief that this GenAI is user-friendly and easy to integrate into their learning processes. An intuitive and accessible interface reduces the effort required to use ChatGPT, thereby increasing students’ intention to adopt it.

Social Influence (SI)

The extent to which individuals perceive that important others believe they should use the technology. In higher education, this could be influenced by peers, instructors, or institutional adoption of ChatGPT. If students feel that their social circle supports the use of ChatGPT, they may be more inclined to use it themselves.

Facilitating Conditions (FC)

The belief that an individual has the necessary infrastructure and support to use the technology. For ChatGPT, this includes access to the technology, availability of support materials, and a supportive learning environment. Adequate resources and support facilitate the use of ChatGPT, encouraging students to adopt it in their learning activities.

Hedonic Motivation (HM)

The fun or pleasure derived from using the technology. ChatGPT could enhance the learning experience by making it more interactive and enjoyable for undergraduates. Enjoyment and intrinsic satisfaction from using ChatGPT may motivate students to adopt it for informal learning.

Habit (HT)

The extent to which people tend to perform behaviors automatically because of learning. In the case of ChatGPT, it’s how integrated the use of the AI becomes in the student’s or educator’s routine. When the use of ChatGPT has become habitual, students are more likely to intend to continue using it in their informal learning.

Behavioral Intention (BI)

In the context of this study, BI captures undergraduate students’ intentions to use ChatGPT for informal learning purposes. The determinants of BI in this study—PE, EE, SI, FC, HM, and HT—highlight the multifaceted nature of technology acceptance, where both practical benefits (like improved learning outcomes) and experiential aspects (such as enjoyment and ease of use) play crucial roles.

Actual Use (AU)

AU measures how frequently and extensively students integrate this AI tool into their informal learning processes. It reflects the transition from intending to use ChatGPT to incorporating it into study habits, coursework, research, and other educational activities.

The relationships among these constructs form a model where PE, EE, SI, FC, HM, and HT directly determine BI to use a technology, which in turn determines actual use behavior (Venkatesh et al., 2012). This study hypothesize that these motivational factors significantly influence students’ intention to use ChatGPT, and that BI influences AU.

Studies have highlighted ChatGPT’s potential as a supportive educational tool and the challenges associated with its adoption, including ethical concerns and the need for guidance in leveraging AI (Crawford et al., 2023; Gilson et al., 2023; Han, 2024; West, 2023). A combination of UTAUT2 and digital competence constructs can form a comprehensive model for understanding ChatGPT’s acceptance and use in higher education, ensuring ethical and effective use to support learning outcomes (Crawford et al., 2023; Gilson et al., 2023).

The aim of the current study is to empirically investigate how both digital competencies, as defined by the DigComp framework, and motivational factors, as outlined in UTAUT2, influence the acceptance and effective utilization of ChatGPT within higher education settings. By integrating constructs from both frameworks, it is aimed to provide a comprehensive understanding of the factors that affect students’ behavioral intention to use and actual use of ChatGPT as an informal learning tool. The outcomes of this research are anticipated to provide actionable guidance on curricular enhancements that can better align with the emerging requirements of a digitally competent and motivated academic community, capable of navigating and maximizing the potential of GenAI in education. By illuminating the interplay between technology acceptance and digital competence, this research contributes to the broader discourse on integrating GenAI tools into educational practices, ultimately enhancing educational outcomes.

Hypotheses Development

Based on the above constructs and their justifications, this study propose the following hypotheses:

UTAUT2 Hypotheses:

Hypothesis 1 (H1): PE positively influences BI to adopt ChatGPT as an informal learning tool within the higher education context. If students believe that ChatGPT will enhance their learning performance, they are more likely to intend to use it.

Hypothesis 2 (H2): EE positively influences BI to adopt ChatGPT as an informal learning tool. An easier-to-use system increases students’ intention to adopt the technology.

Hypothesis 3 (H3): SI positively influences BI to adopt ChatGPT as an informal learning tool. Perceptions of social pressure or encouragement from peers and instructors can motivate students to adopt ChatGPT.

Hypothesis 4 (H4): FC positively influences BI to adopt ChatGPT as an informal learning tool. Access to resources and support increases students’ intention to use ChatGPT.

Hypothesis 5 (H5): HM positively influences BI to adopt ChatGPT as an informal learning tool. Enjoyment and fun derived from using ChatGPT can enhance students’ intention to adopt it.

Hypothesis 6 (H6): HT positively influences BI to adopt ChatGPT as an informal learning tool. When students have formed a habit of using ChatGPT, they are more likely to intend to continue using it in their informal learning.

Hypothesis 7 (H7): BI positively influences AU of ChatGPT as an informal learning tool. Students with a strong intention to use ChatGPT are more likely to actually use it in their learning activities.

DigComp Framework Hypotheses:

Hypothesis 8 (H8): IDL positively influences AU of ChatGPT. Students with higher IDL skills can better utilize ChatGPT for accessing and processing information.

Hypothesis 9 (H9): CC positively influences AU of ChatGPT. Students proficient in CC may use ChatGPT more in collaborative learning activities.

Hypothesis 10 (H10): DCC positively influences AU of ChatGPT. Students skilled in DCC may more frequently use ChatGPT for creating and enhancing digital content.

Hypothesis 11 (H11): PS positively influences AU of ChatGPT. Students with strong PS skills may use ChatGPT to tackle complex problems, increasing their usage.

Hypothesis 12 (H12): S positively influences AU of ChatGPT. Students aware of digital safety may feel more comfortable using ChatGPT, leading to higher usage.

Hypothesis 13 (H13): E positively influences AU of ChatGPT. Students with high ethical awareness may use ChatGPT responsibly, affecting their usage patterns.

Method

Study Design

This study was approved by the Human Research Education Sciences Ethics Committee at [Erzincan Binali Yıldırım] University (Ethics Code: E-88012460-050.04-349157). The research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki. Participants provided informed consent, were assured of the confidentiality and anonymity of their responses, and were informed of their right to withdraw from the study at any time without penalty.

This study employed a quantitative research design consisting of two main components. The first phase involved conducting a Confirmatory Factor Analysis (CFA) of the adapted UTAUT2 model to validate its applicability in assessing the acceptance and use of ChatGPT as an informal learning tool in higher education. The second phase comprised the main study, which explored how digital competencies and the motivational factors that influence the acceptance and utilization of ChatGPT as an informal learning tool among higher education students. This was achieved by integrating constructs from both the UTAUT2 model and the DigComp framework.

Participants and Setting

Participants and Setting for Confirmatory Factor Analysis

The Confirmatory Factor Analysis (CFA) included 140 first-year undergraduate students enrolled in an introductory information technologies course during the 2023 to 2024 academic year. Participants were selected through convenience sampling due to their accessibility and relevance to the study, as they were already utilizing ChatGPT within the scope of their course. This sample was appropriate for validating the adapted UTAUT2 model, given the challenges of recruiting a sufficient number of ChatGPT users at the time of data collection. The sample size met the guideline proposed by Hair et al. (2010), which recommends a minimum of five observations per item for factor analysis; with the UTAUT2 containing 28 items, a sample of 140 ensured stable and reliable factor solutions. Data were collected via Google Forms to facilitate efficient and convenient participation.

Participants and Setting for the Main Study

The main study included 404 undergraduate students from diverse disciplines and institutions during the 2023 to 2024 academic year. Using separate samples for the CFA and main study ensured robustness and generalizability, reducing overfitting risks (Fokkema & Greiff, 2017; Kline, 2015). Practical considerations led to a specific course sample for the CFA, while the main study utilized broader representation. Data collection was conducted via Google Forms.

Data Collection Instruments

Instrument for Confirmatory Factor Analysis

The survey instrument for the CFA was the adapted version of the UTAUT2 framework, tailored to assess the adoption of ChatGPT as an informal learning tool in higher education. This adaptation involved revising the UTAUT2 constructs to align with the context of Generative AI (GenAI) in education. The instrument featured items on a 5-point Likert scale ranging from “Strongly Disagree” (1) to “Strongly Agree” (5), allowing participants to express the degree of their agreement with various statements about their behavioral intentions and actual use behavior concerning ChatGPT.

The UTAUT2 model utilized in this study (Venkatesh et al., 2012), assessed six key constructs that influence BI (3 items, ICR = 0.93) and eventually AU (3 items). These are; PE (4 items, ICR = 0.88), EE (4 items, ICR = 0.91), SI (3 items, ICR = 0.82), FC (4 items, ICR = 0.75), HM (3 items, ICR = 0.86), and HT (3 items, ICR = 0.82). The high Cronbach’s alpha values across all factors indicate excellent internal consistency, confirming the instrument’s reliability for assessing technology acceptance in diverse settings. Each construct was measured using multiple items, with the original model demonstrating strong psychometric properties, including high internal consistency reliability (ICR) with values ranging from 0.75 to 0.93.

Instruments for the Main Study

The main study applied the validated UTAUT2 model to assess higher education students’ acceptance and usage of ChatGPT, focusing on the diverse factors that influence its adoption in informal learning settings. Concurrently, the study utilized the Teacher’s Digital Competence Scale, developed by Gümüş and Kukul (2023), to assess a broad spectrum of digital competencies essential for navigating the contemporary educational digital landscape. The Teacher’s Digital Competence Scale comprises 46 items distributed across six factors: IDL (9 items, Cronbach’s α = .92), CC (7 items, Cronbach’s α = .95), DCC (6 items, Cronbach’s α = .93), PS (9 items, Cronbach’s α = .95), S (10 items, Cronbach’s α = .96), and E (5 items, Cronbach’s α = .91). Responses were measured on a 5-point Likert scale ranging from “Strongly Disagree” (1) to “Strongly Agree” (5). The instrument explained 71.97% of the total variance in digital competence, with each dimension demonstrating high reliability and strong internal consistency.

The Teacher’s Digital Competence Scale was chosen despite its primary focus on teachers for two reasons: the participants were pedagogy students preparing for teaching, and the scale aligns closely with the DigComp 2.1 framework, ensuring relevance to current digital standards. Its cultural sensitivity and focus on ethics made it well-suited for evaluating digital competencies. To increase relevance, “colleagues” was changed to “friends” to better reflect student interactions, improving its applicability and accuracy in capturing competencies related to peer engagement in higher education.

Data Analysis

Data Analysis for CFA

The validation of the adapted UTAUT2 model used CFA conducted in R software, starting with sample size verification to meet the recommended ratio of five subjects per item (Hair et al., 2010). Outliers were flagged using a ± 3 threshold (Osborne & Overbay, 2004), and the data was assessed for multivariate normality following Cain et al. (2017) to ensure robust statistical results.

The CFA assessed the model fit by examining various fit indices. The chi-square to degrees of freedom ratio (χ²/df) provided an initial indication of model fit, with lower values generally indicating a better fit acceptable (Hu & Bentler, 1999). The Root Mean Square Error of Approximation (RMSEA) gaged the model’s approximation to the data, with values less than 0.05 indicating an excellent fit and those up to 0.08 representing an acceptable fit, as suggested by T. A. Brown (2015). The Standardized Root Mean Square Residual (SRMR) was used as another measure of goodness of fit, with values less than 0.08 considered acceptable (Hu & Bentler, 1999).

The Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) were also assessed, with values above 0.90 indicating acceptable fit and values at or above 0.95 indicating very good fit (Hooper et al., 2008). Factor loadings were evaluated for significance, with a threshold of 0.40 based on Stevens (2012) guidelines, to confirm the strength of each item’s relationship to its construct. This CFA process rigorously validated the adapted UTAUT2 model, confirming its suitability for assessing ChatGPT acceptance and use as an informal learning tool in higher education.

Data Analysis for the Main Study

For the main study, data analysis was conducted in R, beginning with checks on data structure, sample adequacy, normality, and outliers to confirm suitability for statistical tests (Hair et al., 2010). Structural Equation Modeling (SEM) was used to test hypothesized relationships between digital competence constructs and factors influencing ChatGPT acceptance. This SEM approach simultaneously analyzed both the measurement model and the structural model, examining interrelations between constructs.

Model fit was evaluated using indices like the CFI and TLI, with values above 0.90 indicating an acceptable fit (Hooper et al., 2008). The RMSEA and SRMR, with values below 0.08, were also used to confirm fit (T. A. Brown, 2015; Hu & Bentler, 1999). Path coefficients were evaluated with bootstrap confidence intervals, highlighting the strength and direction of relationships. This thorough analysis provided insights into how digital competencies influence the adoption and integration of GenAI tools like ChatGPT in higher education. Table 2 summarizes the research process.

Table 2.

Summary of the Research Process.

Step	Description
1. Ethical Approval	Obtained ethical approval from the Human Research Education Sciences Ethics Committee
2. Phase 1: CFA of Adapted UTAUT2 Model
2.1 UTAUT2 Adaptation	Adapted the UTAUT2 questionnaire to assess ChatGPT acceptance in higher education.
2.2 Participant Selection for CFA	Selected 140 first-year undergraduate students from an introductory information technologies course using convenience sampling.
2.3 Data Collection for CFA	Collected data via Google Forms using a 5-point Likert scale.
2.4 Data Analysis for CFA	Performed CFA using R software; assessed model fit using standard indices; validated the measurement model.
3. Phase 2: Main Study
3.1 Instrumentation for Main Study	Used validated UTAUT2 questionnaire and Teacher’s Digital Competence Scale (modified for students).
3.2 Participant Selection for Main Study	Recruited 404 undergraduate students from various disciplines and academic levels; students in pedagogy-related courses.
3.3 Data Collection for Main Study	Collected data via Google Forms using a 5-point Likert scales.
3.4 Data Analysis for Main Study	Conducted Structural Equation Modeling (SEM) using R software; assessed model fit; evaluated relationships between constructs.
4. Results Interpretation and Reporting	Interpreted CFA and SEM results; identified significant predictors; discussed implications; provided conclusions and recommendations.

Results

Confirmatory Factor Analysis: Validating the Adapted UTAUT2 Model for Instructional Acceptance and Utilization of ChatGPT

Skewness and kurtosis values for all variables fell within the acceptable range of −2 to +2 (Hair et al., 2010), indicating no violation of normality. Outliers, identified using a ± 3 z-score threshold, were generally within acceptable limits, and did not significantly distort the analysis, which is typical in real-world educational data.

An initial CFA revealed that item EE3 loaded significantly on both Performance Expectancy and Effort Expectancy (0.301 and 0.532, respectively), violating discriminant validity. To maintain construct integrity, item EE3 was removed based on modification indices and CFA principles, which emphasize each item’s unique contribution to its construct.

The main CFA analysis completed successfully after 67 iterations using the Maximum Likelihood (ML) estimation method with the NLMINB optimization method. The model evaluated 83 parameters with a sample size of 140 observations.

The confirmatory factor analysis, conducted using lavaan 0.6.17, resulted in a significant chi-square statistic (χ² = 520.331, df = 295, p < .001), which is common with larger sample sizes due to the test’s sensitivity (Kline, 2015). However, alternative fit indices indicated an acceptable model fit: CFI = 0.920, TLI = 0.904, Robust CFI = 0.932, and Robust TLI = 0.919. The RMSEA was 0.074, with a 90% confidence interval ranging from 0.063 to 0.084, and the SRMR was 0.070. These indices suggest that the model provides a satisfactory fit to the data (T. A. Brown, 2015; Hooper et al., 2008; Hu & Bentler, 1999).

The CFA results demonstrated that all constructs in the adapted UTAUT2 model exhibited strong and significant factor loadings, confirming that the items effectively measure their intended latent variables (see Table 3). The standardized loadings for the indicators ranged from 0.513 to 0.970, all exceeding the acceptable threshold of 0.40 (Stevens, 2012), indicating robust associations between the observed variables and their respective constructs. Notably, HM showed exceptionally high standardized loadings, ranging from 0.867 to 0.970, suggesting a very strong relationship between the indicators and the latent construct. SI and PE also exhibited high loadings. Overall, all constructs had loadings well above the minimum acceptable level, reinforcing the reliability of the measurement model. The model’s fit indices suggest that while the fit is not perfect, it is acceptable and indicates that the hypothesized relationships are present in the data.

Table 3.

Standardized Estimates for the Adapted UTAUT2 for the Instructional Acceptance and Use of ChatGPT.

Construct	Indicator	Standardized loading
Performance Expectancy	PE1	0.834
	PE2	0.888
	PE3	0.713
	PE4	0.824
Effort Expectancy	EE1	0.794
	EE2	0.724
	EE4	0.760
Social Influence	SI1	0.916
	SI2	0.898
	SI3	0.817
Facilitating Conditions	FC1	0.806
	FC2	0.768
	FC3	0.758
	FC4	0.513
Hedonic Motivation	HM1	0.905
	HM2	0.970
	HM3	0.867
Habit	HA1	0.785
	HA2	0.768
	HA3	0.705
	HA4	0.840
Behavioral Intention	BI1	0.731
	BI2	0.776
	BI3	0.785
Actual Use	AU1	0.881
	AU2	0.887
	AU3	0.822

Measurement Model Validation in the Main Study

Following the initial validation of the adapted UTAUT2 model in the initial sample, a second CFA was conducted on the larger main study sample to validate the measurement model of all constructs prior to SEM analysis. The CFA using the Maximum Likelihood estimation with Robust standard errors (MLR) estimator provided robust measures for evaluating the model fit and parameter estimates for a dataset analyzed in R with the lavaan package. Here is a summary of the CFA results formatted according to APA guidelines:

A confirmatory factor analysis was conducted using the MLR estimator to assess the fit of the specified model to the data. The analysis involved 404 observations with 248 model parameters. The chi-square test of model fit was significant, χ²(2,453) = 3925.584, p < .001, indicating that the model did not perfectly fit the data. However, given the sensitivity of the chi-square test to sample size, additional fit indices were considered. The CFI was 0.930, and the TLI was 0.925, both of which are above the commonly accepted threshold of 0.90, suggesting a good fit of the model to the data. The RMSEA was 0.043, with a 90% confidence interval [0.040, 0.045], and the SRMR was 0.047, both indicating a good fit.

In terms of parameter estimates, all factor loadings were significant (p < .001), demonstrating that each item significantly contributed to its corresponding latent variable. The standardized factor loadings ranged from 0.645 to 0.943 for the different constructs, suggesting strong associations between the items and their respective factors (See Table 4).

Table 4.

Standardized Estimates for the Measurement Model.

Construct	Indicator	Standardizedloading
Performance Expectancy	PE1	0.791
	PE2	0.895
	PE3	0.906
	PE4	0.903
Effort Expectancy	EE1	0.749
	EE2	0.793
	EE4	0.855
Social Influence	SI1	0.876
	SI2	0.905
	SI3	0.897
Facilitating Conditions	FC1	0.838
	FC2	0.862
	FC3	0.843
	FC4	0.676
Hedonic Motivation	HM1	0.921
	HM2	0.942
	HM3	0.919
Habit	HT1	0.806
	HT2	0.910
	HT3	0.916
	HT4	0.885
Behavioral Intention	BI1	0.789
	BI2	0.850
	BI3	0.882
Actual Use	AU1	0.943
	AU2	0.932
	AU3	0.897
Information and Data Literacy	IDL1	0.645
	IDL2	0.746
	IDL3	0.790
	IDL4	0.808
	IDL5	0.785
	IDL6	0.725
	IDL7	0.763
	IDL8	0.710
	IDL9	0.720
Communication and Collaboration	CC1	0.732
	CC2	0.763
	CC3	0.755
	CC4	0.818
	CC5	0.823
	CC6	0.768
	CC7	0.815
Digital Content Creation	DCC1	0.782
	DCC2	0.876
	DCC3	0.902
	DCC4	0.875
	DCC5	0.766
	DCC6	0.751
Problem Solving	PS1	0.654
	PS2	0.672
	PS3	0.748
	PS4	0.690
	PS5	0.785
	PS6	0.739
	PS7	0.828
	PS8	0.829
	PS9	0.778
Safety	S1	0.669
	S2	0.753
	S3	0.686
	S4	0.676
	S5	0.738
	S6	0.691
	S7	0.750
	S8	0.809
	S9	0.808
	S10	0.815
Ethics	E1	0.808
	E2	0.918
	E3	0.901
	E4	0.809
	E5	0.672

Given these results, the specified model can be considered to have a good fit to the observed data, supporting the hypothesized structure of the constructs under investigation.

The evaluation of the measurement model’s reliability and validity was conducted through CFA, employing Cronbach’s alpha, McDonald’s omega (ω), and the Average Variance Extracted (AVE) as key metrics.

Internal consistency reliability, as measured by Cronbach’s alpha, indicated high levels of consistency across all constructs, with values ranging from .866 for EE to .948 for HM. These results suggest that the items within each construct cohesively measure their respective theoretical concepts. Complementary to alpha, McDonald’s omega coefficients were calculated, yielding similar evidence of reliability. Omega values spanned from 0.811 (EE) to 0.948 (HM), reinforcing the constructs’ reliability within our model. Such consistency in the reliability indices supports the internal consistency of our measurement scales, aligning with the recommended threshold of 0.7 for psychological and behavioral research (Nunnally & Bernstein, 1994). The convergent validity of the constructs was assessed through the AVE, with values exceeding the 0.50 benchmark for all constructs. These findings indicate that a significant portion of the variance in the indicators is accounted for by their respective constructs. Notably, HM demonstrated the highest AVE (0.860), suggesting that this construct provides a robust representation of its indicators. The AVE results, complemented by the high reliability coefficients, affirm the constructs’ validity in capturing the essence of the theoretical concepts they represent. Table 5 shows reliability and convergent validity coefficients for measurement constructs.

Table 5.

Reliability and Convergent Validity Coefficients for Measurement Constructs.

Construct	Cronbach’s alpha	McDonald’s omega	AVE
Performance Expectancy	.928	0.929	0.766
Effort Expectancy	.866	0.811	0.643
Social Influence	.922	0.922	0.797
Facilitating Conditions	.876	0.879	0.646
Hedonic Motivation	.948	0.948	0.860
Habit	.931	0.933	0.779
Behavioral Intention	.879	0.881	0.714
Actual Use	.946	0.946	0.854
Information and Data Literacy	.919	0.906	0.556
Communication and Collaboration	.917	0.908	0.610
Digital Content Creation	.928	0.911	0.679
Problem Solving	.920	0.912	0.562
Safety	.925	0.911	0.547
Ethics	.916	0.902	0.687

Overall, the measurement model displayed excellent reliability and satisfactory convergent validity across the all of constructs. The internal consistency, as evidenced by both Cronbach’s alpha and McDonald’s omega, was exceptionally high, illustrating the cohesiveness of the items within each construct. Moreover, the AVE results further validated the constructs’ efficacy in representing their respective indicators, thereby supporting the measurement model’s integrity and its suitability for subsequent structural analysis.

Structural Equation Modeling Analysis: Unraveling the Dynamics of Digital Competence and ChatGPT Adoption in Informal Learning Contexts

First, the analyzes for assumptions were conducted. Skewness and Kurtosis values within the range of −2 to +2 suggest an approximation to normal distribution. The identification of influential outliers was conducted using Mahalanobis distance with a chi-square distribution threshold. A total of 98 observations were identified as outliers from the dataset, indicating deviations in the multivariate space that could potentially influence the SEM analysis. These outliers were determined based on a threshold set at a significance level of .05, with the degrees of freedom corresponding to the number of variables in the numeric data. Mardia’s test for multivariate normality revealed significant deviations from the assumptions of multivariate normality in the dataset. The Mardia skewness statistic was 132,177.89 (p < .001), and the Mardia kurtosis statistic was 131.09 (p < .001), both indicating a substantial departure from normality. Consequently, the dataset was determined not to follow a multivariate normal distribution. Given the identification of influential outliers and the significant deviations from multivariate normality, robust analytical methods were employed to ensure the reliability and validity of the SEM analysis. The MLR was selected as the estimation method, accompanied by a bootstrap approach to enhance the robustness of the SEM analysis. These methodological choices were made to account for the non-normality and the presence of outliers in the dataset, aiming to produce more reliable and accurate model estimates and standard errors.

The SEM analysis was performed using lavaan 0.6.17 with a MLR. The model consisted of 236 parameters after 149 iterations and was tested on a sample size of 404 observations. The SEM analysis revealed the following fit indices: The Chi-square test statistic was significant, χ²(2,465) = 4837.429, p < .001, suggesting a discrepancy between the model-implied and the observed covariance matrices. However, considering the complexity of the model and the large sample size, additional fit indices were examined to assess model fit.

The CFI and the TLI were 0.910 and 0.904, respectively, indicating a good fit to the data. When considering the robust corrections, the Robust CFI was 0.929, and the Robust TLI was 0.924, both of which further substantiate the model’s adequacy.

The RMSEA was 0.049, with a 90% confidence interval ranging from 0.047 to 0.051, and the probability of RMSEA being less than .05 was .832. The robust RMSEA was 0.043, with its 90% confidence interval from 0.040 to 0.045, suggesting a close fit of the model to the observed data. The SRMR was 0.048, further supporting the model’s good fit. In summary, the SEM analysis demonstrated a good fit of the model to the data, with several key relationships being identified among the constructs of interest. The use of MLR estimation with bootstrap methods ensured the robustness of the SEM analysis, addressing potential issues related to non-normality and outliers in the data.

The structural model revealed significant relationships among the constructs. In the structural equation model examining the determinants of Behavioral Intention to use digital tools in educational settings, several predictors were identified as significant contributors. According to the model’s estimates:

Behavioral Intention as a Dependent Variable

BI was significantly influenced by several predictors. PE positively predicted BI (β = .196, p < .001), indicating that higher PE is associated with stronger BI to use ChatGPT as an informal learning tool. Contrarily, EE negatively influenced BI (β = −.141, p < .05), suggesting that despite lower ease of use BI to use ChatGPT as an informal learning tool may increase. FC positively predicted BI (β = .166, p < .05), indicating that higher FC is associated with stronger BI. The path from HM to BI was also significant (β = .212, p < .001), illustrating that greater HM is linked to stronger BI. Furthermore, the relationship between HT and BI was highly significant (β = .688, p < .001), showing a strong positive effect of HT on BI. On the other hand, the effect of SI on BI was not significant (β = −.067, p = .148), indicating that SI might not play a crucial role in determining BI within this study’s context.

Actual Use as a Dependent Variable

AU was significantly predicted by BI (β = .929, p < .001), denoting a strong positive relationship. This indicates that as BI increases, the likelihood of AU also increases. Meanwhile, PS showed a significant positive effect on AU (β = .178, p < .05), indicating that PS skills may slightly enhance AU. E positively influenced AU (β = .087, p < .05), suggesting that higher ethical considerations are associated with greater AU. However, the path from S to AU was nonsignificant (β = −.005, p = .924), suggesting that perceptions of S do not significantly influence the AU of the ChatGPT as an informal learning tool by undergraduate students. The relationship between IDL and AU was not statistically significant (β = −.118, p = .063), indicating that IDL might not directly influence actual technology use in the studied scenario. The path from DCC to AU was nonsignificant (β = −.032, p = .584), suggesting that DCC skills do not have a significant direct effect on AU. Similarly, the effect of CC on AU was not significant (β = −.059, p = .422), indicating that these skills might not directly influence actual technology usage.

The hypothesized relationships among UTAUT2 constructs, digital competence factors, behavioral intention, and actual use are illustrated in Figure 1.

Figure 1.

Path diagram of the structural equation model: influence of UTAUT2 constructs and digital competence factors on behavioral intention and actual use of ChatGPT in higher education. Solid arrows represent significant paths. Dashed arrows represent non-significant paths. Asterisks next to path coefficients indicate level of significance (*p < .05, **p < .001).

The squared multiple correlations in the SEM revealed substantial variance explanations for the criterion variables based on direct effects. Specifically, 60.9% of the variance in BI was accounted for by direct effects of PE, EE, FC, HM, and HT. This indicates a strong predictive capability of these variables on BI, showcasing the combined influence of expectancy beliefs, ease of use, and habitual usage on the intention to engage in the behavior. Further, 92.1% of the variance in AU was explained by the direct effects of BI, PS, and E, with BI being the most significant predictor. These findings, detailed in Table 6, not only validate the UTAUT2 model’s applicability to the educational adoption of GenAI technologies but also illuminate the critical influence of digital competence on this innovative technology adoption.

Table 6.

Impact of Digital Competence Factors and UTAUT2 Constructs on Behavioral Intention and Actual Use of ChatGPT in Higher Education Settings.

Predictor	Criterion	Direct effect	Squared multiple correlations (R²)
Performance Expectancy	Behavioral Intention	0.196*	.609
Effort Expectancy		−0.141*
Social Influence		−0.067 (ns)
Facilitating Conditions		0.166*
Hedonic Motivation		0.212**
Habit		0.688**
Behavioral Intention	Actual Use	0.929**	.921
Information and Data Literacy		−0.118 (ns)
Communication and Collaboration		−0.059 (ns)
Digital Content Creation		−0.032 (ns)
Problem Solving		0.178*
Safety		−0.005 (ns)
Ethics		0.087*

Note. ns = nonsignificant.

p < .05. **p < .001.

Discussion

This study addresses a research gap by exploring how digital competencies and motivational factors impact the acceptance and use of ChatGPT in higher education. By integrating constructs from the DigComp framework with the UTAUT2 model, it aims to provide a comprehensive view of factors influencing students’ BI and AU of ChatGPT as an informal learning tool. This approach examines both the digital skills students have and the motivational drivers that encourage them to adopt GenAI technologies.

Previous researches confirm UTAUT2′s strong predictive power, effectively accounting for significant variance in BI and AU across different technological contexts (Avcı, 2022; Raman & Thannimalai, 2021; Schretzlmaier et al., 2022; Zacharis & Nikolopoulou, 2022). In this study, the hypothesized relationships between the UTAUT2 constructs and BI were partially supported. Specifically, PE, FC, HM, and HT were found to have significant positive effects on BI, supporting Hypotheses 1, 4, 5, and 6. However, EE surprisingly had a significant negative effect on BI, contrary to Hypothesis 2, and SI did not significantly predict BI, not supporting Hypothesis 3.

HT’s strong positive effect on BI highlights its critical role in ChatGPT acceptance among students, showing that habitual use encourages seamless integration into academic routines. This finding aligns with Lewis et al. (2013) and Venkatesh et al. (2012) who emphasize habit’s influence on technology use in educational contexts. Strzelecki (2024) also found similar effects in ChatGPT adoption, suggesting that regular engagement strengthens students’ intention to continue using it. This trend is supported by studies on various educational technologies, including mobile learning platforms (Sitar-Taut & Mican, 2021), speech to text lecture systems (Farooq et al., 2017), and Google Classroom (Alotumi, 2022; Kumar & Bervell, 2019). However, some studies found no direct impact of HT on BI in e-learning (Twum et al., 2022) or LMSs (Ain et al., 2016; Raman & Don, 2013), indicating a complex interplay of contextual factors in technology acceptance.

HM emerged as the second strongest predictor of BI, showing that the enjoyment from using ChatGPT significantly enhances students’ intent to integrate it into their learning, supporting Hypothesis 4. This finding aligns with S. A. Brown and Venkatesh (2005), who identified enjoyment as a key factor in technology usage intentions, as well as with studies by Raman and Thannimalai (2021) and Strzelecki (2024), which highlighted HM’s role in influencing BI for e-learning and ChatGPT.

The strong impact of both HT and HM on ChatGPT adoption underscores the importance of tracking user engagement. Habitual use suggests that students may continue using ChatGPT regularly due to its convenience and effectiveness, underscoring its academic value. However, the influence of HM should be carefully monitored, as the enjoyment derived from ChatGPT could drive students to engage deeply in learning but may also risk creating dependency or distraction. Monitoring these engagement patterns is essential to ensure that ChatGPT remains an educational asset without leading to overuse.

PE demonstrated a significant positive effect on BI which implies that students’ perception of ChatGPT as beneficial and likely to enhance their academic performance directly influences their willingness to use it. This finding is consistent with previous research affirming the predictive power of PE in technology acceptance (Alrawi et al., 2019; Chun & Yunus, 2023; Ezzaouia & Bulchand-Gidumal, 2021; Hartono & Oktavia, 2022; Havidz et al., 2018; Issaramanoros et al., 2018; Li et al., 2023; Nurcholisha, 2022; Pham et al., 2020; Rosmayanti et al., 2022; Strzelecki, 2024; Zhang et al., 2019). These findings indicate that individuals’ perceptions of the usefulness and performance of a technology strongly influence their intention to use it.

Contrary to conventional UTAUT findings where ease of use promotes technology adoption (Venkatesh et al., 2003), EE in this study negatively impacted BI. The literature shows mixed results: some studies report a negative impact of EE on BI (Yu et al., 2021), others find a positive impact (Gupta & Arora, 2020; Jameel et al., 2023; Strzelecki, 2024; Su & Chao, 2022), and some find no significant effect (Mensah et al., 2023). This suggests the EE-BI relationship is complex and context-dependent.

In the context of higher education students using ChatGPT as an informal learning tool, this negative EE-BI relationship presents a paradox. Despite EE implying ease of use, students may not find ChatGPT inherently easy to use for their studies, yet they still intend to use it. This counterintuitive result may stem from transitioning from keyword-based searches to conversational interactions with GenAI, requiring new skills and approaches to information retrieval (Nielsen, 2023). The learning curve associated with mastering ChatGPT’s interaction mode might contribute to EE’s negative impact on BI. Nevertheless, students’ intention to use ChatGPT persists, with HT emerging as the strongest predictor. This is likely due to ChatGPT’s ability to deliver targeted content and assist in navigating learning material, even if it requires a greater initial investment to learn how to use the tool effectively.

FC positively influenced BI, and this indicates that the availability of resources and support for using ChatGPT strengthens students’ intentions to use it, consistent with the notion that adequate FC can enhance the likelihood of technology acceptance (Jameel et al., 2023; Raman & Don, 2013; Yu et al., 2021).

SI did not significantly affect BI in this study, suggesting that peer or social pressures may not be instrumental in students’ decisions to use ChatGPT for learning purposes. This lack of influence could be attributed to the novelty of the technology; at the time of study, ChatGPT’s adoption may not have been widespread enough to establish it as a normative behavior within student communities. Consequently, social norms and peer influences related to ChatGPT usage might not have developed significantly. Students’ decisions to use ChatGPT as an informal learning tool are likely driven by individual factors, such as personal evaluations of its utility and effectiveness, rather than by social pressures, or recommendations from peers or instructors. This finding aligns with previous research indicating that the effect of SI on BI can be minimal when a technology is in the early stages of adoption and social norms have not been established (Lampo & Silva, 2022; Yang & Forney, 2013). The absence of significant social influence might suggest that individual attitudes and perceived usefulness are more critical determinants of technology adoption in the early phases of its introduction.

In the context of higher education students using ChatGPT as an informal learning tool, this study provides critical insights into factors influencing AU. Notably, BI emerged as the strongest predictor of AU, supporting Hypothesis 7 and reinforcing the well-established premise that a strong intention to engage with a technology leads to its usage. This robust correlation aligns with recent research on ChatGPT acceptance, indicating that when students express a clear intent to use an educational tool, it reliably predicts subsequent adoption, and utilization in their academic activities (Maheshwari, 2024; Strzelecki, 2024).

Among the DigComp constructs, only PS and E positively influenced AU, supporting Hypotheses 11 and 13. This relationship underscores the integral role of digital competence, particularly problem-solving skills, in contemporary education. In STEM education, problem-solving is fundamental for effectively navigating and leveraging technological tools (Buckley et al., 2018). Students with strong problem-solving abilities are more likely to adopt and use digital tools, as these skills help them adeptly handle the complexities of digital platforms (Bhat, 2019). Additionally, Alasadi and Baiz (2023) suggest that GenAI tools like ChatGPT act not only as information repositories but also as active participants in the learning process, aiding in the development of students’ problem-solving and critical thinking skills. This indicates a potential bidirectional relationship between problem-solving and the use of GenAI tools in learning.

E was another construct of digital competence that positively influenced AU of ChatGPT as an informal learning tool. The positive influence of ethical considerations on AU might underscore a nuanced understanding of digital citizenship among higher education students, encompassing respect for content ownership, adherence to ethical norms in technology use, mindful interaction within digital communities, and sensitivity to the digital etiquette of diverse audiences (Carretero et al., 2017; Gümüş & Kukul, 2023; Vuorikari et al., 2016). Students’ preference for ChatGPT may be influenced by the platform’s transparent approach to its limitations, reflecting a commitment to ethical standards of honesty and integrity, fostering a sense of trust, and reliability.

Conversely, IDL, CC, DCC, and S did not significantly predict AU, leading to the rejection of Hypotheses 8, 9, 10, and 12. This may reflect limitations within the DigComp framework, as it was developed before GenAI’s emergence and may not fully capture competencies specific to GenAI use. Levy-Nadav et al. (2024) highlight the need to refine DigComp to include skills such as prompt-writing, managing AI dialogs, and critically assessing AI-generated content. For instance, IDL, which traditionally focuses on searching and evaluating information, may lack GenAI-relevant skills like generating effective prompts. Similarly, CC in a GenAI context involves managing human-AI rather than human-human interactions. GenAI-assisted DCC requires guiding AI in content creation and responsibly integrating outputs, while S with GenAI requires understanding biases, privacy concerns, and responsible usage beyond traditional safety measures.

Thus, while students may meet current DigComp competencies, the framework may not fully address GenAI skills essential for tools like ChatGPT. As Levy-Nadav et al. (2024) suggest, expanding DigComp to include GenAI-specific competencies would better reflect the skills necessary for effective GenAI use in education. This study contributes to the discourse on digital competencies in the GenAI era, underscoring the need for updated frameworks that address skills required by modern educational technologies.

In summary, the non-significant impact of certain DigComp competencies on ChatGPT use may stem from both the study’s timing and the limitations of the current framework. As GenAI technologies become more embedded in education and competency frameworks evolve to incorporate GenAI-specific skills, future research may reveal different relationships between these competencies and technology use. This underscores the importance of ongoing research and framework updates to facilitate effective GenAI adoption in education.

Limitations

First, a significant limitation is the relatively low sample sizes in relation to the number of estimated parameters in both the initial (N = 140) and main (N = 404) analyses. These sample sizes fall below the recommended 5:1 cases-to-parameters ratio for SEM (Tabachnick & Fidell, 2007), due in part to ChatGPT’s emerging usage among students, with only about 40% of surveyed students identified as users. Although the model showed satisfactory fit indices and consistency between samples, the low observation-to-parameter ratio may heighten overfitting risks and restrict model stability and generalizability. Future studies should aim for larger samples as ChatGPT adoption grows, interpreting current results cautiously given these constraints.

Second, ChatGPT’s novelty at the time of the data collection may have impacted familiarity and comfort levels, potentially influencing participants’ responses. As both the platform and its users mature with GenAI tools, shifting perceptions and competencies may affect the relevance of these findings in the future.

Third, the rapid advancement of AI technologies introduces limitations. GenAI tools like ChatGPT are constantly evolving, potentially changing user interactions and experiences over time. This ongoing development may impact the applicability of findings as new features emerge or educational integration deepens.

Fourth, the cross-sectional design restricts causal inferences between constructs. Longitudinal studies are needed to track how digital competencies, motivational factors, and technology use develop over time as users gain experience and the technology evolves.

Lastly, the reliance on self-reported measures may introduce biases, such as social desirability or inaccurate self-assessment, potentially impacting findings. Future research could include objective data sources, like usage logs or behavioral observations, to complement self-reported information, and improve accuracy.

Implications

The findings of this study have several implications for theory, practice, and future research.

Theoretical Implications

This study contributes to understanding technology acceptance and digital competence in the context of GenAI technologies in higher education. The integration of UTAUT2 and DigComp provides a comprehensive model capturing both motivational drivers and necessary digital skills for GenAI adoption. The findings reveal that HT is the strongest positive predictor of BI to use ChatGPT, highlighting the pivotal role of habitual engagement in adopting advanced AI technologies. HM and PE also significantly influence BI, emphasizing that enjoyment and perceived usefulness drive students’ intentions. These results extend UTAUT2 by demonstrating that in the GenAI context, habitual use and enjoyment may outweigh traditional factors like social influence. The positive effects of PS and E on actual use AU highlight that advanced skills are essential for effectively utilizing GenAI technologies. The current DigComp framework does not fully address these competencies, revealing a theoretical gap. Therefore, this study implies that the DigComp framework should be updated to include GenAI-specific competencies such as prompt engineering, managing ongoing dialogs with AI tools, analyzing the pros and cons of AI technologies, and critically evaluating AI-generated content.

Practical Implications

Since HT is the strongest predictor of BI, educators might integrate GenAI tools into regular learning activities to foster habitual engagement. This could involve incorporating ChatGPT into assignments, discussions, and research projects to normalize its use.

The positive effects of PS and E on AU indicate that curricula could emphasize these areas. Incorporating ethical discussions about AI use and problem-solving exercises involving GenAI can enhance students’ competencies and responsible usage. Furthermore, educational institutions might revise their digital competence programs to include GenAI-specific skills such as prompt engineering, managing ongoing dialogs with AI tools, analyzing AI technologies’ pros and cons, and critically evaluating AI-generated content. This ensures students are equipped with the necessary skills to effectively engage with GenAI tools like ChatGPT.

Policymakers and educational authorities should consider updating frameworks like DigComp to incorporate GenAI competencies. This alignment ensures that educational standards reflect the skills required in an AI-driven landscape.

Implications for Future Research

Future research should explore the complex relationship between EE and BI in GenAI-enhanced learning tools. The unexpected negative relationship observed here suggests a need to understand how perceived effort impacts GenAI adoption in education.

As ChatGPT’s usage becomes more common among students, the influence of SI on BI may grow. Longitudinal studies could provide insights into how social norms and peer recommendations shape technology acceptance over time as GenAI tools move from novelty to mainstream. Additionally, examining the impact of instructor endorsements and institutional support on SI could enhance understanding of various sources of social influence on technology adoption.

Future studies should investigate the bidirectional relationship between PS skills and GenAI use. Longitudinal research could capture how sustained ChatGPT interaction influences PS skill development, while experimental studies might assess changes in students’ PS abilities before and after extended use. Qualitative studies, such as interviews or focus groups, could offer insights into how students perceive ChatGPT’s impact on their problem-solving processes, revealing its value beyond immediate informational support.

Longitudinal research employing objective measures along with self-reported data is essential to understand the evolving role of ChatGPT in higher education. As the technology matures, larger and more diverse samples will enable more robust, generalizable findings. Additionally, research should examine how updating digital competence frameworks to include GenAI-specific skills influences GenAI adoption and usage.

Conclusion

This study examined how digital competencies and motivational factors affect higher education students’ acceptance and use of ChatGPT as a learning tool. Integrating UTAUT2 and DigComp constructs, the research found that HT was the strongest positive predictor of BI to use ChatGPT, underscoring the critical role of habit in GenAI technology adoption. HM and PE were the second and third most effective constructs influencing BI, indicating that enjoyment and perceived usefulness significantly drive students’ intentions. Conversely, EE negatively impacted BI, and SI was not significant, possibly due to ChatGPT’s novelty. BI strongly predicted AU, while specific digital competencies—PS and E—positively affected AU. These findings suggest that updating digital competence frameworks to include GenAI-specific skills is essential, helping educators and policymakers support students in effectively using emerging technologies like ChatGPT.

Footnotes

ORCID iD

Sonay Caner-Yıldırım

Ethical Considerations

This study was conducted with the approval of the Educational Sciences Ethics Committee (Ref. No: E-88012460-050.04-349157) of Erzincan Binali Yıldırım University,Erzincan,Turkey.

Author Contribution

The author designed the study,collected and analyzed the data,interpreted the results,and wrote the manuscript.

Funding

The author received no financial support for the research,authorship,and/or publication of this article.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research,authorship,and/or publication of this article.

Data Availability Statement

Data related to this study are available from the corresponding author upon reasonable request. The data include responses to the UTAUT2 and DigComp-based surveys used to investigate the acceptance and use of ChatGPT as a learning tool among undergraduate students.

Declaration of Generative AI and AI-Assisted Technologies in the Writing Process

During the preparation of this work the author used ChatGPT in order to improve the language quality of the manuscript. After using this tool/service,the author reviewed and edited the content as needed and takes full responsibility for the content of the publication.

References

Abdelghani

Sauzéon

Oudeyer

P. Y.

(2023). Generative AI in the classroom: Can students remain active learners?. arXiv preprint arXiv:2310.03192.

Ain

Kaur

Waheed

(2016). The influence of learning value on learning management system use: An extension of UTAUT2. Information Development, 32(5), 1306–1321.

Ajzen

Fishbein

(1988). Theory of reasoned action-Theory of planned behavior (pp. 67–98). University of South Florida.

Akgun

Greenhow

(2021). Artificial intelligence in education: Addressing ethical challenges in K-12 settings. AI and Ethics, 2, 431–510. https://doi.org/10.1007/s43681-021-00096-7

Alasadi

E. A.

Baiz

C. R.

(2023). Generative ai in education and research: Opportunities, concerns, and solutions. Journal of Chemical Education, 100(8), 2965–2971. https://doi.org/10.1021/acs.jchemed.3c00323

Alenezi

(2023). Digital learning and digital institution in higher education. Education Sciences, 13, 88. https://doi.org/10.3390/educsci13010088

Al Farsi

. (2023). The efficiency of UTAUT2 model in predicting student’s acceptance of using virtual reality technology. International Journal of Interactive Mobile Technologies (iJIM), 17, 17–27. https://doi.org/10.3991/ijim.v17i12.36951

Alotumi

(2022). Factors influencing graduate students’ behavioral intention to use Google Classroom: Case study-mixed methods research. Education and Information Technologies, 27(7), 10035–10063.

Alrawi

Samy

Yusoff

Shanmugam

(2019). Factors influencing the technology acceptance of mobile commerce in Malaysia by using the revised UTAUT model. International Journal of Recent Technology and Engineering, 8(4), 694–699. https://doi.org/10.35940/ijrte.c6653.118419

10.

Artacho

E. G.

Martínez

T. S.

Martín

J. L. O.

Marín-Marín

García

G. G.

(2020). Teacher training in lifelong learning—The importance of digital competence in the encouragement of teaching innovation. Sustainability, 12(7), 2852. https://doi.org/10.3390/su12072852

11.

Avcı

(2022). Examining the factors affecting teachers’ use of digital learning resources with UTAUT2. Malaysian Online Journal of Educational Technology, 10(3), 200–214.

12.

Bai

Liu

(2023). ChatGPT: The cognitive effects on learning and memory. Brain-X, 1(3), e30. https://doi.org/10.1002/brx2.30

13.

Bergdahl

Nouri

Fors

(2020). Disengagement, engagement and digital skills in technology-enhanced learning. Education and Information Technologies, 25, 957–983. https://doi.org/10.1007/s10639-019-09998-w

14.

Bhat

(2019). Learning styles in the context of reasoning and problem solving ability: An approach based on multivariate analysis of variance. International Journal of Psychology and Educational Studies, 6(1), 10–20. https://doi.org/10.17220/ijpes.2019.01.002

15.

Bond

Buntins

Bedenlier

Zawacki-Richter

Kerres

(2020). Mapping research in student engagement and educational technology in higher education: A systematic evidence map. International Journal of Educational Technology in Higher Education, 17, 1–30. https://doi.org/10.1186/s41239-019-0176-8

16.

Boscardin

C. K.

Gin

Golde

P. B.

Hauer

K. E.

(2023). Chatgpt and generative artificial intelligence for medical education: potential impact and opportunity. Academic Medicine, 99(1), 22–27. https://doi.org/10.1097/acm.0000000000005439

17.

Brown

S. A.

Venkatesh

(2005). Model of adoption of technology in households: A baseline model test and extension incorporating household life cycle. MIS Quarterly, 29, 399–426.

18.

Brown

T. A.

(2015). Confirmatory factor analysis for applied research. Guilford Publications.

19.

Buckley

Seery

Canty

(2018). Investigating the use of spatial reasoning strategies in geometric problem solving. International Journal of Technology and Design Education, 29(2), 341–362. https://doi.org/10.1007/s10798-018-9446-3

20.

Cain

M. K.

Zhang

Yuan

K. H.

(2017). Univariate and multivariate skewness and kurtosis for measuring nonnormality: Prevalence, influence and estimation. Behavior Research Methods, 49, 1716–1735. https://doi.org/10.3758/s13428-016-0814-1

21.

Carretero

Vuorikari

Punie

(2017). DigComp 2.1: The digital competence framework for citizens with eight proficiency levels and examples of use. https://doi.org/10.2760/38842

22.

Chen

Lin

(2020). Artificial intelligence in education: A review. IEEE Access, 8, 75264–75278. https://doi.org/10.1109/ACCESS.2020.2988510

23.

Chun

T. W.

Yunus

M. M.

(2023). Factors affecting Malaysian ESL teachers’ behavioral intentions for technology use in the post-COVID-19 era. Frontiers in Psychology, 14, 1127272. https://doi.org/10.3389/fpsyg.2023.1127272

24.

Crawford

Cowling

Allen

K. A.

(2023). Leadership is needed for ethical ChatGPT: Character, assessment, and learning using artificial intelligence (AI). Journal of University Teaching and Learning Practice, 20(3), 02. https://doi.org/10.53761/1.20.3.02

25.

Cross

(2007). Informal learning: Rediscovering the natural pathways that inspire innovation and performance. Pfeiffer. http://www.ir.harambeeuniversity.edu.et/bitstream/handle/123456789/1086/Informal%20%20learning.pdf?sequence=1&isAllowed=y

26.

Davis

F. D.

(1985). A technology acceptance model for empirically testing new end-user information systems: Theory and results [Doctoral dissertation, Massachusetts Institute of Technology].

27.

Dickey

Bejarano

Garg

(2023). Innovating computer programming pedagogy: The AI-lab framework for generative AI adoption. arXiv preprint arXiv:2308.12258.

28.

Evangelinos

Holley

(2016). Investigating the digital literacy needs of healthcare students: Using mobile tablet devices for the assessment of student-nurse competency in clinical practice [Conference session]. E-Learning, E-Education, and Online Training: Second International Conference, eLEOT 2015, Novedrate, Italy.

29.

Ezzaouia

Bulchand-Gidumal

(2021, January). A model to predict users’ intentions to adopt contact-tracing apps for prevention from COVID-19 [Conference session]. Information and Communication Technologies in Tourism 2021: Proceedings of the ENTER 2021 eTourism Conference.

30.

Farooq

M. S.

Salam

Jaafar

Fayolle

Ayupp

Radovic-Markovic

Sajid

(2017). Acceptance and use of lecture capture system (LCS) in executive business studies: Extending UTAUT2. Interactive Technology and Smart Education, 14(4), 329–348.

31.

Ferrari

(2013). DIGCOMP: A framework for developing and understanding digital competence in Europe. EUR 26035. Publications Office of the European Union. https://doi.org/10.2788/52966

32.

Fishbein

Ajzen

(1977). Belief, attitude, intention, and behavior: An Introduction to theory and research. Philosophy and Rhetoric, 10(2), 130–132.

33.

Fokkema

Greiff

(2017). How performing PCA and CFA on the same data equals trouble: Overfitting in the assessment of internal structure and some editorial thoughts on it [Editorial]. European Journal of Psychological Assessment, 33(6), 399–402. https://doi.org/10.1027/1015-5759/a000460

34.

Gilson

Safranek

C. W.

Huang

Socrates

Chi

Taylor

R. A.

Chartash

(2023). How does ChatGPT perform on the United States medical licensing examination? The implications of large language models for medical education and knowledge assessment. JMIR Medical Education, 9(1), e45312.

35.

GovTech. (2024, October 30). California law requires schools to teach students about AI. https://www.govtech.com/education/k-12/california-law-requires-schools-to-teach-students-about-ai

36.

Gümüş

M. M.

Kukul

(2023). Developing a digital competence scale for teachers: Validity and reliability study. Education and information technologies, 28, 2747–2765. https://doi.org/10.1007/s10639-022-11213-2

37.

Gupta

Arora

(2020). Investigating consumer intention to accept mobile payment systems through unified theory of acceptance model: An Indian perspective. South Asian Journal of Business Studies, 9(1), 88–114. https://doi.org/10.1108/SAJBS-03-2019-0037

38.

Hair

J. F.

Black

W. C.

Babin

B. J.

Anderson

R. E.

(2010). Multivariate data analysis: A global perspective. Pearson Education.

39.

Han

(2024). Chatgpt in and for second language acquisition: A call for systematic research. Studies in Second Language Acquisition, 46(2), 301–306. https://doi.org/10.1017/S0272263124000111

40.

Hartono

K. W.

Oktavia

(2022). The influence of cryptocurrency transaction as a currency in NFT-based game transactions. International Journal of Emerging Technology and Advanced Engineering, 12(8), 167–179. https://doi.org/10.46338/ijetae0822_20

41.

Havidz

I. L. H.

Aima

M. H.

Wiratih

H. W. R.

(2018). Determinants of intention to recommend WeChat mobile payment innovation in China to be implemented in Indonesia. International Journal of Advanced Engineering Research and Science, 5(7), 297–310. https://doi.org/10.22161/ijaers.5.7.38

42.

Hockly

(2023). Artificial intelligence in English language teaching: The good, the bad and the ugly. RELC Journal, 54(2), 445–451. https://doi.org/10.1177/00336882231168504

43.

Hooper

Coughlan

Mullen

(2008, September). Evaluating model fit: a synthesis of the structural equation modelling literature [Conference session]. 7th European Conference on Research Methodology for Business and Management Studies.

44.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. https://doi.org/10.1080/10705519909540118

45.

Ifenthaler

Schumacher

(2023). Reciprocal issues of artificial and human intelligence in education. Journal of Research on Technology in Education, 55(1), 1–6. https://doi.org/10.1080/15391523.2022.2154511

46.

Issaramanoros

Khlaisang

Pugsee

(2018). Auto mechanic students’ perceptions and readiness toward mobile learning in Thailand. International Journal of Interactive Mobile Technologies (iJIM), 12(5), 28. https://doi.org/10.3991/ijim.v12i5.8906

47.

Jameel

A. S.

Hamdi

S. S.

Karem

M. A.

Awqati

A. J.

Ahmad

A. R.

(2023, August). Understanding the determinants of intention to use mobile payment systems: An extended UTAUT perspective [Conference session]. AIP Conference Proceedings.

48.

Jonsson

Tholander

(2022). Cracking the code: Co-coding with AI in creative programming education. Creativity and Cognition. https://doi.org/10.1145/3527927.3532801

49.

Kean

(2024, September 11). Kean’s bill to improve artificial intelligence literacy for K-12 passes committee. https://kean.house.gov/media/press-releases/keans-bill-improve-artificial-intelligence-literacy-k-12-passes-committee

50.

Kline

(2015). Personality (psychology revivals): Measurement and theory (1st ed.). Routledge.

51.

Kohnke

Moorhouse

B. L.

Zou

(2023). Chatgpt for language teaching and learning. RELC Journal, 54(2), 537–550. https://doi.org/10.1177/00336882231162868

52.

Korzynski

Mazurek

Krzypkowska

Kurasinski

(2023). Artificial intelligence prompt engineering as a new digital competence: Analysis of generative AI technologies such as ChatGPT. Entrepreneurial Business and Economics Review, 11(3), 25–37. https://doi.org/10.15678/eber.2023.110302

53.

Kreinsen

Schulz

(2023, March 18). Towards the triad of digital literacy, data literacy and AI literacy in teacher education – A discussion in light of the accessibility of novel generative AI. https://doi.org/10.35542/osf.io/xguzk

54.

Kumar

J. A.

Bervell

(2019). Google Classroom for mobile learning in higher education: Modelling the initial perceptions of students. Education and Information Technologies, 24, 1793–1817.

55.

Laato

Morschheuser

Hamari

Björne

(2023, July). AI-assisted learning with ChatGPT and large language models: Implications for higher education [Conference session]. 2023 IEEE International Conference on Advanced Learning Technologies (ICALT).

56.

Lampo

Silva

(2022). How social influence and image impact on the intention to use a technology: A study from the battery electric vehicle domain [Conference session]. Proceedings of the 2022 13th International Conference on E-business, Management and Economics. https://doi.org/10.1145/3556089.3556123

57.

Levy-Nadav

Shamir-Inbal

Blau

(2024). Integrating generative artificial intelligence tools to develop digital competences in secondary schools. In Ferreira Mello

Rummel

Jivet

Pishtari

Ruipérez Valiente

J. A.

(Eds.), Technology Enhanced Learning for inclusive and equitable quality education EC-TEL 2024. Lecture Notes in Computer Science (Vol. 15160, pp. 113–118). Springer.

58.

Lewis

C. C.

Fretwell

C. E.

Ryan

Parham

J. B.

(2013). Faculty use of established and emerging technologies in higher education: A unified theory of acceptance and use of technology perspective. International Journal of Higher Education, 2(2), 22–34.

59.

Lieberman

(2024, May 17). A bipartisan bill aims to boost AI education for K-12 teachers. Education Week. https://www.edweek.org/technology/a-bipartisan-bill-aims-to-boost-ai-education-for-k-12-teachers/2024/05

60.

Lin

(2016). Integrating ethical guidelines and situated ethics for researching social-media-based interactions: Lessons from a virtual ethnographic case study with Chinese Youth. Journal of Information Ethics, 25, 114.

61.

Liu

(2023). The digital competence of chinese university students: A survey study. Journal of Education and Educational Research, 2(1), 35–38. https://doi.org/10.54097/jeer.v2i1.5132

62.

Gui

Luo

Yang

Zhang

Tang

(2023). Determinants of intention with remote health management service among urban older adults: A unified theory of acceptance and use of technology perspective. Frontiers in Public Health, 11, 1117518. https://doi.org/10.3389/fpubh.2023.1117518

63.

Livingstone

D. W.

(1999). Adults’ informal learning: Definitions, findings, gaps and future research (NALL Working Paper No. 49). Centre for the Study of Education and Work, Ontario Institute for Studies in Education, University of Toronto. https://utoronto.scholaris.ca/bitstreams/7e50ec16-1083-4aa3-84b0-4b9616804be1/download

64.

Maheshwari

(2024). Factors influencing students’ intention to adopt and use ChatGPT in higher education: A study in the Vietnamese context. Education and Information Technologies, 29, 12167–12195. https://doi.org/10.1007/s10639-023-12333-z

65.

Martzoukou

Kostagiolas

Lavranos

Lauterbach

Fulton

(2021). A study of university law students’ self-perceived digital competences. Journal of Librarianship and Information Science, 54(4), 751–769. https://doi.org/10.1177/09610006211048004

66.

Mensah

M. S.

Arthur

K. N.

Mensah-Williams

(2023). Antecedents of E-learning in undergraduate entrepreneurship education. E-Learning and Digital Media, 21(5), 496–516. https://doi.org/10.1177/20427530231167642.

67.

Mittal

Sai

Chamola

Sangwan

(2024). A comprehensive review on generative ai for education. IEEE Access, 12, 142733–142759. https://doi.org/10.1109/ACCESS.2024.3468368

68.

Nielsen

(2023). AI: First new UI paradigm in 60 years. Nielsen Norman Group. https://www.nngroup.com/articles/ai-paradigm/

69.

Nunnally

J. C.

Bernstein

I. H.

(1994). Psychometric theory (3rd ed.). McGraw-Hill.

70.

Nurcholisha

D. A.

(2022). Determinants of the Moka POS adoption by micro, small, and medium enterprises in Jakarta using unified theory of acceptance and use of technology model. Journal of International Conference Proceedings, 5(3), 150–159. https://doi.org/10.32535/jicp.v5i3.1812

71.

Nygren

Nissinen

Hämäläinen

De Wever

(2019). Lifelong learning: Formal, non-formal and informal learning in the context of the use of problem-solving skills in technology-rich environments. British Journal of Educational Technology, 50, 1759–1770. https://doi.org/10.1111/BJET.12807

72.

OpenAI. (2024). ChatGPT: Optimizing language models for dialogue. OpenAI. https://openai.com/chatgpt

73.

Osborne

J. W.

Overbay

(2004). The power of outliers (and why researchers should ALWAYS check for them). Practical Assessment, Research, and Evaluation, 9(1), 6. https://doi.org/10.7275/qf69-7k43

74.

Ouyang

Zheng

Zhang

Jiao

(2023). Integration of artificial intelligence performance prediction and learning analytics to improve student learning in online engineering course. International Journal of Educational Technology in Higher Education, 20(1), 4. https://doi.org/10.1186/s41239-022-00372-4

75.

Pavlik

J. V.

(2023). Collaborating with chatgpt: considering the implications of generative artificial intelligence for journalism and media education. Journalism & Mass Communication Educator, 78(1), 84–93. https://doi.org/10.1177/10776958221149577

76.

Peters

Romero

(2019). Lifelong learning ecologies in online higher education: Students’ engagement in the continuum between formal and informal learning. British Journal of Educational Technology, 50, 1729–1743. https://doi.org/10.1111/BJET.12803

77.

Pham

T. B. T.

Dang

L. A.

T. M. H.

T. H.

(2020). Factors affecting teachers’ behavioral intention of using information technology in lecturing-economic universities. Management Science Letters, 10(11), 2665–2672. https://doi.org/10.5267/j.msl.2020.3.026

78.

Pinto

Leite

(2020). Digital technologies in successful academic itineraries of higher education non-traditional students. Educação e Pesquisa, 46, e216818. https://doi.org/10.1590/S1678-4634202046216818

79.

Raman

Don

(2013). Preservice teachers’ acceptance of learning management software: An application of the UTAUT2 model. International Education Studies, 6(7), 157–164.

80.

Raman

Thannimalai

(2021). Factors impacting the behavioural intention to use E-learning at higher education amid the covid-19 pandemic: UTAUT2 Model. Psychological Science and Education, 26(3), 82–93.

81.

Rosmayanti

Noni

Patak

A. A.

(2022). Students’ acceptance of technology use in learning English pharmacy. International Journal of Languages Education, 6(3), 314. https://doi.org/10.26858/ijole.v6i3.24144

82.

Ryan

R. M.

Deci

E. L.

(2000). Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55(1), 68–78.

83.

Rybakova

Shcheglova

Bogatov

Alieva

(2021). Using interactive technologies and distance learning in sustainable education. E3S Web of Conferences, 250, 07003. https://doi.org/10.1051/e3sconf/202125007003

84.

Schretzlmaier

Hecker

Ammenwerth

(2022). Extension of the unified theory of acceptance and use of technology 2 model for predicting mHealth acceptance using diabetes as an example: A cross-sectional validation study. BMJ Health & Care Informatics, 29(1), e100640. https://doi.org/10.1136/bmjhci-2022-100640

85.

Shrivastava

S. K.

Shrivastava

(2022). The impact of digitalization in Higher Educational Institutions. International Journal of Soft Computing and Engineering, 11, 7–11. https://doi.org/10.35940/ijsce.b3536.0111222

86.

Sitar-Taut

D. A.

Mican

(2021). Mobile learning acceptance and use in higher education during social distancing circumstances: An expansion and customization of UTAUT2. Online Information Review, 45(5), 1000–1019.

87.

Stevens

J. P.

(2012). Exploratory and confirmatory factor analysis. In Applied multivariate statistics for the social sciences (5th ed., pp. 337–406). Routledge.

88.

Strzelecki

(2024). Students’ acceptance of ChatGPT in Higher Education: An extended unified theory of acceptance and use of Technology. Innovative Higher Education, 49, 223–245. https://doi.org/10.1007/s10755-023-09686-1

89.

C. Y.

Chao

C. M.

(2022). Investigating factors influencing nurses’ behavioral intention to use mobile learning: Using a modified unified theory of acceptance and use of technology model. Frontiers in Psychology, 13, 673350.

90.

Suwanroj

Leekitchwatana

Pimdee

(2019). Confirmatory factor analysis of the essential digital competencies for undergraduate students in thai higher education institutions. Journal of Technology and Science Education, 9(3), 340. https://doi.org/10.3926/jotse.645

91.

Tabachnick

B. G.

Fidell

L. S.

(2007). Using multivariate statistics (5th ed.). Allyn & Bacon/Pearson Education.

92.

Taylor

Todd

(1995). Decomposition and crossover effects in the theory of planned behavior: A study of consumer adoption intentions. International Journal of Research in Marketing, 12(2), 137–155.

93.

Touvron

Lavril

Izacard

Martinet

Lachaux

Lacroix

Rozière

Goyal

Hambro

Azhar

Rodriguez

Joulin

Grave

Lample

(2023). LLaMA: Open and efficient foundation language models. ArXiv, abs/2302.13971. https://doi.org/10.48550/arXiv.2302.13971.

94.

Twum

K. K.

Ofori

Keney

Korang-Yeboah

(2022). Using the UTAUT, personal innovativeness and perceived financial cost to examine student’s intention to use E-learning. Journal of Science and Technology Policy Management, 13(3), 713–737.

95.

Venkatesh

Davis

F. D.

(2000). A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science, 46(2), 186–204.

96.

Venkatesh

Morris

M. G.

Davis

G. B.

Davis

F. D.

(2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27, 425–478.

97.

Venkatesh

Thong

(2012). Consumer acceptance and use of information technology: Extending the unified theory of acceptance and use of technology. Behavioral Marketing eJournal, 36, 157. https://doi.org/10.2307/41410412

98.

Vuorikari

Kluzer

Punie

(2022). DigComp 2.2: The Digital Competence Framework for Citizens - With new examples of knowledge, skills and attitudes (EUR 31006 EN). Publications Office of the European Union. https://doi.org/10.2760/115376

99.

Vuorikari

Punie

Gomez

S. C.

Van Den Brande

(2016). DigComp 2.0: The digital competence framework for citizens. Update phase 1: The conceptual reference model (No. JRC101254). Joint Research Centre (Seville site). https://doi.org/10.2791/607218

100.

Watkins

K. E.

Marsick

V. J.

Wofford

M. G.

Ellinger

A. D.

(2018). The evolving Marsick and Watkins (1990) theory of informal and incidental learning. New Directions for Adult and Continuing Education, 2018(159), 21–36. https://doi.org/10.1002/ace.20285

101.

Wekerle

Daumiller

Kollar

(2020). Using digital technology to promote higher education learning: The importance of different learning activities and their relations to learning outcomes. Journal of Research on Technology in Education, 54(1), 1–17. https://doi.org/10.1080/15391523.2020.1799455

102.

West

C. G.

(2023). AI and the FCI: Can ChatGPT project an understanding of introductory physics?. arXiv preprint arXiv:2303.01067.

103.

Ouyang

(2022). The application of ai technologies in stem education: A systematic review from 2011 to 2021. International Journal of Stem Education, 9(1), 59. https://doi.org/10.1186/s40594-022-00377-5

104.

Yang

Forney

J. C.

(2013). The moderating role of consumer technology anxiety in mobile shopping adoption: Differential effects of facilitating conditions and social influences. Journal of Electronic Commerce Research, 14(4), 334.

105.

Yan

Greiff

Teuber

Gašević

(2024). Promises and challenges of generative artificial intelligence for human learning. Nature Human Behaviour, 8(10), 1839–1850.

106.

Yilmaz

F. G. K.

Yilmaz

Ceylan

(2023). Generative artificial intelligence acceptance scale: A validity and reliability study. International Journal of Human–Computer Interaction, 40, 8703–8713.

107.

C. W.

Chao

C. M.

Chang

C. F.

Chen

R. J.

Chen

P. C.

Liu

Y. X.

(2021). Exploring behavioral intention to use a mobile health education website: An extension of the utaut 2 model. Sage Open, 11(4), 21582440211055721.

108.

Zacharis

Nikolopoulou

(2022). Factors predicting university students’ behavioral intention to use eLearning platforms in the post-pandemic normal: An UTAUT2 approach with ‘learning value’. Education and Information Technologies, 27(9), 12065–12082. https://doi.org/10.1007/s10639-022-11116-2

109.

Zhang

Liu

Luo

Xie

Liu

Zhou

(2019). Factors influencing patients’ intentions to use diabetes management apps based on an extended unified theory of acceptance and use of technology model: Web-based survey. Journal of Medical Internet Research, 21(8), e15023. https://doi.org/10.2196/15023

110.

Zhao

Sánchez-Gómez

M. C.

Pinto Llorente

A. M.

Zhao

(2021). Digital competence in higher education: Students’ perception and personal factors. Sustainability, 13(21), 12184. https://doi.org/10.3390/su132112184

Modeling ChatGPT Adoption Among Undergraduates: An Integrated UTAUT2 and Digital Competence Framework

Abstract

Plain Language Summary

Keywords

Introduction

Theoretical Framework

Model Generation

Information and Data Literacy (IDL)

Communication and Collaboration (CC)

Digital Content Creation (DCC)

Problem-Solving (PS)

Safety (S)

Ethics (E)

Performance Expectancy (PE)

Effort Expectancy (EE)

Social Influence (SI)

Facilitating Conditions (FC)

Hedonic Motivation (HM)

Habit (HT)

Behavioral Intention (BI)

Actual Use (AU)

Hypotheses Development

Method

Study Design

Participants and Setting

Participants and Setting for Confirmatory Factor Analysis

Participants and Setting for the Main Study

Data Collection Instruments

Instrument for Confirmatory Factor Analysis

Instruments for the Main Study

Data Analysis

Data Analysis for CFA

Data Analysis for the Main Study

Results

Confirmatory Factor Analysis: Validating the Adapted UTAUT2 Model for Instructional Acceptance and Utilization of ChatGPT

Measurement Model Validation in the Main Study

Structural Equation Modeling Analysis: Unraveling the Dynamics of Digital Competence and ChatGPT Adoption in Informal Learning Contexts

Behavioral Intention as a Dependent Variable

Actual Use as a Dependent Variable

Discussion

Limitations

Implications

Theoretical Implications

Practical Implications

Implications for Future Research

Conclusion

Footnotes

ORCID iD

Ethical Considerations

Author Contribution

Funding

Declaration of Conflicting Interests

Data Availability Statement

Declaration of Generative AI and AI-Assisted Technologies in the Writing Process

References