AI Literacy in Achieving Sustainable Development Goals: The Interplay of Student Engagement and Anxiety Reduction in Northern Cyprus Universities

Abstract

Technological development in artificial intelligence (AI) has significantly transformed the learning context, and university-level students are now required to possess AI literacy. Effective research, however, has not been conducted to study factors influencing AI literacy. Grounded in engagement theory, self-efficacy theory, and transactional distance theory, this research investigates how anxiety, self-efficacy, and AI literacy are associated among Northern Cyprus University students. A cross-sectional survey was conducted, gathering data from 222 participating students from different universities in the region. Findings indicate that for university students in Northern Cyprus, student engagement significantly influences AI literacy. Also, the relationship between student engagement and AI literacy is mediated by anxiety reduction, which denotes that higher engagement decreases anxiety, enhancing AI literacy. Moreover, it is found that self-efficacy mediates the relationship between student engagement and AI literacy, which indicates that higher levels of engagement result in higher levels of self-efficacy, resulting in higher levels of AI literacy outcomes. Smart PLS 4 structural equation modeling (SEM) was used in data analysis and gaining meaningful insight into these relationships. The study contributes to Sustainable Development Goals (SDGs) 3 and 4 through the facilitation of mental well-being and inclusive quality education via improved AI competencies, proposing evidence-based perceptions into how engagement, anxiety reduction, and self-efficacy boost well-being and education. The findings of the study will enable educators, policymakers, and curriculum developers to design curricula and educational strategies that reduce anxiety, strengthen the self-efficacy of learners, and thereby strengthen their AI literacy level.

1. Introduction

In recent years, the groundbreaking impetus of AI technology has covered the entire globe [1]. The significant progression in AI technologies has transformed many areas, including but not limited to education. With the integration of AI expertise into the education sector, the use of AI literacy has become crucial for students in universities [2]. As the use of artificial intelligence grows, it offers more opportunities to improve education. This integration with learning techniques and environments should be understood from the perspective of students about the use of these inventions and their effect [3]. This study aims to explore the relationship between student embracement, anxiety reduction, and efficiency within the student populations in Northern Cyprus when becoming involved with AI-driven objectives such as chatbots and adaptive learning platforms. For instance, AI chatbots offer instant feedback, enabling students to learn at their own speed [4], while adaptive platforms differentiate content to bridge skill gaps, endorsing comprehensive education [4]. Through the exploration of these facets, we aim to offer further insight to improve comprehension of university students after they engage with AI education and create a positive approach when utilizing AI literacy within an academic setting [2]. Although we are concentrating on AI literacy, it is crucial to keep in mind that Northern Cyprus is a unique cultural center that combines European, Middle Eastern, and Mediterranean elements. Cross-regional research demonstrates that cultural differences exist in sustainability consciousness, which is the interaction of knowledge, attitudes, and behaviors [5].

The past two decades have seen increasingly productive literature on student engagement [6,7,8,9]. Focused mainly on social–ecological analysis and social–cultural speculation, the engagement is theorized as a dynamic system of social and psychological constructs, together with a collaboration process [10]. Although the interaction is not considered to be an attribute of the student alone, it is considered to be a changeable state of mind that is influenced by the ability of the school, family, and peers to offer consistent expectations and learning support [11]. The engagement of students must be one of the most precious motivators for the improvement of teaching within the university sector. Taking into consideration that the student body is the direct beneficiary of superior teaching standards, they offer the opportunity to provide critical feedback on a combination of positive and negative aspects [12]. The engagement of students has interested researchers due to the positive influence of differing outlooks on students’ learning—for example, their critical thinking, cognitive progress, moral and ethical mental expansion, satisfaction, and perseverance [13]. Student engagement helps to improve their results, including advanced academic performance, reduced dropout, and increased emotional security, together with long-term adult education and occupational outcomes [6]. In previous research, the effect of engagement in university life [14], life fulfillment (a central element of well-being) [15], educational performance [3], and motivation [16] was highlighted. The implementation of these new technologies, including AI, and their effect on these dimensions is currently underexplored. Furthermore, nervousness about education, more specifically in the relationship to technology, has negative results, including feeling insecure, nervous, apprehensive, fearful, and intimidated, which can impact the overall experience [17]. Anxiety has been exhaustively explored as a significant variable in education [18], especially differentiated anxiety, comprising trait anxiety and state anxiety: trait anxiety is noted as a comparatively stable characteristic of a person, while state anxiety is a minor feeling that occurs occasionally. MacIntyre and Gardner [19] offer that trait anxiety is “an overall personality characteristic that is pertinent in several occasions”, whereas state anxiety is “a more current experience of anxiety as an emotional state” and specific situation anxiety is “a specific form of apprehension that appears regularly over sometime within a certain circumstance”. Also, Bandura [20] explains self-efficacy as “people’s judgments of their abilities to arrange and complete courses of action necessary to complete specified types of accomplishments”. Furthermore, it can be called self-assurance, as it relates to someone’s personal belief in their capability to complete a task successfully [21]. This study aims to cover these historical beliefs with current issues, specifically concerning AI literacy integration in education.

Following the review of the current literature, which offers numerous papers that have explored the effects of AI in higher education [22,23,24,25,26], there remains a lack of comprehensive research concentrating on the relationship between students’ cognitive, social, and effective learning and the expansion of AI literacy [2,27]. Prominent research has discovered the connection between engagement, self-efficacy, and anxiety in large language models (LLMs) within higher education establishments [27]. Nevertheless, the specific methods through which self-efficacy and anxiety reduction predict AI literacy continue to remain insufficiently tackled in the context of Northern Cyprus. This study addresses these gaps by proposing a complex model within an underexplored socio-educational context, while explicitly linking outcomes to Sustainable Development Goals (SDGs) 3 and 4. This study explicitly contributes to SDGs 3 and 4 by promoting mental well-being and inclusive quality education through the enhancement of AI competencies among university students. By concentrating on these underrepresented people, we aim to focus on contributing important perceptions that will redirect attention to educational practices and policymaking in higher education.

AI is increasing exponentially within the context of higher education internationally [22]. It is considered one of the most impactful game changers in higher education. As a result, students will need data literacy, technological literacy, and human literacy, together with substantial experimental educational methods within each area of literacy, to enable the development of skills required to benefit their futures [28]. The research focused on three techniques of AI in higher education: personalization processes (adaptation for the student on knowledge and individualization), software (intellectual programs and autonomous robots and their ability to learn), and ontologies and semantic web (data and knowledge collection from multiple sources, big data) [26]. The incorporation of AI technology can assist schools and teachers alike to promote a greater comprehension of students’ learning, development, and personal requirements, while also offering specific support to increase teaching strategies and specific education for individuals [1]. AI can furthermore improve education in higher education establishments to initiate a more receptive, varied, and innovative learning environment, which will then enhance the general quality of higher education [29]. Conversely, the growing use of AI in higher education raises alarms concerning academic integrity and creates ethical worries, as this could potentially culminate in plagiarism problems, reduced critical thinking, stifled creativity, and an undermining of uniqueness in education, research, and scholarship [22]. Therefore, the over-dependence on AI technologies by educators and students alike could lead to a reduction in cognitive skills, including but not limited to originality, critical thinking, intellectuality, and how to solve problems. Additionally, concerns surrounding privacy and security of data are areas of apprehension that need to be addressed [24]. This paper aims to investigate the dynamics concerning student engagement, reduction in anxiety, self-efficacy, and AI literacy through a quantitative survey to capture a comprehensive view of Northern Cyprus students in designing tasks. By systematically addressing these components, we aim to provide an exhaustive comprehension of the relationship between students’ perspectives of AI and their personal educational experiences, ultimately enhancing the effectiveness of AI in learning establishments and environments.

2. Literature Review

2.1. Theoretical Background and Hypothesis Development

The theoretical background of this paper is based on a combination of frameworks, including engagement, self-efficacy, and transactional theories, that together allow us to comprehend how university students interact with AI within an educational setting. Engagement within an education environment is centered on the socio-constructivist theory, which considers individual and social involvement to affect learning by creating a meaning of what is being learned [30]. Kearsley and Schneiderman [31] generated a framework for technology-based education and learning, which focused on the consideration that students should be engaged in meaningful learning activities through worthwhile objectives and tasks, along with working in conjunction with others [32]. In their theory, they generated three primary targets to create engagement: (1) a focus on collaboration, centered around communication, planning, management, and social skills; (2) assignments that are project-based to include education through creative and purposeful activities; and (3) a focus on non-academic subjects that involves external customers to engage with [33]. Self-efficacy theory, as generated by Bandura, focuses on the social cognitive theory, which explains that individuals can be influenced by their personal feelings, behaviors, thoughts, beliefs, and motivations [34]. Individuals who have self-confidence are more stimulated to confront challenges straight on and find a solution themselves. This comes from their ability and conviction that they are personally enabled by a suitable skill set to be able to overcome issues, therefore generating more confidence and improving their self-efficacy. In reverse, individuals who have a lack of confidence in their ability may experience frustration and anxiety and will try to evade these types of challenges completely [35]. Transactional distance theory is a well-known current modern theory on the engagement between individuals, surroundings, and interaction designs. Gokool-Ramdoo [36] expects students’ interaction to generate an effective educational experience, creating learning perseverance. He proposed that emotional behaviors like enjoyment and stress could influence this [37]. The collaboration of these theories suggests that while AI technology can boost students’ motivation, self-efficacy, and involvement [38], it could also increase stress and diminish the students’ educational experience [39].

2.2. Student Engagement and AI Literacy

Engagement within the context of the education framework refers to the amount of interaction, interest, and participation between teachers and students through the process of learning. Examining engagement levels is critical for comprehending the effectiveness of the teaching methods and highlighting areas for development and improvement [40]. As a result, educators should generate an engaging and inspiring atmosphere where the students can generate their learning process to improve their understanding and educational results [15]. Student engagement is a complicated and multi-layered experience that usually comprises emotional, behavioral, and cognitive areas [6]. Behavioral engagement comprises features like involvement, awareness, focus on the task, compliance with requirements, demonstration of effort, consistency, attentiveness, deliberation, extensive inputs, and contribution to overall school activities [41]. Emotional engagement comprises emotions of awareness of the content that connects to students who are interested in it [42]. Engaging cognitively in education is a yearning to extend further than the minimum requirements and enjoy the challenge [43]. The collaboration of these dimensions has been utilized to define, gauge, and research students’ engagement; the holistic involvement of all three is required to comprehend what student engagement looks like [44]. In the recent past, AI has revolutionized many different areas of life, and education is one of the most innovative areas that has been transformed by artificial intelligence. The concept of AI literacy has been scientifically categorized into a four-dimensional framework through reviews of existing literature. These dimensions encompass (1) knowing and understanding AI, (2) using and applying AI, (3) evaluating and creating AI, and (4) addressing AI ethics [45,46,47]. This framework provides an inclusive approach to developing the competences required to appropriately categorize, use, evaluate, and collaborate with AI-related products [47]. AI-powered educational platforms such as ChatGPT 4o and generative AI platforms offer a promise to impact the enhancement of student engagement and the development of the education sector [48]. In the ever-increasing development of AI in the education sector, engaging crucial AI literacy within higher education students has become critical. However, the aspects that control AI literacy continue to be insufficiently explained [2]. AI literacy is a central concept of AI in computational ideology, encompassing design considerations of the perception, presentation, and consideration of AI technology, machine understanding, human involvement, and social impact [49]. AI literacy is a combination of competencies that allows people to appraise AI technologies in fine detail, to communicate and cooperate efficiently with AI, and to utilize AI as an online tool from home to work. Users who have more knowledge of and experience in AI, i.e., higher “AI literacy”, could require differing demands and opportunities in comparison to a novice [50]. There are three areas of AI literacy, with the first being the acquisition of basic comprehension of AI—for example, “machine learning, classifiers, decision processing, reasoning, and prediction”. The second aspect is the generation of knowledge from the initial understanding to acquire more extensive knowledge through the application of AI concepts to make conclusions of AI independently, thereby using AI concepts for results, and thirdly, to enable the use of AI to understand the world through problem-solving [51]. As AI evolves and is utilized more, the ability to engage with AI is critical for students to be successful in a world that has developed through technology [52]. The target of AI literacy education is to grow individuals who are not just users of AI but also contributors in this area. This entails designing AI systems that can effectively improve society [52]. Furthermore, improving AI literacy as a fundamental aspect of students’ digital literacy is vital for their chances to engage dynamically and constructively with AI technologies. The integration of generative AI supports constructionist ideology, focusing on active, significant, and perspective-based learning [53].

This paper highlights that fulfilling the three psychological requirements—autonomy, competence, and comprehension—will significantly improve student’s AI literacy, thereby endorsing the use of specific self-regulated learning strategies (SRLSs), comprising intellectual engagement, metacognitive knowledge, resource management, and inspirational beliefs [2]. Furthermore, the effectual use of ChatGPT can noticeably improve student engagement with educational content, thereby creating a prolonged motivation to study [15]. The outcomes suggest that integrating AI into English as a Foreign Language instruction has a positive benefit on students’ cognitive, emotional, and social interactions [25]. As a result, we hypothesize that:

H1. 

Student engagement affects AI literacy among university students of Northern Cyprus.

2.3. Reduction in Anxiety as a Mediator

The term anxiety indicates a conscious, apprehensive emotional state. Technology-related anxiety can be defined as a person being fearful about the use of technology. It demonstrates negative sentiments associated with technology, which potentially could impact the general educational experience while using technology. Feelings of anxiety, misperception, annoyance, and anger can impact not only the education process but also efficiency, social relationships, and the overall welfare of people [39]. Trait anxiety negatively impedes academic accomplishments [54]. The use of AI chatbot instruction demonstrates a reduction in anxiety for students learning to write in English. AI chatbots offer an immediate response to enable students to correct errors instantaneously. The rapid feedback stops students from overthinking their mistakes, therefore reducing anxiety. The use of AI chatbots ensures that students can learn at their own speed [55]. Furthermore, the utilization of LLMs can assist with reducing language learning anxiety, which is a common barrier to acquiring languages, as it offers a low-risk, sympathetic environment in which students can practice without being concerned by any form of judgment [27]. Higher academic accomplishment was related to lower anxiety levels and increased confidence once students had achieved either a singular skill advancement session or multiple [56]. It is considered that anxiety in children reduces their concentration, control, and understanding of memory, as proposed in cognitive theories, resulting in students’ academic skills being affected [57]. In this research, we assert that anxiety reduction mediates the relationship between student engagement and AI literacy.

The research demonstrates that the competitive environment of and concern about disapproval in university life could elevate anxiety among the student population, thereby impacting their academic achievements and overall social well-being [58]. As a result, a system of providing feedback should be implemented to take advantage of social norms to improve standards and promote contemporary accountability, thereby generating a feeling of community and constructive social awareness, which are a vital part of increasing engagement and reducing any negative sentiments [59]. As AI technology is a relatively fresh topic, to encourage students’ interest in being educated in this subject, it is necessary to shape students’ fundamental knowledge regarding AI, determine its relevance, build their conviction, and reduce anxiety through deliberate and considered curriculum design [39]. By developing an atmosphere that lessens anxiety, we anticipate that we will see a significant impact on both commitment and the use of AI [27]. Thus, we propose:

H2. 

Anxiety reduction mediates the relationship between student engagement and AI literacy among students in Northern Cyprus.

2.4. Self-Efficacy as a Mediator

With the increasing use and prevalence of AI, having the conviction to connect with these techniques is vital for students to accomplish their goals in a technology-driven environment. This confidence precisely impacts their ability to educate themselves and utilize AI concepts appropriately. Fears that have been highlighted suggest that young students may struggle to distinguish between AI self-efficacy and personal creative self-efficacy [52]. AI self-efficacy suggests a belief in a person’s aptitude to generate the necessary actions to implement future AI-related challenges [20]. Creative self-efficacy is crucial in AI literacy education, as it builds confidence in students’ belief in their capability to develop and generate solutions using AI technologies [52]. Support services within academic environments are probably going to be important sources for creating a student’s self-efficacy by offering (1) chances to develop their abilities, (2) the generation of constructive feedback on tasks completed successfully, and (3) environments to reduce anxiety in students [56].

Research on self-efficacy in education reports that academic self-efficacy offers the chance to improve students’ health and performance when managing stressful circumstances. Academic self-efficacy as a powerful factor in enhancing academic performance has received much attention in educational psychology [60]. Furthermore, superior engagement actively plays a fundamental role in comprehending, forecasting, and promoting inspiration, as engagement is a substantial positive predictor of psychological need satisfaction and self-efficacy. In turn, this indicates that emotional engagement is an important positive predictor of self-efficacy and inherent value [61]. In former research, it was found that self-efficacy positively impacts AI literacy and future education goals, and lifelong learning is positively associated with employment security and employability [62]. Increased self-efficacy in turn encourages students to take on AI challenges with increased confidence and enthusiasm, generating improved literacy in this area of technology. The conceptual model illustrating the direct and mediating relationships among student engagement, anxiety reduction, self-efficacy, and AI literacy is presented in Figure 1. Therefore, we hypothesize that:

H3. 

Self-efficacy mediates the relationship between student engagement and AI literacy among university students of Northern Cyprus.

3. Materials and Methods

3.1. Sampling and Data Collection Procedure

To assess the scale’s reliability and validity as well as investigate the relationships between the variables, the study used SmartPLS 4.0. Despite the use of a sophisticated model and a considerable number of samples, the PLS-SEM analysis yields the best results [1]. PLS-SEM is appropriate for exploratory research with complex structural models and moderate sample sizes, as it emphasizes prediction and theory development over model fit [63]. Using a criterion consistent with the current model, G*power 3.1 was employed to calculate the sample size required to conduct the present research [64]. Maintaining a sufficient sample size is necessary to achieve a statistical power of 0.8 or 0.9 and a Type I error of 0.05 or 0.01 [65]. To ensure a precise sample size calculation, the study’s optimum alpha (α = 0.01), statistical power (1− β = 0.95), effect size (f2) = 0.1 [66], and three predictors were chosen. A minimum sample size of 176 individuals was suggested via G*Power analysis. Students from three different Northern Cyprus universities and colleges constituted the study’s target population. A certain subset of Northern Cyprus students from the colleges of Art and Culture, Engineering, Medical Sciences, and Business were participants in this study. Northern Cyprus students were selected due to the widespread use of AI in higher education and the diversity of its academic institutions. Simple random sampling, which involved manually distributing questionnaires [Table S1], provided a sample of 320 participants. A simple random sampling method was used to guarantee that each student in the target population had an equal chance of being selected. A pilot study with 30 students was conducted before the main data collection, which utilized an online survey distributed to students in Northern Cyprus. The pilot study facilitated refining the questionnaire based on feedback about precision and relevance. We obtained participants’ consent through a cover letter that explained the study’s purpose and assured anonymity. A final dataset of 222 responses, representing a 69.38% response rate, was deemed eligible for analysis and is considered acceptable. The questionnaire utilized established scales, adapted from previous studies, to measure the study’s key variables.

3.2. Respondents’ Profile

The demographic profile of the student body in Northern Cyprus was diverse. Although women made up a sizable fraction of the group (42.3%), men made up a slight majority of the group (57.7%). The students’ ages ranged widely, with a significant concentration of people in their twenties, especially those who were 25 years of age or younger (37.4%) and in the 25–30 age range (30.6%). The students’ fields of study were varied, although there was a noticeable focus on creative fields like art and design (52.3%). A sizeable percentage of the student body also pursued other subjects, such as engineering (17.1%), business and economics (16.2%), and medicine (11.7%), while other majors constituted 2.7%. Regarding educational attainment, the majority of participants (48.6%) had bachelor’s degrees, followed by master’s degree students (33.3%) and doctorate candidates (17.1%). Only 0.9% of students had completed high school and had no major of study yet.

3.3. Measurement

The researcher used standard questions generated by multiple scholars to develop the research questionnaire for this study. Sections of the questionnaire were designed to assess the study’s main variables: anxiety, self-efficacy, engagement among students, and AI literacy. The questionnaire included items drawn from Wang et al.’s [2] study on the influence of need satisfaction on AI literacy in higher education to assess the AI literacy variable. For instance, the statement, “I can distinguish between smart devices and non-smart devices” was presented to the participants to respond to. Items from Oh et al.’s study [67], which examined job applicants’ attitudes toward learning in the face of digital disruptions, were utilized for assessing self-efficacy. One such item was, “I will have no problem learning new skills”. The Italian Higher Education Student Engagement Scale (I-HESES), which was initially verified by Marcionetti and Zammitti [68], was used to measure the student engagement concept. One example item was, “I get a lot of satisfaction from studying”. Lastly, components from the study by Hwang and Wu [69], which investigated the impact of generative AI on anxiety reduction in design students, were incorporated into the questionnaire to evaluate anxiety reduction. For example, the statement, “I come up with a creative solution to a problem” was communicated to participants to answer. An extensive analysis of the relationships between the variables that were identified in the context of higher education in Northern Cyprus was made possible by this methodical approach.

Cronbach’s alpha was utilized to evaluate the measurement scales’ reliability. All scales had great internal consistency, according to the results. In particular, the AI literacy scale showed a Cronbach’s alpha of 0.931, the anxiety reduction scale a Cronbach’s alpha of 0.905, the self-efficacy scale a Cronbach’s alpha of 0.878, and the student engagement scale a Cronbach’s alpha of 0.909. These numbers imply that the study’s measures had a high level of internal consistency and reliability. A 5-point Likert scale, with 1 denoting “strongly disagree” and 5 denoting “strongly agree”, was used in the survey.

Table 1. Scales, dimensions, outer loadings, reliability, and validity of the construct.

Variable	Dimensions			OL	Cronbach’s α	CR (rho_a)	CR (rho_c)	AVE
AI literacy	Awareness	AIL1	I am capable of differentiating between smart and non-smart devices.	0.759	0.931	0.932	0.942	0.646
		AIL2	I’m not sure how AI technology can assist me.	0.856
		AIL3	I can recognize the artificial intelligence (AI) technology used in the programs and goods I utilize.	0.779
	Usage	AIL4	I am proficient in using AI products or programs to assist me in my day-to-day tasks.	0.792
		AIL5	Generally speaking, I have little trouble learning how to use new AI products or applications.	0.827
		AIL6	I can increase my productivity at work by using AI tools or applications.	0.818
	Evaluation	AIL7	I can assess an AI product’s or application’s strengths and weaknesses after using it for some time.	0.801
		AIL8	I can select the best option from a range of options that a smart agent provides.	0.764
		AIL9	I can select the best AI product or application from a range for a given assignment.	0.691
	Ethics	AIL10	I always use AI goods or applications in accordance with ethical standards.	0.680
		AIL11	When utilizing AI products or applications, I am mindful of privacy and information security concerns.	0.748
		AIL12	I am constantly aware of the misuse of artificial intelligence.	0.688
Self-efficacy		SE1	Learning new abilities will not be an issue for me.	0.810	0.878	0.883	0.911	0.673
		SE2	I’ll be able to manage the demands of work and training.	0.793
		SE3	I do not doubt that I can finish the work training.	0.854
		SE4	I’m determined to learn as much as I can during my work training.	0.860
		SE5	I do not doubt that job training will enable me to secure employment.	0.782
Student engagement	Cognitive engagement	SEN1	My studies give me a great deal of satisfaction.	0.723	0.909	0.911	0.928	0.647
		SEN2	I consider my course to be intellectually engaging.	0.693
		SEN3	Usually, I am inspired to study.	0.627
	Social engagement with the teacher	SEN4	I interact with teachers to help them comprehend the challenges I have when studying.	0.718
		SEN5	I actively seek appropriate constructive feedback from teachers regarding my progress.	0.711
		SEN6	I talk about my work with my teachers.	0.712
	Social engagement with peers	SEN7	I frequently meet with other students to talk about classes.	0.754
		SEN8	I frequently work with other students.	0.799
		SEN9	I feel like I belong to a group of learners who are dedicated to learning.	0.726
	Affective engagement	SEN10	I truly enjoy attending this school.	0.789
		SEN11	This course has exceeded my expectations.	0.785
		SEN12	I enjoy my time at this school a lot.	0.769
Anxiety reduction		AN1	When faced with a challenge, I think of an original answer.	0.837	0.905	0.905	0.933	0.778
		AN2	I consider something from a different angle.	0.858
		AN3	My thought process is creative and open-ended.	0.891
		AN4	I improvise.	0.880
		AN5	I think “outside the box”.	0.686

Note: OL = outer loading, CR = composite reliability, AVE = average variance extracted. Developed by the authors.

contains comprehensive details about the survey questions and their dimensions, outer loadings, T-values, and p-values.

In the construct, the outer loading value indicates how reliable the indicator is. A number greater than 0.7 is advised for outer loading. To do this, the authors modified the model by removing AIL9, AIL10, AIL12, SEN1, SEN2, SEN3, SEN4, SEN5, and AN5. Composite reliability (CR) evaluations and Cronbach’s alpha were used to evaluate the scales’ reliability. Convergent validity was evaluated for this scale using the average variance extracted (AVE) values. The model’s convergent validity was satisfactory because all of the constructs had AVE values over the 0.5 criterion [65] (see Table 1).

4. Analysis and Results

Using a variety of criteria, the research findings provided a PLS-SEM analysis on the research model using SmartPLS through a measurement model or an outer model assessment, and a structural model evaluation or an inner model. Outer loadings, convergent validity, discriminant validity, internal consistency reliability, and model fit tests were used to evaluate the measurement/outer model. The significance value (p-value and T-value), R-square, and Q-square were all part of the structural model evaluation.

4.1. Outer Model Assessment

In 2015, Henseler, Ringle, and Sarstedt introduced the heterotrait–monotrait ratio of correlations (HTMT) as a new criterion for discriminant validity. The heterotrait–monotrait (HTMT) ratio is seen in

Table 2. Heterotrait–monotrait ratio (HTMT) test.

	AI Literacy	Anxiety	Self-Efficacy	Student Engagement
AI literacy
Anxiety	0.753
Self-efficacy	0.826	0.639
Student engagement	0.770	0.885	0.696

. There is sufficient evidence of a pair of constructs’ discriminant validity when the HTMT value is much less than 1 or less than 0.9 [70].

The results showed satisfactory values within the current study and supported the discriminant validity of the model (see Table 2).

4.2. Structural Model Assessment

Table 3. Hypothesis test.

H	Path	β	T-Statistics	p-Values	Remark
H1	Student engagement → AI literacy	0.205	2.320	0.020	Significant
H2	Student engagement → self-efficacy → AI literacy	0.305	8.238	0.000	Significant
H3	Student engagement → anxiety → AI literacy	0.199	3.264	0.001	Significant

SRMR = 0.067, NFI = 0.812.

displays the findings of path analyses, which showed a significant relationship between AI literacy, anxiety, self-efficacy, and student engagement. A positive relationship between student engagement and AI literacy was shown in H1 (β = 0.205, p = 0.020).

4.2.1. Mediation Analysis

The mediation analysis was conducted via SmartPLS 4.0 to study the indirect effects of student engagement on AI literacy through the mediators of self-efficacy and anxiety reduction. The results displayed in Table 3 show that both mediation paths were statistically significant.

Particularly, the indirect effect of student engagement on AI literacy via self-efficacy was determined to be significant (β = 0.305, t = 8.238, p < 0.001), representing that increased engagement improves students’ self-efficacy, which sequentially resulted in higher AI literacy (H2). This result endorses the role of self-efficacy as a mediator, signifying that psychological empowerment contributes expressively to students’ capability to engage with AI devices. Likewise, the indirect effect of student engagement on AI literacy through anxiety reduction was also significant (β = 0.199, t = 3.264, p = 0.001). H3 proposes that superior student engagement helps relieve anxiety, thus allowing for more efficient development of AI literacy expertise.

4.2.2. Model Fit

An excellent model fit was indicated by the SRMR value (SRMR = 0.067) and NFI value (NFI = 0.812). According to Wang [2], the SRMR should be below the widely accepted limit of 0.08. However, an NFI score above 0.80 may be appropriate based on the conditions and other fit indices [71] (see Table 3).

The PLS-predicted statistics for the three dependent variables (DVs)—AI literacy, anxiety, and self-efficacy—are shown in

Table 4. Model evaluation metrics.

DV	R2	Q2	RMSE	MAE
AI literacy	0.682	0.497	0.720	0.559
Anxiety	0.652	0.648	0.600	0.468
Self-efficacy	0.395	0.389	0.788	0.627

. With a Q² of 0.497, a root mean squared error (RMSE) of 0.720, and a mean absolute error (MAE) of 0.559, the model had a solid predictive potential and explained a significant 68.2% of the variance for AI literacy (R² = 0.682). Additionally, anxiety had a high predictive relevance (Q² = 0.648) and a strong fit (R² = 0.652), with reduced error metrics (RMSE = 0.600, MAE = 0.468), indicating a more accurate model. In comparison to the other two DVs, self-efficacy had lower explanatory power and higher error metrics (RMSE = 0.788, MAE = 0.627), suggesting a less accurate model for predicting self-efficacy, even though it still exhibited a decent fit (R² = 0.395) and predictive relevance (Q² = 0.389).

5. Discussion

The discussion is structured to directly indicate and interpret the empirical findings, aligning each hypothesis with its theoretical basis and practical significance to represent a coherent narrative. The results of this research highlight the detailed interaction between the engagement of the student, AI literacy, anxiety reduction, and self-efficacy among the students of Northern Cyprus universities. The results of this paper indicate a significant relationship between student engagement and AI literacy (H1), corroborating the crucial role of active participation in enhancing the comprehension and understanding of AI [2]. Furthermore, we investigated the mediating role of reduction in anxiety within this relationship. Hypothesis 2 (H2) demonstrated a reduction in anxiety as a mediator, confirming that engagement alleviates anxiety [58], which then facilitates AI skill achievement [27]. Likewise, Hypothesis 3 (H3) reaffirmed self-efficacy’s mediating role, thereby implying that students with greater confidence in their skills are more likely to engage with learning facilities, therefore improving their AI skills [62]. Self-efficacy is highlighted as a vital mediating role (H3), suggesting that students with greater confidence in their skills are more likely to get involved with learning material and thereby improve their AI skills. Hence, the study revealed that the indirect effect of student engagement on AI literacy through self-efficacy (β = 0.305, p < 0.001) is stronger than the corresponding effect through anxiety reduction (β = 0.199, p = 0.001). This proposes that while both pathways are significant, the development of self-efficacy plays a more outstanding role in mediating the influence of student engagement on AI literacy. Additionally, the stronger effect of self-efficacy aligns with findings in prior research indicating that students with higher self-efficacy demonstrate greater confidence and effectiveness when engaging with technology-enhanced learning environments [72]. Through carrying out this investigation, we expand our comprehension of how effective engagement performance can enable students to accomplish improved performance. It is possible to control their feelings by demonstrating the indirect consequences of self-efficacy to improve in an AI-enhanced educational future. The stronger mediating role of self-efficacy can be explained through Bandura’s [20] social cognitive theory, which highlights that self-efficacy not only boosts motivation but also drives cognitive engagement. Moreover, engagement, self-efficacy, and anxiety represent a triad of psychological constructs that are inherently interconnected through large language models. Our findings are in line with the previous literature that correlates high engagement, high self-efficacy, and low anxiety with improved learning outcomes [27].

Our findings emerge from universities in Northern Cyprus, a politically distinct and culturally diverse region. The cultural differences may have an impact on how anxiety, self-efficacy, student engagement, and AI literacy interact. Because of its location at the intersection of Middle Eastern and European sociocultural frameworks, Northern Cyprus combines socialist ideals with local ecological issues. For example, peer collaboration (social engagement) may facilitate AI literacy in this region through communal educational practices, which is consistent with the observed mediation of anxiety reduction and self-efficacy.

6. Conclusions

To conclude, this research investigates the perceptions of university students about AI, concentrating on interaction of engagement, reduction in anxiety, and self-efficacy with the design of tasks in Northern Cyprus. The research corroborates that student engagement considerably impacts AI literacy (Hypothesis 1). The hypothesis was supported, demonstrating that engaging students with AI educational actions improves students’ AI literacy by applying prompt strategies and appraising AI’s strengths and weaknesses. Hypothesis 2 proposed that anxiety reduction mediates the relationship between AI literacy and student engagement. The effects of anxiety on students result in a reduced opportunity to succeed in science subjects, which in turn leads to poor accomplishments, thereby reducing understanding and the ability to focus on topics. Hypothesis 2 discloses that anxiety reduction performs as a crucial mediating method that connects the engagement of students with learning results, specifically in technology-based educational environments. Finally, Hypothesis 3, which theorized that self-efficacy mediates the relationship, was supported. Hypothesis 3 indicates that students who have faith in their ability to comprehend and exceed expectations are more prone to setting goals that challenge them, interact in the learning environment, and overcome obstacles. These results indicate that students who are engaged with AI tasks also demonstrate increased levels of self-efficacy, while planning tasks that reduce anxiety can positively impact student engagement with AI. Education and higher education establishments must authorize investment in programs that will improve the understanding and knowledge of AI tools. Techniques for relaxation are indispensable tools for students to manage educational anxiety and enhance educational results. If anxiety is effectively controlled and managed, students demonstrate improved scores on tests and overall educational accomplishments. The requirement to ensure that top-quality educational systems are available to all is shown in SDG 4. SDG3, “Good health and Well-Being”, calls for measures to ensure progression in mental health and human well-being. SDGs 3 and 4 interconnect at this stage. Through the mechanism of perception of well-being based on action and personal emotion, SDG 4 contributes to SDG 3 in the practical integration of goals set for the 20230 Agenda. By emphasizing AI tools’ role in task design, this study demonstrates how technology-driven pedagogy advances SDG 4 (Quality Education) and SDG 3 (Well-Being). Therefore, by amending the curricula to include AI skills, this paper offers practical implications and guidance for teachers seeking to prepare their students for the AI future.

6.1. Theoretical Implications

The results of this research highlight the delicate relationships between student engagement, AI literacy, reduction in anxiety, and self-efficacy among students in Northern Cyprus universities, demonstrating the requirement for engaging in AI literacy in universities. Previous research has investigated the role of apparent independence, comprehension, and relatedness, which offers a crucial role in progressing AI literacy within student bodies in universities [2]. This paper considers how self-efficacy and a reduction in anxiety compare with students’ engagement and the expansion of AI literacy. This research uncovers that student engagement incorporates cognitive (fulfillment from studying, intellectual inspiration, and motivation), social (peer teamwork, teacher involvement), and affective (establishment of ownership) dimensions [68]. Intellectual interaction likely motivates and stimulates cognitive improvement by generating critical consideration, while social interactions with peers and teachers will bring about more collaboration in problem-solving, falling in line with Wang et al.’s study [2], highlighting the need for participation in education in AI learning. The attachment of emotional and affective engagement directly to the learning process enhances motivation, allowing students to explore AI tools and enabling principled consideration, effective usage, and operator awareness [27]. Furthermore, this paper investigates the relationship between engagement, self-efficacy, and anxiety for students utilizing large language models (LLMs). The results of the study underline the critical importance of adapting LLM teaching to cater to the varied requirements and inclinations of students, showing the way forward to more efficient and inclusive teaching strategies [27]. Previous research has highlighted that enhanced involvement in education is demonstrated by specified fields of education—for example, higher interaction with different topics and more social interactions with teachers and students alike [15]. Continuing the research by Bandura, student engagement in education can be improved through students’ social interactions with other people in the context of learning, embracing parents, teachers, peers, and other supporting staff [44]. Social and emotional skills are skills that can be improved. Students with sound social and emotional abilities generally have better performance in school, both academically and socially, which highlights predicted success in life. Within the last decade, there have been frequent requests for the instruction of social and emotional skills within teacher training programs [44].

The findings reinforce Bandura’s social cognitive theory by representing that both emotional and cognitive dimensions—namely, anxiety and self-efficacy—mediate the enhancement of complex competencies like AI literacy [21]. Incorporating these aspects encourages a holistic impression of AI usage, focusing not only on technical skills but also on the critical consideration, inspiration, and responsibility in the generation of AI technologies for the future [66]. Literacy should involve a gradual comprehension of generative AI mechanisms, restrictions, and societal suggestions to generate a foundation of extended digital aptitudes. The development of AI literacy as a fundamental part of students’ digital comprehension is imperative for their capability to absorb information critically and beneficially with this technology. By ignoring these cornerstones, the implementation of generative AI technology turns out to be superficial and misses out on the potential to develop empowering skills and intellect [53]. The analysis proposes that AI can improve personal development, social engagement, constructive participation, and the growth of humanistic traits, which are all vital for enhancing educational performance and retention rates [67].

6.2. Practical Implications

The research results offer noteworthy practical suggestions for educators who are looking to improve student AI literacy. The primary goals should be to highlight active teaching strategies. Educators should integrate AI tools such as AI chatbots and analytics-driven platforms to create adaptive, low-anxiety learning environments. This suggests that through group meetings, debates, problem-solving, and practical approaches, the students can become vital parts of their learning process. Furthermore, this encourages a more comprehensive understanding of theories, and it advances information retention while also generating a supportive, low-anxiety learning environment. A supportive environment can also assist students in generating a feeling of belonging, which is a vital element, possibly even a precondition, of student inspiration. Students’ feeling of being part of the community in the classroom assists in developing a sense of worth for the class’s involvement together with a sense of comprehension, or self-efficacy, concerning the assignments. Together, these inspiring factors forecast student interaction and higher academic accomplishment. Additionally, teachers should encourage the building of student self-efficacy. This combines creating activities that ensure students achieve success by offering challenges that are attainable with clear learning targets and expectations; meaningful, time-specific, productive feedback; opportunities to practice and understand concepts; and tasks that are separated into manageable pieces. Relaxation practices, management of stress skills, and self-help techniques assist students in generating an emotionally stable environment. In the long run, to ensure a successful approach is developed, a holistic curriculum within the framework of content, emotional, spiritual, and engagement-linked aspects will be included. At the point of constructing a curriculum, teachers ought to set difficulty-plus-sequence assignments as opposed to lower intellectual difficulty compared to higher intellectual complexity. By tackling these practical concerns, teachers can prepare students with a combination of technical skills and the emotional strength to complete their educational journey with AI technologies. This paper offers valuable insights for educators, developers of curricula, and legislators to create educational standards and proposals for AI literacy. Moreover, the findings promote curricula that drive AI literacy across fields, structured in three stages: (1) foundational AI concepts (e.g., machine learning, ethics), (2) applied AI skills (e.g., using tools like ChatGPT for problem-solving), and (3) critical evaluation (assessing AI’s societal impacts). Practically, the results support a plan of AI-integrated education that directs both affective and cognitive dimensions of learning. By identifying how these variables affect AI literacy, educators and curriculum designers can better shape learning environments that not only equip students with technical skills but also foster confidence and emotional enthusiasm.

7. Limitations and Future Research Recommendations

While the study presents meaningful insights, it is important to critically evaluate the limitations that may affect the generalizability of the results. This research paper has some limitations. First, the research focused mainly on student engagement, reduction in anxiety, and self-efficacy, which possibly discounted other potential variables, including, for example, the possibility of exploring the role of AI anxiety. Future research could examine how AI literacy interacts with students’ engagement and the development of AI anxiety. Secondly, the data were collected from three universities in Northern Cyprus, restricting the generalization of the results. Furthermore, it is crucial to recognize that the results may not speak for all universities; therefore, future research could consider an expanded number of colleges and universities, both within Northern Cyprus and globally, to accomplish a more thorough representation dataset. Thirdly, this research is mainly quantitative in focus. The absence of qualitative data subsequently restricts analysis to explore more intense insights into the personal experiences involving AI interactions and their effect on emotions and potential challenges. This limitation highlights the requirement for qualitative investigations to encapsulate inferred perspectives and more comprehensive experiences, which quantitative methods on their own cannot offer. To alleviate this, any forthcoming research should include qualitative methods, including in-depth interviews or focus groups, to enable a more balanced exploration of students’ AI-related perceptions, experiences, and challenges. Moreover, our research solely concentrated on the viewpoint of students about the impact of AI, ignoring the position of teachers and educators alike. Future research should consider investigating these effects of AI from the viewpoint of teachers, too, examining how AI integration impacts teaching practices and pedagogical strategies. Finally, our results, generated in the cultural setting of Northern Cyprus, underline the need to treat culture as an active moderator in the AI literacy-to-sustainability pathway. Future studies should replicate this model in diverse cultural contexts.