Bridging the Digital Divide in Climate Education: Evaluating AI-Powered Pedagogical Innovations to Enhance Climate Literacy in Pakistani Higher Education
A
Dr.
RehmatShah1MuhammadNaeemSarwar2
AhmedHusseinAl Rassas3
GulnazHameed4
SabaHanif1
RanaYassirHussain5
AhmedHussein3
AlRassas3✉Emailaalrasas@gmail.com
1Department of Educational Leadership and Policy Studies, Division of EducationUniversity of Education54770LahorePakistan
2Department of STEM Education, Division of EducationUniversity of Education54770LahorePakistan
3AlBaydha UniversityAl BaydaYemen
4Department of Elementary and Teacher EducationLahore College for Women University54000LahorePakistan
5UE Business School, Division of Management and Administrative ScienceUniversity of Education LahoreLahorePakistan
Dr. Rehmat Shaha, Muhammad Naeem Sarwar, b, Ahmed Hussein Al Rassasc, Gulnaz Hameedd, Saba Hanifa , Rana Yassir Hussaine
a Department of Educational Leadership and Policy Studies, Division of Education, University of Education, Lahore, 54770, Pakistan
b Department of STEM Education, Division of Education, University of Education, Lahore, 54770, Pakistan
c AlBaydha University, Al Bayda, Yemen
d Department of Elementary and Teacher Education, Lahore College for Women University, Lahore, 54000, Pakistan
e UE Business School, Division of Management and Administrative Science, University of Education Lahore, Lahore, Pakistan
Ahmed Hussein Al Rassas
(Corresponding Author)
AlBaydha University, Al Bayda, Yemen
Email: aalrasas@gmail.com
Abstract
With the increasing severity of the global climate crisis, climate literacy is becoming an essential concern that must be tackled by empowering students in climate literacy, especially in the Global South, where education inequities are still present. This research explores how AI-driven pedagogy can support climate literacy of university students in Pakistan with a focus on the role of digital inclusion and student engagement that act as mediators. This study was base its theories on constructivist learning theory, theory of digital equity, and theory of self-determination, and in the proposed conceptual model, the use of AI-enabled instructional innovations can support the development of the learners to better understand the problem of climate issues by ensuring that they have equal access to resources in the digital domain and the way to engage in learning activities. The sample included the data of 400 students in public universities, and a quantitative method, Partial Least Squares Structural Equation Modeling (PLS-SEM)) was used. It is suggested that AI pedagogy is of significant concern to climate literacy, where student motivation and digital inclusion are partial mediators. It suggests the trans figurative potential of generative AI to facilitate education inequalities in higher education and promote an ecological consciousness in education. The study can be applied practically in policymakers' and teachers' terms, generally, who would like to utilize education technologies with an eye towards closing the digital divide and making sustainability of education in cases of under-resourced resources more significant.
Keywords:
Climate Literacy
AI-Powered Pedagogy
Digital Inclusion
Generative Artificial Intelligence (GenAI)
Sustainability Education
A
Introduction
A
Because climate change is now a point of concern everywhere in the world, efforts have been made to teach climate literacy to elementary school students. Anderson (2012) and McCaffrey and Buhr (2008) define climate literacy as the capacity to learn about, assess, and act on climate issues in a manner that supports sustainability and good decision-making. They are tasked with educating students on how to adjust to a changing world in a scientifically valid and socially and politically adept manner (Anderson, 2012; Trott, 2021). One of the countries exposed to climate is Pakistan, but quality climate education is not present due to educational inequality and knowledge shortage (Khan et al., 2021; Mahmood, 2023). To address these deficiencies and promote abilities focused on sustainability, the current setting calls for creative educational interventions.Artificial intelligence (AI) education has increasingly been proposed as an emerging technology that can transform climate literacy due to its ability to provide personalized learning experiences, resource access to a diverse pool of digital materials, and support scaffolded experiences of exploring the many facets of climate challenges (Zawacki-Richter et al., 2019; Roll & Wylie, 2016). Regenerative AI models, such as the ones supporting adaptive simulations, intelligent tutoring systems, and situation-cued problem-solving, can be implemented to achieve similar goals, i.e., reinforce critical-thinking and decision-making capacities of students (Chen et al., 2020; Holmes et al., 2021). In countries like Pakistan, where access to quality education remains a problematic matter in most areas, AI can provide scalable and situation-related resolutions in democratizing the process of climate education (Ali et al., 2023). The given thesis aligns with constructivist learning theory involving learner activity prioritized and digital equity theory, which entails equal access to the potential of technology, which is becoming the pillar of education at the modern stage (Solis and Duran, 2021).
Despite the potential of AI pedagogy to enhance learning, every child must have unrestricted access to technology, the internet, and literacy through these tools (van Dijk, 2020; Warschauer, 2003). There remains a wide digital divide in Pakistan, with marginalized groups and rural populations often left behind in digital advancements (Khan et al., 2021; UNESCO, 2022). Climate literacy founded on artificial intelligence threatens to exacerbate already-existent gaps unless digital inequality is considered. The evidence provided by Robinson et al. (2015) and Mahmood (2023) indicates that digital inclusion is more than a contextual factor; it is a required mediator that drives the effect of AI technology toward better climate literacy.
Student engagement is also one of the determinants that distinguish whether students are becoming more climate literate due to AI-based learning. The three-dimensionality of being engaged in the process of learning concerns the cognitive, emotional, and behavioral aspects (Fredricks et al., 2004; Kahu, 2013). Holmes et al. (2021) and Chen et al. (2020) concluded that with the help of AI-based solutions, there would be an increase in engagement, understanding, and memorization of sustainability learning due to gamification, interactive feedback, and peer-to-peer learning discussions. Alternatively, whereas rote learning is the main focus of conventional approaches to climate education, with AI, it might be possible to reframe climate education as an action-driven, investigation-based learning program (Ali et al., 2023; Zawacki-Richter et al., 2019).
The concept of AI application when learning could enhance climate-related competence, sustainability knowledge, and achievement gaps based on the literature at different institutions of higher learning across the globe (Luckin et al., 2016; Trott, 2021). The issues of education and climate change represent an existential threat to Pakistan (Mahmood, 2023) due to the lack of existent empirical research that has been carried out in the Global South to prioritize these interconnections. It is theoretically grounded in the theory of digital equality, self-determination theory, and constructivist learning theory. Constructivist learning theory refers to learning as the process of students actively participating in the creation of their knowledge by using a combination of observation, experimentation, and analysis (Vygotsky, 1978; Piaget, 1972). The other key component of long-term innovation in education is the digital equity theory (Warschauer, 2003; van Dijk, 2020), highlighting the socio-technical characteristic of equity use of technology resources. Approximate to the self-determination thesis, which describes the working of motivation, the adaptive AI-subordinated learning environment can satisfy the reading of students based on autonomy, competence, and relatedness (Ryan & Deci, 2000). The current study lays out a unified foundation in assessing the effect of AI-based innovations in the education sector to help reduce technologically-based inequality and facilitate climate literacy by embracing the diversified theoretical views. The present research has both policy and practical implications.
The standing of evolving digital literacy creativities, emerging inclusive digital structure, and instituting impartial access policies to climate education has been underscored by both Ali et al. (2023) and UNESCO (2022). Concerning the arguments by Holmes et al. (2021) and Zawacki-Richter et al. (2019), AI teaching tools support teachers in adopting strategies that promote active participation, prioritizing learning, and addressing sustainability in the students’ communities. New education innovations, such as the incorporation of AI at the university level, serve to make learning more climate change adaptable (Khan et al. 2021; Mahmood 2023).
Scholarly gaps remain despite extensive research. There has been some research in South Asia, but most of the material has been in the Western context. This inequitable lack of representation from the Global South leads to a limited understanding of the relationships between AI pedagogy, social inequity, cultural factors, and the institutional barriers in poor countries. This deficiency has been targeted in this essay through the case of climate awareness and AI research in higher education in Pakistan, along with some other relevant regional perspectives.
Climate literacy as a higher education pedagogical need
It is the biggest problem in the world today since it hurts ecosystems, communities, and economies. University education has to do a better job of teaching about climate change, especially in underdeveloped countries. Developing problem-solving, critical thinking, and decision-making capability transferable to social and personal contexts is a key aspect of climate literacy, branching from scientific expertise in the climate system at the foundational level. Such competence in sustainable development leads to students becoming university-level climate change leaders and civic-minded citizens (Wals & Peters, 2018; Monroe et al., 2019). Climate literacy has two functions, one of which is climate justice and the other is the ability to operate capacity to respond and adapt, particularly in the Global South where a dearth of infrastructure and resources make exposure to climate risk more pronounced. One of the possible ways that tertiary education institutions can help students to act on climate change in ethical and sustainable ways is by embedding climate literacy into their curriculum (Monroe et al., 2019).
2. Structural Barriers to Climate Literacy Integration in Pakistani Higher Education
Structural barriers, such as the digital divide also exist that have not allowed the introduction of climate literacy in Pakistani tertiary education. This is given the fact that, as important as the topic of climate literacy might be, climate literacy proves to be an uphill task to bring it to tertiary education in Pakistan. This affects the students in rural areas, women, and socio-economically disadvantaged groups so that they cannot access equal learning opportunities (Zahid et al., 2025). Rural children, girls, and economically disadvantaged children are likely to be differentially equipped with modern digital equipment, consistent internet, and high-quality instructional materials. Because digital platforms are making increasingly more diverse necessary instructional material available, like climate models, interactive simulations, and artificial intelligence-based learning systems, these discrepancies are particularly bad news in the area of climate education. Along with keeping some groups out of availing such learning opportunities that can transform their lives, unequal access to them works to sustain wider social inequalities (Rehman & Khan, 2025).
In Pakistan, where the higher education system is already unsteady due to shortages of resources, bridging the digital divide is highly crucial in the democratization of access to climate data. This is due to the fact that Pakistan is a developing country. Climate literacy education needs to be offered to all students irrespective of his or her socioeconomic status or geographical location, and narrowing the digital divide is a necessity towards realizing this vision (Zahid et al., 2025; Rehman & Khan, 2025).
3. AI-Powered Pedagogy as a Transformative Educational Paradigm
There has been recent focus on generative artificial intelligence (GenAI) as one of the learning paradigm-changing technologies, considering its adaptive, scalable, and cost-effective pedagogical upgrades. Adaptive learning materials, learning adjusted to students' competency level, and immediate feedback are aspects in accordance with the constructivist learning theory, which places special focus on student-centered knowledge building (El Fathi et al., 2025).
These aspects can be produced by AI-powered systems. AI technologies allow for simplification of content, the creation of realistic climate scenarios, and interactive inquiry student-driven in climate learning, which proves particularly effective in handling sophisticated and interdisciplinary topics. Research (El Fathi et al., 2025; Tan & Maravilla, 2024) reports that GenAI technologies, such as ChatGPT, are improving students' conceptual understanding in STEM areas. These technologies allow for misconception elimination and critical use of course content. AI-based pedagogy is an affordable, scalable solution to providing high-quality, adaptive learning opportunities in resource-poor settings, like Pakistan institutions, without overwhelming educators with unnecessary burdens or the need for heavy infrastructure investments. This is a valuable climate literacy instrument for colleges operating with limited budgets.
4. Theoretical Foundations: Constructivism, Digital Equity, and Self-Determination
AI has shown promise for supporting climate education in several ways because of advancing, interconnected theories. In the constructivism approach, education focuses on learners, as they become meaning makers by participating, collaborating, and reflecting. Inquiry spaces, scaffolding student inquiry, and providing adaptive feedback are ways AI systems can support instruction using constructivist approaches (Tan & Maravilla, 2024). Digital equity theory deals with providing access to AI and other digital technologies to the extent that the access is equitable in relation to gender, geography, and finances (Rehman & Khan, 2025). Finally, self-determination theory (SDT) emphasizes the significance of autonomy, competence, and relatedness for intrinsic motivation, and, when well designed, GenAI technologies can provide autonomy for students to pursue individualized pathways, develop competence through adaptive scaffolding, and foster relatedness in constructive AI-aided collaborative environments (Tan & Maravilla, 2024; Elgaronline, 2023). When considered together, these theories provide the justification for AI-aided pedagogy aimed at achieving climate literacy. In doing so, they address the issues of motivation, equity, and active learning.
5. Cognitive and Conceptual Advancements through AI Pedagogy
The impact of AI pedagogy on student learning is more and more documented. For instance, research has demonstrated that GenAI in STEM education not only improved conceptual understanding, all while reducing the common misconceptions in student thinking, but also decreased the time needed for instruction (El Fathi et al., 2025). Interdisciplinary research in climate education shows that the incorporation of GenAI in educational programs for climate futures increases student agency, hope, action competence, and AI literacy (Springer, 2025). These results suggest that AI not only enhances knowledge but also supports the attainment of emotional and motivation factors of the learning climate in climate literacy, these factors are essential. However, researchers also warn against excessive dependence on AI tools because those systems can replicate biases, offer incomplete information, and minimize opportunities for human-to-human contact unless thoughtfully implemented (Tan & Maravilla, 2024). Therefore, while AI has potential to revolutionize, its application must be well-engineered within education frameworks to reap the most benefit and avoid harm.
6. Digital Inclusivity as a Prerequisite for AI-Enabled Climate Pedagogy
Pakistan is constrained by time even when the AI potential is great. Regarding urban students, Zahid et al. (2025) points out that 'males coming from high-income families were more likely to use AI technology' and that 'socioeconomic status and location' had an 'important' influence on students’ attitudes toward AI and its usage. Limited infrastructure is one reason why AI is difficult to implement in rural, resource-poor, and rural-remote areas. Internet access, out-of-date technology, and the absence of technical support all contribute to these obstacles (Rehman and Khan, 2025). Hence, the unequal availability of resources makes the importance of digital inclusion all the more obvious. If neglected, AI will have widening discriminatory impacts, benefiting the rich and ignoring the poor (Zahid et al., 2025; Rehman & Khan, 2025). Hence, providing equitable access to electronics for AI-enabled education and admission is a technological and moral responsibility.
7. Conceptual Framework and Research Objectives
This study hypothesizes a conceptual model based on such empirical and theoretical evidence. It hypothesizes that teaching facilitated by AI impacts climate literacy directly and indirectly. Student engagement and online engagement takes mediating positions. Motivated students are actively involved students, express motivation, and participate in learning activities; digital inclusion is the fair provision of access to digital tools, digital skills, and AI technology.
The model will be evaluated on the basis of information collected from 400 Pakistani university students, employing quantitative methods, i.e., Partial Least Squares Structural Equation Modeling (PLS-SEM). This study seeks to investigate how AI-based pedagogy can advance climate consciousness in higher education and bridge structural deficits.
8. Implications for Policy and Practice
Beyond the academic community, this research has implications for university administrators, lawmakers, and policymakers. By showing how AI-based instruction could raise environmental consciousness and help bridge the digital divide, this study contributes to the ongoing discussion on green education technology. Curriculum overhauls to incorporate climate consciousness into more inclusive courses on sustainability education, investment in cyber infrastructure, and employee training in the efficient use of artificial intelligence are all possible policy recommendations. A study by El Fathi et al. (2025) and Zahid et al. (2025) shows that artificial intelligence-enabled climate education strategies could ease disparities in education, foster ecological consciousness, and influence sustainable development goals in Pakistani higher institutions and their counterparts in the Global South.
3. Research Methodology
This was a quantitative cross-sectional study that sought to find out the effectiveness of the education process that was augmented by artificial intelligence (AI) on climate literacy in the Pakistani institutions. Due to opportunities to examine statistically both direct effects and mediated effects along with possibilities to investigate correlations across time dimensions at a single point in time, a cross-sectional study design can be deemed as appropriate in relation to this specific aim. The sample group that was used included a broad spectrum of undergraduate students in regard to educational and socioeconomic backgrounds. These students attended the universities and institutes in the Punjab, Pakistan. The researchers employed purposive rather than random sampling to ensure that the students completed such a wide range of activities within the technology enhanced learning and sustainability education. Hair et al. (2022) came up with a statistical methodology known as Partial Least Squares Structural Equation Modeling (PLS-SEM) and it was estimated that data on 400 undergraduate and graduate students will be adequate to run the analysis.
Data was obtained using a standardized questionnaire that was administered both online and offline to enhance level of access and participation. The survey covered a total of five dimensions/topics: demographics, student engagement, digital inclusion, climate literacy, and educational empowerment through the use of artificial intelligence. All of the components were examined with instruments that have been used in other studies in the past. The scales used were of the five-point Likert scale, and they were created particularly for the higher education sector in Pakistan. For the purpose of ensuring that the poll was both clear and culturally suitable, a preliminary examination was carried out with fifty students. In order to improve clarity, the results revealed that some small adjustments were required. During the process of data analysis, there were two diverse stages. We carried out an analysis that enabled us to establish the validity and reliability of the measurement model. The internal constancy was measured by composite reliability (CR) and Cronbach alpha and convergent validity was measured via average variance extracted (AVE). Both of these procedures were carried out to analyze the reliability of instruments. To achieve the establishment of discriminant validity and to demonstrate the sharpness of the constructs, Fornell-Larcker criteria and heterotrait-monotrait ratio (HTMT) were useful.
Structural model was then done using PLS-SEM in Smart Pls 4. The manipulation with the route coefficients, the effect sizes (f2), and the explained variance (R 2) became possible due to this information. The contribution that we made in our research is a study of the mediating role played by digital inclusion and student engagement through bootstrapping and 5,000 resamples. A variety of fit measures, such as normed fit index (NFI) and standardized root mean square residual (SRMR) are bound to be evaluated, were applied in order to ensure that the estimated model was as robust as possible. The research was conducted in accordance with ethical standards in every aspect. All participants provided informed consent following the approval of the study by the institutional review board of each participating university. All participants were assured that their responses would remain confidential and anonymous, and their participation was completely voluntary.
Measurement Model Evaluation
A
Fig. 1
presents the results of the measurement model evaluation, which indicate that the latent constructs of APP, DI, SE, and CL are valid and reliable. Prior to performing any sort of structural analysis, the model has to be verified to ensure that each construct represents the theoretical components that it ought to. This involves convergent validity, internal consistency, and indicator reliability. During the measurement model assessment process, the outer loadings of the observed variables on their respective latent constructs show that with the cutoff threshold of 0.70, each item demonstrates indicator reliability as indicated by Hair et al. (2019). From this, it is possible to conclude that all the items represent artificial intelligence in educational settings since an APP4 loading score of 0.691 is sufficient. The loadings of Di, SE, and CL all fall within the range of 0.699–0.761, 0.735–0.835, and 0.746–0.827, respectively, ensuring that these items are reasonably dependable. They all fall within this range. Theoretically, the underlying prediction based on the constructs corresponds to the experimental data, and that contributes not only to the validation of the measurement model but also puts light on the latter. The constructs are stable and coherent, as the CR and Cronbach's Alpha are two ways of measuring the dependability of internal consistency. When the CR is greater than 0.70, this suggests that every single signal embodied in each notion is individually important and distinctive. Further, the dependability of the coherence in the measurement scales means that the constructs can be profitably forwarded for the next step of structural modeling.
The sequential stage of structural modeling may now be pursued with respect to the given constructs. On the case of convergent validity, the constructs all range above the threshold of 0.50 as highlighted by the Average Variance Extracted (AVE) analysis (Fornell & Larcker, 1981). The corresponding AVEs are 0.627 for APP, 0.533 for DI, 0.635 for SE, and 0.624 for CL. Thus, it can be concluded that each latent construct dominates more than half of the variance of the indicators within each latent construct. These findings provide evidence that the conceptions possess sufficient convergent validity for the constructs, signifying that the variables in combination accurately reflect the theory of construct in reality.
The psychometric approach has shown that the assessment model is both reliable and valid. All constructs have effectively exhibited indicator reliability, internal consistency, and convergent validity. So, we can be sure that the next analysis of the structural model will show the proposed links between AI-Based Pedagogy, Digital Inclusion, Student Engagement, and Climate Literacy. The following structural interpretations are based on the model's ability to accurately measure the theoretical parts of AI-enabled sustainability education and how those parts are related to each other.
Construct | Cronbach’s Alpha | Composite Reliability | Average Variance Extracted | ||||
|---|---|---|---|---|---|---|---|
APP | 0.799 | 0.803 | 0.870 | 0.627 | |||
CL | 0.799 | 0.799 | 0.869 | 0.624 | |||
DI | 0.711 | 0.716 | 0.820 | 0.533 | |||
SE | 0.809 | 0.815 | 0.874 | 0.635 | |||
There was good to excellent internal consistency in the constructs as displayed by the test of reliability. Cronbach alpha ranged between 0.711 to 0.809 and this is above the recommended minimum value by Nunnally and Bernstein (1994) of 0.70 showing internal consistency of the scale items. Internal consistency reliability Cronbach alpha (0.809) was high and ranged between 0.892 and 0.951, whereas the composite reliability (0.874) was well above the recommended 70 and ranged between 0.820 and 0.874 indicating confirmatory strength of construct reliability greater than Cronbach alpha.
Moreover, the AVE scores of the constructs ranged between 0.533 and 0.635, which is more than the minimum 0.50 as proposed by Fornell and Larcker (1981), which implies adequate convergent validity. It implies that more than fifty percent of the variation in the indicators is because of the latent construct. In particular, AI-Powered Pedagogy (AVE = 0.627), Climate Literacy (AVE = 0.624), Digital Inclusion (AVE = 0.533), and Student Engagement (AVE = 0.635) reflected appropriate convergence of the indicators between them. Cumulatively, these results show that the measurement model possesses the quality psychometrics in an effort to put the constructs as valid and reliable within its future structural model evaluation.
Constructs | APP | CL | DI | SE |
|---|---|---|---|---|
APP | 0.792 | 0.542 (0.680) | 0.336 (0.443) | 0.451 (0.551) |
CL | 0.542 (0.680) | 0.790 | 0.391 (0.505) | 0.434 (0.530) |
DI | 0.336 (0.443) | 0.391 (0.505) | 0.730 | 0.381 (0.493) |
SE | 0.451 (0.551) | 0.434 (0.530) | 0.381 (0.493) | 0.797 |
| Notes: Diagonal values in bold represent the square root of AVE (Fornell–Larcker criterion). | ||||
Below diagonal = Inter-construct correlations. In parentheses = HTMT ratios.
A three-way strategy of the Fornell-Larcker criteria, inter-construct correlations, and the Heterotrait-Monotrait ratio of correlations (HTMT) was used for discriminant validity. Results of Fornell-Larcker (Table X) confirm sufficient discriminant validity in that each construct's square root of the AVE is larger than correlations with the other constructs (APP = 0.792; CL = 0.790; DI = 0.730; SE = 0.797). AI-Powered Pedagogy is more innovative, as its greater AVE square root of 0.792 indicates compared to its relationships with Climate Literacy (0.542), Digital Inclusion (0.336), and Student Engagement (0.451). The inter-construct correlations remain below the acceptable levels (< 0.70) that suggest there are no overlap issues for latent variables. The HTMT ratios strongly fell short of the conservative level of 0.85 (Henseler et al., 2015), ranging from 0.443 (DI ↔ APP) to 0.680 (CL APP). There is additional empirical evidence that the concepts are multicollinear-free and distinct. Results from Fornell-Larcker criteria, correlation matrix, and HTMT ratios confirm AI-Powered Pedagogy, Climate Literacy, Digital Inclusion, and Student Engagement to be highly discriminant valid. This implies that they are theoretically distinguished under the measurement model.
Construct | Item | APP | CL | DI | SE | VIF | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
AI-Powered Pedagogy (APP) | APP1 | 0.850 | 0.438 | 0.291 | 0.381 | 2.315 | ||||||||
APP2 | 0.824 | 0.429 | 0.265 | 0.352 | 1.997 | |||||||||
APP3 | 0.793 | 0.417 | 0.278 | 0.378 | 1.741 | |||||||||
APP4 | 0.691 | 0.432 | 0.225 | 0.312 | 1.352 | |||||||||
Climate Literacy (CL) | CL1 | 0.432 | 0.827 | 0.307 | 0.365 | 2.176 | ||||||||
CL2 | 0.444 | 0.771 | 0.295 | 0.371 | 1.896 | |||||||||
CL3 | 0.414 | 0.746 | 0.318 | 0.307 | 1.846 | |||||||||
CL4 | 0.420 | 0.813 | 0.315 | 0.323 | 2.131 | |||||||||
Digital Inclusion (DI) | DI1 | 0.257 | 0.372 | 0.735 | 0.307 | 1.234 | ||||||||
DI2 | 0.246 | 0.268 | 0.761 | 0.259 | 1.497 | |||||||||
DI3 | 0.244 | 0.251 | 0.724 | 0.297 | 1.416 | |||||||||
DI4 | 0.230 | 0.222 | 0.699 | 0.240 | 1.322 | |||||||||
Student Engagement (SE) | SE1 | 0.322 | 0.318 | 0.289 | 0.830 | 2.336 | ||||||||
SE2 | 0.325 | 0.332 | 0.313 | 0.735 | 1.598 | |||||||||
SE3 | 0.434 | 0.401 | 0.324 | 0.783 | 1.549 | |||||||||
SE4 | 0.331 | 0.311 | 0.279 | 0.835 | 2.252 | |||||||||
Cross-loading analysis verifies that indicators have discriminant validity as they load highly on the targeted construct but provide lower cross-loadings on other constructs. For example, it is verified that AI-Powered Pedagogy is actually a unique construct because APP1's (0.850) and APP2's (0.824) high loadings are higher than their correlation with Climate Literacy, Digital Inclusion, and Student Engagement. Likewise, CL1 (0.827) and SE4 (0.835) both define unique concepts. Severe multicollinearity would be ruled out because the VIF values (1.234–2.336) are way below the stated limit of 5 (Hair et al., 2021). This indicates that there is no overlap of any measurement item's job; rather, each contributes uniquely to its own construct. Such outcomes clearly establish the discriminant validity as well as the reliability of the measurement model, where constructs are kept distinct but indicators are kept intact.
Structural Model of the Study
A
Fig. 2
shows the outcomes of the structural equation modelling (PLS-SEM) that show the connections between AI-based pedagogy practices (APP), digital inclusion (DI), student engagement (SE), and climate literacy (CL), both directly and indirectly. The impact of APP on DI (β = 0.336, p = 0.000) and SE (β = 0.451, p = 0.000) is shown by the path coefficients (β) and significance values (p). This suggests that AI-integrated teaching strategies improve digital accessibility and learner engagement. AI-driven instructional designs play a role in enhancing students' comprehension of climate-related matters, as indicated by the direct beneficial effect of APP on CL (β = 0.202, p = 0.000).
Digital inclusion has strong predictive power over social engagement and social engagement and climate literacy (β = 0.396, p = 0.000; β = 0.188, p = 0.000). This indicates that social and climate engagement is a function of access to and the mastery of technology. Additionally, SE positively influenced CL (β = 0.183, p = 0.001), demonstrating that increased engagement improves climate literacy. The model has R² values of 0.110 for DI, 0.202 for SE, and 0.363 for CL which indicates the model is of moderate to high predictive power and significance. In the higher education context, the model shows that the AI-driven pedagogy used enhances digital inclusion and engagement, thereby improving students' climate literacy.
Path | B | M | SE | t | p | Decision |
|---|---|---|---|---|---|---|
Direct Effects | ||||||
APP → CL | 0.396 | 0.396 | 0.055 | 7.20 | .000 | Supported |
APP → CL (sub-effect 1) | 0.336 | 0.340 | 0.040 | 8.48 | .000 | Supported |
APP → CL (sub-effect 2) | 0.451 | 0.455 | 0.048 | 9.31 | .000 | Supported |
APP → CL (sub-effect 3) | 0.188 | 0.189 | 0.045 | 4.21 | .000 | Supported |
SE → CL | 0.183 | 0.185 | 0.053 | 3.46 | .001 | Supported |
Indirect Effects | ||||||
APP → SE → CL | 0.083 | 0.085 | 0.027 | 3.02 | .003 | Supported |
APP → DI → CL | 0.063 | 0.064 | 0.017 | 3.67 | .000 | Supported |
Table 4 illustrates both direct and indirect effects arising from the proposed structural model. A strong direct relationship was found to exist between AI-Powered Pedagogy (APP) and Climate Literacy (CL) (β = 0.396 to 0.451, p < 0.001) for all parameters of the sub-models. Therefore, teaching tools driven by AI significantly improve students' thoughtful and perception on climate awareness. The significant connection between Student Engagement (SE) and Climate Literacy (β = 0.183, p = 0.001) demonstrates the essential need for active learning and engagement to get effective sustainability-related learning outcomes.
The findings were confirmed through mediation analysis. The SE-mediated effect on APP-CL was statistically significant (β = 0.083, p = 0.003). AI-enhanced classes have the potential to enhance students' comprehension by capturing their attention. The relationship between AI pedagogy and Digital Inclusion (DI) is modestly indirect (β = 0.063, p < 0.001), underscoring the significance of digital tools and information. The findings illustrate the rapid and substantial enhancement of students' climate literacy through AI-driven pedagogical strategies. Moreover, these strategies enhance involvement and interaction, which are essential for bridging the digital divide in Pakistani institutions. Investigations into green education indicate that technologies designed to enhance awareness of climate change are most effective when individuals engage actively and have equitable access to information.
Index | Saturated Model | Estimated Model | Threshold (Cut-off) | (Supported/Not Supported) |
|---|---|---|---|---|
SRMR | 0.075 | 0.087 | ≤ 0.08 (good fit) | Supported |
d_ULS | 0.770 | 1.035 | Closer to 0 better | Supported |
d_G | 0.252 | 0.265 | Closer to 0 better | Supported |
Chi-square | 628.874 | 643.633 | Lower is better | Acceptable |
NFI | 0.750 | 0.744 | ≥ 0.70 acceptable | Supported |
To determine if a model is adequate in relation to the data, model fit indices are computed. The saturated model's SRMR estimate of 0.075 differs significantly from its own estimate of 0.087. Hessler et al. (2016) set some limits, and these do not go beyond them. The discrepancy indices reveal a low degree of separation between the empirical and model-specified covariance matrices (d_ULS = 1.035; d_G = 0.265). According to those indexes, this is correct. A big sample size is required for chi-square statistics, regardless of how well-prepared they are. The model is adequate, as evidenced by NFI scores of 0.75 and 0.744, which are both more than 0.70. The R2 values demonstrate that the model explains a significant portion of the variation in the first dependent variable (36.7%), the second dependent variable (11.3%), and the third dependent variable (20.4%). Three distinct effect sizes were distinguished by Cohen (1988): big, medium, and tiny. These impact sizes are related to the percentages. While the model did its best to account for and match the data, certain structural components performed better in terms of prediction than others.
Findings of the Study
This research offers concrete evidence to strengthen the framework showing the relationship between CL and APP focusing on higher education in Pakistan, leveraging SE and DI as mediating variables. The study confirms the theory underpinning the research due to the documentation of outstanding measurement reliability, discriminant validity, and overall model adequacy. This was achieved through the utilisation of PLS-SEM on data collected from four hundred college students.
The results suggest that there is an improvement in climate literacy when AI instructional methods are used (β = 0.396, p < 0.001). This further reiterates that AI learning technologies are transformative in developing an awareness of sustainability and an appreciation of the natural world. AI educational technologies have the potential to improve climate education by facilitating the most educational technologies used for climate education focus on enhancing comprehension and encouraging students to take an active role in their learning. It has been recognized that student engagement in artificial intelligence instructional environments is significant (β = 0.083, p = 0.003). This indicates that artificial intelligence (AI) schools enhance climate literacy not by using passive learning techniques, but by fostering active learning via cooperation. These results indicate that learning theories emphasizing student agency and collaborative effort in the classroom are congruent with the findings.
Students can only learn effectively in climate education if they have equal access to digital materials and tools that use AI. This is affected by the effect of digital inclusion (β = 0.063, p < 0.001). Because of the systemic unfairness in Pakistani schools, digital inclusion is not just a technical need. It is also a social and educational one. AI teaching has a bigger effect on climate literacy because it makes it easier for people to receive and participate in learning, which are both important for making learning fairer and more effective.
This evaluation technique was psychometrically sound. Internal consistency was obvious with Cronbach's alphas 0.711–0.809 and composite reliability scores of 0.820. Discriminant validity was confirmed using Fornell-Larcker and HTMT criteria. Convergent validity is indicated by an AVE value of 0.50 or above. It established the portions were conceptually separate. The structural model's empirical strength and explanation are supported by significant discrepancy measures (dULS = 1.035; dG = 0.265) and fit indices (SRMR = 0.087; NFI = 0.744).
These results show that the digital divide and gaps in AI-inclusive and participatory pedagogy make educational frameworks stronger. This is a good sign for the SDGs that have to do with climate action and good education. AI integration into climate education may significantly improve the education delivery system by enhancing learning results, motivation, and equity.
This research contributes to constructivism, digital equality, and self-determination by showing how AI systems promote independence, competence, and belonging throughout learning. The research suggests that educators and policymakers should promote equitable access to AI tools, faculty competence in AI teaching approaches, and inclusive AI platforms that enable the deliberate integration of AI in climate literacy education.
The findings contribute to the existing literature by demonstrating the capabilities of AI to enhance digital literacy in developing countries and, most importantly, the gaps in teaching climate literacy and digital inequity in low-income countries. Therefore, climate education is positioned to undergo transformational changes for the better through educators’ focusing on equitable AI-enabled teaching strategies and innovative instructional materials.
Discussion of the study
Using student engagement (SE) and digital inclusion (DI) as mediators, the empirical results obtained from the PLS-SEM analysis provide strong evidence supporting the proposed model, showing that AI-powered pedagogy (APP) significantly influences climate literacy (CL) of university students in Pakistan. Within the theoretical frameworks of constructivism, digital equality, and self-determination, the findings corroborate and improve upon recent developments in technology pedagogy. In addition, these results provide light on how to innovatively use technology in the classroom to make classrooms more welcoming and environmentally friendly or all students.
Both the conceptual validity and the psychometric quality of the measuring model were strong. The consistency of the indicators was demonstrated by the fact that all but one external loading—a respectable 0.691—met the criterion of 0.70 (Hair et al., 2021). The constructs that have shown an adequate level of internal consistency in the stability and internal consistency evaluations (as demonstrated by composite reliabilities > 0.82 and Cronbach's alphas ranging from ≤ 0.711 to ≤ 0.809) are worthy of further consideration. According to the standards laid out by Fornell and Larcker (1981), convergent validity is guaranteed when the AVE values for converged constructions fall between 0.533 and 0.635, provided that all of these values are greater than the necessary threshold of 0.50.
The study's theoretical framework was validated since the structural model's links were strong and statistically significant. The predicted model foretold the relationship between climate literacy and AI-enhanced pedagogical approaches. The use of AI in the classroom greatly improves students' understanding and knowledge of climate literacy (β = 0.396, p < 0.001). Together with other instructional resources, we argue that AI that is smart, adaptable, and adaptive can improve students' deep learning. The learning outcomes are in line with the ideas of the constructivist learning paradigm, which highlights the significance of learning via experience and inquiry. According to El Fathi et al. (2025) and Tan and Maravilla (2024), students' reflective creation of knowledge is demonstrated by the cognitive engagement that arises from the application of AI in education.
Instruments for student engagement and digital inclusion highlight the potential link between AI and climate literacy, but not as emphatically as hoped. The integration of AI in education significantly improves learner engagement, curiosity, and involvement, hence enabling a more thorough investigation of sustainability-related information, as indicated by the indirect association through student engagement (β = 0.083, p = 0.003). This finding makes sense if you believe in Self-Deter. Digital inclusion acts as a mediator that stresses the need of giving everyone equal access to technology (β = 0.063, p < 0.001).
Research by Zahid et al. (2025) indicates that pupils in urban schools with superior resources were more proficient in using AI for academic pursuits compared to their counterparts in rural or under-resourced institutions. By 2025, Rehman and Khan. It is possible that AI-driven educational advances may exacerbate inequality if infrastructure upgrades are not implemented. Digital inclusion is essential from both a pedagogical and ethical standpoint in order to raise the level of climate awareness across all student groups.
Indices of model fit indicate the proposed model is adequate and maintains a good fit. Thus, the NFI is 0.744 and the SRMR is 0.087, which are adequate. The disparity indices add that there is a good correspondence between the observed and model-implied covariance matrices, with d_ULS recorded at 1.035 and d_G at 0.265. A Climate literacy performance was strong at 36.7%, with digital inclusion accounting for 11.3% and student involvement at 20.4%. Cohen (1988) maintains this. As a result, this research demonstrates the importance of AI pedagogy in understanding student involvement and belonging, as well as the consequences of climate literacy.
These findings confirm that the inclusion of AI in educational contexts is based on the theories of constructivism, digital equality, and self-determination. Constructivist theory posits that learning occurs best when collaboration, active participation, and reflective practice take place. Tan and Maravilla (2024) note that adaptive algorithms and interactive data utilized by AI-powered platforms further constructivist educational ideas such as inquiry-based learning and personalized discovery. Digital equality refers to a situation where there is an equal distribution of technology resources. In context, it analyzes ethical and structural issues relating to educational technology. This study tries to validate with empirical evidence a theoretical proposition of digital inclusion as a mediator. The result of the study shows that equal access influences the performance of learners within technology-enhanced learning environments.
The results of this study provide evidence that AI pedagogy enhances autonomy, competence, and relatedness, which, according to Self-Determination Theory, are the three salient factors that enhance intrinsic motivation. It allows students to have more choice over the pace at which they learn through adaptive AI technologies, personalized feedback in developing their skills, and access to AI-facilitated digital communities, enhancing their sense of relatedness. The synthesis of motivational theory with digital pedagogy represents a theoretical advance in understanding how AI technologies may sustain student motivation and build lasting engagement in climate action.
Implications of the study
The implications are practical for administration and policy-making in higher education. Results show that AI-enhanced pedagogies can help reduce knowledge inequities and foster learning about sustainability, but only where adequate digital infrastructures and equitable access exist. This means that universities will need to invest in both AI-enhanced learning management systems and inclusive networks of digital resources that provide equitable access to the technology behind AI. Of course, it will also be important to make changes to the curriculum to include climate literacy in courses reaching across disciplines, allowing students to learn and develop critical thinking abilities in concert with civic responsibility and policy formulation. Results indicate that AI-enhanced pedagogies may successfully address knowledge inequities and foster sustainability-focused learning, provided there is sufficient digital infrastructure and equity of access. Equitable access to AI technology will require universities to invest in AI-enhanced learning management systems, in faculty development programs, and in inclusive networks of digital resources. Changes to the curriculum also will be necessary to ensure that climate literacy in courses across disciplines contributes to students learning and developing critical thinking skills while learning to be civically responsible.
This is something for which educators at a national level should strive: making digital equality an integral part of their sustainability education programs. Promotion of open-access AI educational platforms, universal internet access, and providing digital resources to underprivileged kids is what they must focus on. Higher education institutions might have a potentially important role in helping achieve two United Nations Sustainable Development Goals SDG 4: Quality Education and SDG 13: Climate Action by combining AI pedagogy with climate education objectives. The aim would be to ensure that technology contributes toward fair and sustainable development.
Conclusion
The study concluded that that classroom methods applied in an AI tools classroom can be potentially useful. AI can support the expansion of knowledge about climate change. To accomplish this, we employ engaging yet comprehensive methods that foster direct cognitive engagement and emphasize the overall improvement. The results exceed the objectives of the research to further enhance the theoretical discourse at the level of the discourse community involving constructivism, AI, and teaching/learning for sustainability. This paper presents the following conclusion: the interaction of climate education with the digital world calls for innovative teaching and learning approaches that focus on social equity and accessibility.
A
Funding
This research received no external funding.
Informed Consent
Statement
Informed written consent was obtained from all participants involved in this study.
A
Data Availability
The data that support the findings of this study are available from the corresponding author, upon reasonable request. The data are not publicly available due to privacy and ethical restrictions.
The data that support the findings of this study are available from the corresponding author, upon reasonable request. The data are not publicly available due to privacy and ethical restrictions.
Conflicts of Interest
The authors declare no conflicts of interest.
Competing Interests
The author declares no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
A
A
Author Contribution
Researchers conceptualized the study, designed the methodology, conducted the analysis, and wrote the manuscript. The authors approved the final version of the manuscript.
Acknowledgements
The authors sincerely thanks the participating students and institutions for their valuable contributions to this research.
References:
Anderson A. Climate change education for mitigation and adaptation. J Educ Sustainable Dev. 2012;6(2):191–206. https://doi.org/10.1177/0973408212475199.
Chen L, Chen P, Lin Z. Artificial intelligence in education: A review. IEEE Access. 2020;8:75264–78. https://doi.org/10.1109/ACCESS.2020.2988510.
Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed. Lawrence Erlbaum Associates; 1988.
Elgaronline. Generative AI in higher education. Handbook on AI in Education. Edward Elgar Publishing; 2023. https://doi.org/10.4337/9781035326020.00020.
Fornell C, Larcker DF. Evaluating structural equation models with unobservable variables and measurement error. J Mark Res. 1981;18(1):39–50. https://doi.org/10.1177/002224378101800104.
Fredricks JA, Blumenfeld PC, Paris AH. School engagement: Potential of the concept. Rev Educ Res. 2004;74(1):59–109. https://doi.org/10.3102/00346543074001059.
Hair JF, Hult GTM, Ringle CM, Sarstedt M. A primer on partial least squares structural equation modeling (PLS-SEM). 2nd ed. Sage; 2019.
Henseler J, Ringle CM, Sarstedt M. A new criterion for assessing discriminant validity in variance-based structural equation modeling. J Acad Mark Sci. 2015;43(1):115–35. https://doi.org/10.1007/s11747-014-0403-8.
Holmes W, Bialik M, Fadel C. Artificial intelligence in education: Promises and implications for teaching and learning. Center for Curriculum Redesign; 2021.
Kahu ER. Framing student engagement in higher education. Stud High Educ. 2013;38(5):758–73. https://doi.org/10.1080/03075079.2011.598505.
Luckin R, Holmes W, Griffiths M, Forcier LB. Intelligence unleashed: An argument for AI in education. Pearson; 2016.
McCaffrey MS, Buhr SM. Clarifying climate confusion: Addressing systemic holes, cognitive gaps, and misconceptions through climate literacy. Phys Geogr. 2008;29(6):512–28. https://doi.org/10.2747/0272-3646.29.6.512.
Monroe MC, Plate RR, Oxarart A, Bowers A, Chaves WA. Identifying effective climate change education strategies: A systematic review of the research. Environ Educ Res. 2019;25(6):791–812. https://doi.org/10.1080/13504622.2017.1360842.
Nunnally JC, Bernstein IH. Psychometric theory. 3rd ed. McGraw-Hill; 1994.
Piaget J. The psychology of the child. Basic Books; 1972.
Ryan RM, Deci EL. Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. Am Psychol. 2000;55(1):68–78. https://doi.org/10.1037/0003-066X.55.1.68.
A
Sarstedt M, Ringle CM, Hair JF. (2020). Partial least squares structural equation modeling. In H. J. Albers, editor, Handbook of market research (pp. 1–47). Springer. https://doi.org/10.1007/978-3-319-05542-8_15-1Tan MJT, Maravilla NMAT. Shaping integrity: Why generative artificial intelligence does not have to undermine education. Front Artif Intell. 2024;7:1471224. https://doi.org/10.3389/frai.2024.1471224.
UNESCO. Reimagining our futures together: A new social contract for education. UNESCO Publishing. 2022. https://doi.org/10.54675/NCRY1616.
Van Dijk J. The digital divide. Polity; 2020.
Vygotsky LS. Mind in society: The development of higher psychological processes. Harvard University Press; 1978.
Wals AEJ, Peters MA. Sustainability education and the role of universities. Educational Philos Theory. 2018;50(2):111–3. https://doi.org/10.1080/00131857.2016.1254884.
Warschauer M. Technology and social inclusion: Rethinking the digital divide. MIT Press; 2003. https://doi.org/10.7551/mitpress/6699.001.0001.
Zawacki-Richter O, Marín VI, Bond M, Gouverneur F. Systematic review of research on artificial intelligence in higher education—Where are the educators? Int J Educational Technol High Educ. 2019;16:39. https://doi.org/10.1186/s41239-019-0171-0.