The era of Society 5.0 has brought about significant changes across various sectors. This rapid progression of technology and science in the smart society era fundamentally alters how pedagogical frameworks operate globally (Susilo et al., 2022). Consequently, modern educators face intense pressure to transform conventional methodologies into technology-integrated environments that prepare students for automated societal tasks (Suratmi & Sopandi, 2022). These changes require individuals to adapt by mastering essential twenty-first-century competencies, including critical thinking within the 4C framework. Critical thinking refers to a person’s cognitive process in addressing a problem through interpretation, analysis, explanation, evaluation, inference, and self-regulation (Putri et al., 2021). This high-level mental ability is essential for students to make

reflective, logical decisions when facing complex data (Mustikasari, 2022). In education, critical thinking skills enable students to solve complex problems more accurately (Hamedani et al., 2024). Recent bibliometric evidence by (Khusna et al., 2024) also emphasizes that critical thinking has become a leading focus in mathematics education, underscoring its importance in twenty-first-century learning. This academic consensus shows that improving classroom models is essential for systematically developing students' reasoning and analytical skills (Rensburg & Rauscher, 2022). Therefore, critical thinking has become a crucial competency that students must possess to navigate ongoing societal changes.

Despite being recognized as an essential competence, students’ critical thinking skills in Indonesia remain relatively low (Ramlawati et al., 2025). According to the Program for International Student Assessment (PISA), students who demonstrate strong critical thinking skills are typically classified at Level 4 or above. However, from 2000 to 2022, the proportion of Indonesian students achieving this level has never exceeded 5%. More specifically, the detailed percentages are presented in Table 1. This percentage even reached its lowest point in 2022, at approximately 0.5%. This figure is well below the OECD average of 14.9% for high-performing students (O.E.C.D., 2023). This low percentage is due to learning that is insufficiently focused on developing critical thinking skills.

One model that supports students’ development of critical thinking is Challenge-Based STEM Context Learning, a learning model that presents challenges and problems for students with STEM focused content (Putri & Alyani, 2023). By integrating structured projects and context-rich problems, this approach effectively enhances students' higher-order thinking (Usman, 2024). Additionally, differentiated learning offers varied learning pathways, allowing students with different abilities to reach the same learning goals (Masinading & Gaylo, 2022). This model acknowledges that adapting instructional methods to specific student profiles is critical, as diverse initial skills significantly dictate how learners process complex tasks (Sutama, 2022). This relevance is supported by (Shamim et al., 2022), who found that STEM-context pedagogical beliefs are increasingly emphasized as a foundation for developing higher-order thinking skills. Blending digital environments with tailored instructions enhances math skills and reduces cognitive barriers (Nida et al., 2020). As a result, this combination provides a flexible instructional foundation that can be further enhanced by integrating adaptive digital tools. Therefore, using realistic, contextualized situations is vital for enhancing students' math problem-solving skills in modern classrooms (Nugraheni & Marsigit, 2021).

Building on these digital tools, an AI-Website becomes an adaptive innovation that can further facilitate the development of students’ critical thinking skills. AI-Website learning supports the development of critical thinking by providing alternative learning pathways to enhance students’ learning abilities (Wu et al., 2025). The structural optimization of integrated web learning platforms substantially accelerates cognitive transformation by encouraging students to dynamically reconstruct visual math representations (Cevikbas & Kaiser, 2022). Website integration also enables the development of flexible features aligned with various learning objectives. However, implementing an AI-Website requires an appropriate learning model to ensure optimal impact on students’ critical thinking skills (Omer et al., 2025). A systematic review by (Melisa et al., 2025) also concludes that AI integration in higher education can enhance critical thinking, but its impact heavily depends on the learning model used. AI development has shifted many cognitive tasks to digital tools, making critical thinking an even more vital human skill in the Society 5.0 era. As software systems easily automate complex arithmetic computations, modern educational design must prioritize the cultivation of non-mechanized, reflective human reasoning to drive sustainable classroom innovation (Tehlan, 2025). Thus, integrating a differentiated, challenge-based STEM context with an AI website could be an innovation that enhances students’ critical thinking skills. In addition to these several factors, there are also internal factors, such as affective factors, that can influence students’ critical thinking skills.

One of the affective factors influencing students’ critical thinking skills is the Adversity Quotient (AQ) (Astiantari et al., 2022). According to (Stoltz, 1997), Adversity Quotient is an individual’s ability to respond to, endure, and overcome difficulties, measured through the four CORE dimensions: control over situations (Control), responsibility for problems (Ownership), the extent to which difficulties affect other aspects (Range), and endurance in facing obstacles (Endurance). In modern educational psychology, these psychological dimensions are highly structural in defining whether a student will actively search for alternative problem-solving pathways or completely abandon the task (Rahmadani et al., 2025). In the learning context, AQ is related to students’ persistence in dealing with complex problems and their tendency not to give up easily. Therefore, AQ can be one of the factors examined in this research to improve critical thinking skills. Consequently, integrating this affective evaluation within a technology-enhanced framework provides a more holistic analysis of the actual mechanics behind student learning development (Chiu, 2022).

Although in theory, AQ is one of the factors that can influence the improvement of critical thinking skills, there has not been much research examining the improvement of critical thinking skills using AQ as one of the variables (Manousou, 2025). Research integrating AI-Websites, CBL-STEM Context and differentiated instruction are still mostly conducted separately, and no studies have examined their integration, resulting in a gap in this topic. Therefore, this study aims to address the research gap by analyzing how students’ AQ influences the improvement of critical thinking skills in a differentiated CBL-STEM Context learning integrated with an AI-Website.

Based on this background, the present study was designed to: (1) measure the effectiveness of differentiated challenge-based STEM Context learning integrated AI-Website in improving students’ critical thinking skills; (2) analyze the influence of AQ on students’ critical thinking skills; and (3) describe students’ critical thinking based on their Adversity Quotient categories within mathematics learning. This study is expected to provide an initial overview of the potential of integrating an AI-Website into the CBL-STEM Context, differentiated learning to strengthen students’ critical thinking skills in responding to the challenges of the Society 5.0 era.

This study shows that differentiated CBL-STEM Context learning integrated with AI-Website effectively enhances students' critical thinking skills. Through the CBL model, students are supported in analyzing and evaluating contextual problems both independently and collaboratively. The statistical results strengthen this conclusion, with the experimental group achieving an average posttest score of 82.91, a 75.6% increase in critical thinking skills, and a learning completion rate of 86.67%. Comparative tests further confirmed the superiority of the experimental group over the control group. These findings align with previous studies showing that AI-Website and STEM- Context learning models systematically promote higher-order thinking skills, including critical thinking. This result is consistent with (Zhu et al., 2021), who demonstrated through a randomized experiment that CBL-STEM Context learning significantly enhances students’ higher-order thinking skills. These findings also align with (Zabian, 2025) who found that AI-Website learning systems can improve academic mastery through adaptive, structured learning. The architectural integration of such intelligent web-based tutoring models systematically optimizes individual student performance by deploying advanced neural algorithms to predict learning behavior and adjust content pathways accordingly (Gomede et al., 2021). In line with this, (Panggabean et al., 2025) demonstrated that a validated Canva-IBL-STEM model significantly enhances higher-order thinking skills, particularly in analytical and problem-solving dimensions. This empirical correlation highlights that the strategic implementation of structured digital learning systems serves as an essential pathway to systematically accelerate and elevate students' higher-order cognitive performances (Mat et al., 2024). These findings support the argument that integrating STEM with digital learning platforms can systematically foster students’ higher-level cognitive processes, including critical thinking. Thus, the learning success in this study is consistent with recent empirical evidence on the effectiveness of collaboration among CBL-STEM Context, AI-Website, and differentiated learning in developing students' critical thinking skills. By synthesizing these three educational innovations, the blended environment successfully orchestrates a comprehensive learning ecosystem where student autonomy and cognitive growth are nurtured concurrently (Chiu, 2021).

Interview results revealed differences in the mastery of critical thinking indicators across AQ categories. All groups achieved the basic indicators of interpretation and analysis, but variations appeared in explanation, evaluation, inference, and self-regulation. Climber students generally demonstrated stronger mastery of these higher-level indicators than Campers and Quitters. Previous studies also support this pattern, indicating that students with higher AQ tend to solve problems more accurately and use more logical reasoning than those with lower AQ. This pattern is also supported by (Zhao & Sang, 2023), who found that AQ contributes to cognitive performance, but its influence remains strongly shaped by the surrounding learning conditions. Which indicates that systematic reviews show that the contribution of AQ to critical thinking skills is often small and depends on other variables, such as learning strategies and environmental support. Other studies also show that although students with high AQ tend to demonstrate better learning performance, the effect of AQ does not stand alone and is still shaped by various learning conditions (Anggraini & Mahmudi, 2021). Thus, differentiated challenge-based STEM Context learning integrated AI-Website can bridge the AQ gap, enabling all students to achieve relatively equal critical thinking skills.

Variations in N-gain scores were also influenced by learning preferences across the three AQ categories: Climbers, Campers, and Quitters. Although AQ’s impact on critical thinking is relatively small, it still affects how students respond to instructional support. The use of AI-Website and differentiated CBL-STEM Context learning provided structured scaffolding that addressed each group’s needs. Climbers favored AI integration and differentiated tasks because they offer advanced challenges and opportunities for exploration. Campers preferred the gradual structure of STEM Context learning combined with differentiation, which helped them connect ideas more effectively. Quitters depended on structured, differentiated learning that offers step-by-step guidance and reinforces prerequisite knowledge. This AI-Website scaffolding helped both Campers and Quitters exceed expectations in critical-thinking indicators based on their AQ profiles. A similar tendency was reported by (Pramuditya et al., 2022), who showed that structured support within an open-ended learning environment significantly improves students’ mathematical problem-solving abilities. According to the qualitative study by (Putri & Alyani, 2023), Participants with the climber type can solve problems more accurately. In contrast, those with the quitter type tend to meet only the interpretation indicator. Students with high AQ were better able to reason and provide logical-mathematical explanations than students with low AQ when solving non-routine problems. This operational difference reflects a core pattern where resilient learners instinctively deploy more effective self-regulated metacognitive strategies when untangling non-routine tasks (Veenman, 2021). Overall, the findings of this study indicate that the higher a student's AQ, the more complete the indicators of mathematical critical thinking they can achieve in the problem-solving process. However, the tactical integration of digital technology helps reconstruct how students approach challenging tasks, allowing vulnerable learners to effectively emulate advanced analytical pathways (Attard & Holmes, 2020). The use of data-driven predictive frameworks, as highlighted by (He et al., 2025), further underscores the capacity of adaptive technologies to optimise learning pathways in mathematics education.

The results of this study show that differentiated CBL-STEM Context learning integrated with an AI-Website effectively improves students’ critical thinking skills while also reducing the limiting effect of low AQ on these outcomes. This addresses the research gap, as prior studies didn't combine AI-Website, CBL-STEM Context, and differentiated learning in a single framework. AI-Website learning boosts students’ critical thinking and interacts positively with AQ by offering adaptive support beyond traditional methods. Qualitative data also reveal that students across all AQ categories, Climbers, Campers, and Quitters, reached various critical thinking levels. Even students with lower AQ scores mastered indicators through scaffolding provided by differentiated tasks, challenges, and AI pathways. These findings align with (Nasir et al., 2022), who reported that STEM guided inquiry improves students’ conceptual understanding and scientific explanation skills across varying levels of ability. This instructional alignment demonstrates that when digital scaffolding is paired with variable task structures, it effectively broadens cognitive access, allowing less resilient or lower-achieving groups to successfully engage with advanced problem-solving phases (Setyaningrum et al., 2024). These outcomes also surpass those typically seen in traditional classrooms, where students with low AQ often meet only basic critical-thinking standards. Emphasizes that although AQ influences critical thinking, its impact is relatively small and heavily dependent on the instructional strategies used. Therefore, models that offer adaptive, differentiated support, like here, help close performance gaps among diverse AQ profiles. This research addresses the unexamined integration gap and shows that AI-supported, differentiated CBL-STEM Context learning can promote fairer development of critical thinking skills across all AQ categories.

Overall, differentiated CBL-STEM Context learning integrated with an AI-Website proved effective in improving critical thinking skills and in reaching all AQ categories. Although the influence of AQ was relatively small, this shows that differentiated learning models can accommodate learners with diverse backgrounds and initial abilities. These findings reinforce the conclusion that integrating AI into differentiated CBL-STEM Context learning is a relevant innovation for comprehensively improving critical thinking skills.

This study concludes that the integration of differentiated CBL-STEM Context integrated with AI-Website instruction is effective in improving students’ critical thinking skills. This effectiveness is demonstrated by significantly higher post-test scores, stronger N-Gain results, and better mastery of KKTP in the experimental group compared to the control group. Furthermore, findings indicate that the Adversity Quotient (AQ) has an impact on learning at the lower level, suggesting that other factors influence the development of critical thinking skills among students. Interviews revealed distinct patterns in the fulfillment of critical thinking indicators. Students in the higher AQ category tended to consistently fulfill critical thinking indicators, while students with lower AQ initially struggled to meet some indicators. All subjects tend to be able to consistently perform well on interpretation and analysis indicators, but differences appear in indicators such as evaluation and inference. This was found to be due to differences in learning preferences, where students in the Climber group were better able to adapt to the learning process because it aligned with the AI-Website and the gradually increasing level of difficulty, while the Camper group preferred learning that linked the material to everyday life or other topics through CBL-STEM Context; whereas the Quiter group itself tends to rely more on scaffolding through differentiated learning, because the learning still focuses on improving critical thinking skills based on their initial abilities, causing the Quiter group to have some difficulty adapting to the learning process. Therefore, in general, differentiated CBL-STEM Context learning integrated with an AI-Website is effective in enhancing students' critical thinking skills.

Research focused on improving critical thinking skills in general found that subjects in the low AQ category did not meet as many indicators as those in other categories. Therefore, future research is expected to integrate differentiated learning by AQ categories. The small sample size is a limitation of this study; therefore, it is hoped that future research will be conducted on other populations to enrich existing findings by integrating differentiated CBL-STEM context-based learning with an AI-website based on students’ AQ.

We want to thank the Directorate of Research and Community Service (DPPM) of the Ministry of Education, Culture, Research, and Technology for funding this research, and we would also like to thank SMP Negeri 30 Semarang, our partner, for making this research possible.