Introduction

Competency in constructing scientific explanations is essential in science learning. It enhances reasoning skills, strengthens argumentation, and deepens conceptual understanding(McNeill & Krajcik, 2006). This competency involves the ability to comprehend, select, and apply scientific knowledge to develop rational explanations for natural phenomena (Hundelhausenet al., 2024). An effective scientific explanation integrates three key components. It begins with a claim that addresses a scientific question or phenomenon, supported by relevant and sufficient evidence from credible scientific data. Finally, reasoning employs scientific principles to logically connect this evidence to the claim (McNeill & Krajcik, 2008). Blending these components forms the foundation for constructing scientific explanations that empower students to interact with scientific ideas and demonstrate their understanding through coherent explanations.

The Programme for International Student Assessment (PISA) evaluates three core scientific competencies, including the competency to explain phenomena scientifically, a key focus of the Organisation for Economic Co-operation and Development (OECD, 2025)’s 2025 assessment framework. Constructing scientific explanations is a key indicator of students' ability to apply scientific knowledge, making it a fundamental criterion in PISA’s evaluation of scientific literacy. The integration of making claims, using evidence, and applying reasoning provides a structured approach to assess how effectively students apply scientific knowledge to real-world phenomena.

PISA assesses 15-year-old students worldwide, including those from OECD and partner countries such as Thailand.The 2022 PISA science assessment results revealed that Thai students scored below the OECD average, ranking 58th out of 81 participating countries (OECD, 2023). These results suggest that many Thai students face challenges in applying scientific knowledge to construct explanations of real-world phenomenaand struggle to connect theoretical concepts with authentic situations.These findings underscore the urgent need to enhance science teaching approaches that focus on strengthening students' competency to construct scientific explanations.

In Thailand, education continues to rely heavily on traditional teaching methods that emphasize rote memorization rather than encouraging students to engage in analytical thinking and constructing scientific explanations (Sukumal, 2024; Wannathai & Pruekpramool, 2024). This reliance is a significant challenge in the development of students’ competency in constructing scientific explanations. To overcome the limitations of traditional teaching, phenomenon-based learning has been introduced as an effective approach to connecting scientific concepts with real-world contexts (Symeonidis & Schwarz, 2016). Phenomenon-based learning emerges as a relevant instructional strategy, commencing with authentic, real-world phenomena to systematically support students in developing their competency in constructing scientific explanations (Islakhiyah et al., 2017). This approach encourages students to build a more coherent understanding by integrating knowledge across disciplines and actively engaging in the learning process through investigation, hands-on activities, and collaborative discussions (Symeonidis & Schwarz, 2016). Such engagement fosters critical thinking, scientific reasoning, and interdisciplinary connections, making phenomenon-based learning a valuable tool for enhancing students' scientific explanation competencies (Adipat, 2024; Akkas & Eker, 2021; Penuel et al., 2019).

The Role of Generative AI in Education

Beyond pedagogical strategies like phenomenon-based learning, the integration of emerging technologies such as Generative Artificial Intelligence (Gen AI) presents new opportunities for education (Jauhiainen & Guerra, 2023, 2024; Mittal et al., 2024). This technology can generate content based on user prompts via a chat interface (Balasubramaniam et al., 2024; Zhou & Lee, 2024). One of its notable roles in education is helping students find relevant sources, articles, and documents within a short time (Lee et al., 2023; Rasul et al., 2023). Gen AI also helps overcome language barriers, as many students rely on Thai-language references, which limits their access to a wide range of academic resources (Deng et al., 2025).

Beyond these general applications, GenAI offers significant potential in science education, particularly for competency in constructing scientific explanations. These tools can assist students in accessing and processing complex scientific information and synthesizing data from diverse sources, thereby fostering a deeper understanding essential for developing robust scientific explanations (Zapata-Rivera et al., 2024). Moreover, Gen AI can provide students with illustrative examples of well-structured scientific explanations (Cooper, 2023). This exposure can help learners grasp the critical components of a scientific explanation, such as claims, evidence, and reasoning, and guide them in developing their competency in constructing scientific explanations (Hsu et al., 2015; Zhang et al., 2006).

For this study, Microsoft Copilot was selected as the specific Gen AI tool based on several key advantages for the educational context. First, its free version provides high-quality, referenced information from credible academic sources, making it a reliable tool for scientific inquiry (Gupta et al., 2025). Second, it is highly efficient in systematically summarizing and organizing information, a feature that directly supports students in developing their competency in constructing scientific explanations. Finally, its native integration with Microsoft 365 applications (such as Word, PowerPoint, and Excel), which are widely used in educational settings, enhances its practical usability for classroom implementation.

Research Rationale and Objectives

Although previous studies have explored phenomenon-based learning and the use of Gen AI in education, there has been no direct investigation into the combined effects of Microsoft Copilot and phenomenon-based learning on developing competency in constructing scientific explanations. Furthermore, there is a notable gap in understanding the specific factors that influence the development of this competency when Gen AI tools are integrated with phenomenon-based learning approaches. Therefore, this research aims to address these gaps by implementing Microsoft Copilot within the phenomenon-based learning framework, specifically during the third step–Investigation. In this step, Microsoft Copilot will be utilized as a tool for investigating problems related to predetermined phenomena. This study has two primary objectives:

1) to compare eighth-grade students' competency in constructing scientific explanations before and after implementing phenomenon-based learning with Microsoft Copilot.

2) to examine the key factors that influence the development of this competency when phenomenon-based learning is integrated with Microsoft Copilot.

This research is expected to contribute to the field in three key ways. First, it provides the first empirical evidence on the effectiveness of integrating Microsoft Copilot with phenomenon-based learning for developing scientific explanation competency. Second, it identifies specific factors that influence this competency development in a Gen AI-enhanced learning environment. Third, it offers evidence-based guidelines for educators on implementing AI tools within structured pedagogical frameworks.These contributions advance both theoretical understanding and practical implementation of Gen AI-assisted science education.

Methodology

Research Design

This study employed a mixed-methods quasi-experimental design with one-group pretest-posttest methodology. Quantitative data measured changes in students' competency in constructing scientific explanations, while qualitative data provided insights into the learning process. This approach was selected to comprehensively evaluate the effectiveness of phenomenon-based learning integrated with Microsoft Copilot and to identify factors influencing the development of students' competency in constructing scientific explanations.

Data collection involved a mixed-format questionnaire to capture students’ perceptions, reflections, and experiences with the integrated approach ofphenomenon-based learning and Microsoft Copilot. Additionally, a competency test was used to quantitatively assess students' improvement in constructing scientific explanation competency. The questionnaire responses were analyzed using thematic analysis, while the competency test scores were statistically examined to determine learning gains. The research was conducted during the second semester of the 2024 academic year, with the study spanning a total of 6 weeks (3 hours/week).

Research Context and Participants

This study was conducted at a government-funded all-girls secondary school in Bangkok, Thailand, and was targeted at eighth-grade students during the 2024 academic year. At this grade level, there were a total of five classrooms, comprising 105 students, who were assigned to their classes through a mixed-ability grouping system. An analysis of the average academic achievement scores from the first semester of the 2024 academic year shows that students in all five classrooms demonstrated similar levels of academic achievement.

The study focused on one purposively selected classroom with 23 students. This specific classroom was chosen because all students had their own smartphones, an internet connection, and were ready to install the Microsoft Copilot application on their personal devices. These factors were important for integrating Microsoft Copilot into phenomenon-based learning as a key component of the study.

Lesson Content and Instructional Process

The lesson content covered five major topics related to Earth and its Changes, including Earth's structure and changes, weathering processes, soil formation, water resources, and fossil fuels. These topics were integrated into a phenomenon-based learning approach with Microsoft Copilot, following a structured five-step instructional design that was adapted from the research of Islakhiyah et al. (2017), detailed in Appendix A (Table A1). The lesson framework was adapted from the basic science and technology textbook by Khamcheensri et al. (2023). This textbook was selected as it is designed in accordance with the 2017 revision of the Thailand Basic Education Core Curriculum (Ministry of Education Thailand, 2017).

Research Instruments

Lesson Plans for Phenomenon-Based Learning with Microsoft Copilot

This research developed five lesson plans on Earth and Its Changes, integrating phenomenon-based learning with Microsoft Copilot. These plans, designed according to phenomenon-based learning principles and incorporating Microsoft Copilot as an instructional tool, followed a five-step instructional framework (see Appendix A). To ensure quality and validity, the lesson plans were evaluated by a panel of three experts in science education and technology integration, who assessed them on criteria such as curriculum alignment and instructional design, providing both quantitative and qualitative feedback. The evaluation results demonstrated exceptionally high appropriateness of the lesson plans, with a mean score of 4.86 (SD = 0.24).Following modifications based on expert recommendations, a feasibility study was conducted with aclassroom of eighth-grade students who were not part of the main research sample. This pilot study aimed to assess (a) the practicality of the instructional design, (b) students' engagement and interaction with Microsoft Copilot, and (c) the accuracy and appropriateness of the Microsoft Copilot-generated content. Observational data and student feedback were collected to further refine the lesson plans before full implementation.

Competency Test for Constructing Scientific Explanations

A competency test for constructing scientific explanations, comprising five constructed-response items based on real-world phenomena, was developed. Aligned with McNeill and Krajcik's (2008) framework, the items required students to formulate claims, provide evidence, and use reasoning. Each item was scored using a rubric assessing these three components (2 points maximum per component), yielding a total maximum score of 30 points.

The test underwent a systematic validation process. Initially, ten items related to Earth and Its Changes were drafted. Three science education experts then evaluated these items for content relevance, clarity, and alignment with McNeill and Krajcik’s framework. Items demonstrating high content validity, as indicated by an Index of Item-Objective Congruence (IOC) value of ≥ 0.67, were selected for further refinement.

Following expert feedback, the revised instrument was pilot-tested with a comparable group of eighth-grade students (not part of the main study). This pilot testing enabled psychometric analysis, including determination of difficulty indices (ranging from 0.50 to 0.65), discrimination indices (ranging from 0.60 to 0.80), and reliability (calculated at 0.89). Based on these analyses, five items with optimal psychometric properties were selected for the final instrument administered to the main study group.

A Mixed-Format Questionnaire

A mixed-format questionnaire, comprising nine items(labeled Q1–Q9), was developed to explore students’ experiences with Microsoft Copilot in the phenomenon-based learning environment (see Appendix B for the full questionnaire). The instrument integrated various question types to capture a comprehensive range of student experiences. Initially, two multiple-choice questions assessed students' prior use of Microsoft Copilot and their perceived learning improvement (Q1–Q2). Following this, several other multiple-choice questions, which allowed for multiple selections and an other option (Q4, Q5, and Q7), explored their perceptions regarding Microsoft Copilot’s role, factors aiding competency in constructing scientific explanations, and influential phenomenon-based learning activities. Notably, specific multiple-choice questions (Q2, Q5, and Q7) were each paired with a subsequent open-ended question (Q3, Q6, and Q8, respectively) to prompt students for further elaboration on their chosen multiple-choice question responses. Finally, the questionnaire concluded with an exclusively open-ended question (Q9) designed to solicit overall feedback on the integration of Microsoft Copilot with science learning. This mixed-format design was chosen to suit the participants' age group (approximately 13 years old), aiming to gather both structured and in-depth qualitative data effectively. The inclusion of multiple-choice questions with clear options alongside open-ended spaces was intended to facilitate comprehensive responses.

To ensure the instrument's quality, it underwent a rigorous validation process. Three science education experts reviewed the questionnaire for clarity, wording, alignment with research objectives, and appropriateness of question formats for all items, including open-ended questions. Feedback from experts guided initial refinements.Subsequently, a pilot test was conducted with one classroom of non-participant eighth-grade students. The main purpose of this pilot test was to assess the clarity and comprehensibility of all questionnaire items and to refine the response options for multiple-choice questions.This involved asking students to articulate their understanding of each question and conducting cognitive interviews to ensure questions were interpreted as intended. Based on student feedback from the pilot test, several questions were rephrased for better clarity, and new response options derived from common other responses were added to relevant multiple-choice items to enhance their comprehensiveness. After final revisions, the questionnaire was deployed via Google Forms for efficient data collection. Qualitative responses underwent thematic analysis, where data were systematically coded using an inductive approach to identify emerging themes and patterns. The coding process adhered to Braun and Clarke’s (2006) framework, ensuring reliability in thematic categorization.

Data Collection

The research process commenced with a comprehensive explanation of the study objectives to all participants. Informed consent was obtained following ethical guidelines to ensure voluntary participation and data confidentiality (further details on these ethical procedures are provided in the Ethics Statements section). Participants were assured that results would be presented in an aggregated format to protect their privacy.

Before starting the implementation of the learning intervention, students individually completed a five-item pre-test to evaluate students’ initial competency in constructing scientific explanations. Students were introduced to phenomenon-based learning with Microsoft Copilot through detailed instructions to ensure they understood the approach. The intervention was conducted over six weeks, following the phenomenon-based learning framework. After completion of the learning intervention, students completed the post-test (identical to the pre-test) to assess competency gains. They then responded to a mixed-format questionnaire that included a mix of choice and open-ended questions. The purpose of this survey was to explore students’ experiences and perceptions regarding the integration of Microsoft Copilot within the phenomenon-based learning approach, with a focus on identifying key factors contributing to their competency in constructing scientific explanations.

Analyzing of Data

Students’ competency in constructing scientific explanations was analyzed by both quantitative and qualitative approaches. For the quantitative analysis of competency test scores (pre-test and post-test), the assumption of normality for the differences between paired scores was first assessed using the Shapiro-Wilk test. As this assumption was violated (W = 0.889, p = .015) the non-parametric Wilcoxon Signed-Ranks Test was employed to compare students' competency before and after the intervention. Descriptive statistics, including means and standard deviations for pre-test and post-test scores, were also calculated to provide an overview of the data. To analyze data from the mixed-format questionnaire, which was completed by all students, thematic analysis was used to identify key themes and patterns of meaning in the data (Braun & Clarke, 2006; Clarke & Braun, 2013). Thematic analysis was chosen because it is a flexible method that provides a clear framework for the analysis, and at the same time, it can be adapted to the type of data. As an initial step in the analysis, students’ responses were first sorted systematically by question, and then all the answers from all the participants were grouped under each question in a structured format. These compiled responses were then transferred into a Microsoft Word document, with each individual response clearly labeled by its corresponding participant number to maintain data traceability and organization. Identifiable information was anonymized during the data compilation. To enhance the credibility and reliability of the findings, investigator triangulation was employed alongside thematic analysis to reduce potential bias that could arise from a single researcher’s interpretation. Next, the data were duplicated into three separate sets and distributed to three researchers for the familiarization process. Each researcher independently conducted coding on the same dataset to identify meaningful codes and extract key elements that represented significant patterns within the data. After the individual coding process, the research team conducted collaborative meetings to compare their codes, merge similar codes, and collaboratively organize them into thematic groups. These themes were then further refined to make sure they included only those that properly represented the important issues related to the research objectives. Any discrepancies in theme designation were resolved through theoretical considerations and empirical evidence. Once the themes were confirmed, researchers systematically assigned names and descriptions to ensure consistency in presenting the final results.

Results

Students’ Competency in Constructing Scientific Explanations Before and After the Intervention of Phenomenon-Based Learning with Microsoft Copilot

Table 1. Comparison of Competency Before and After Intervention Using Wilcoxon Signed-Rank Test (N=23)

Assessment	Maximum Score	Mean Score	SD	Z	p-value
Pre-test	30	3.48	3.20	4.213	< .001
Post-test		25.30	4.81

As detailed in Table 1, the results from the Wilcoxon Signed-Ranks Test revealed a statistically significant improvement in students' competency scores following the intervention (Z = 4.213, p < .001). The mean score rose substantially from 3.48 (SD = 3.20) at pre-test to 25.30 (SD = 4.81) at post-test. The median scores confirmed this large increase, shifting from 4.0 to 26.0, which underscores a considerable learning gain across the group.

Factors Influencing the Development of Students' Competency in Constructing Scientific ExplanationsThrough Phenomenon-Based Learning with Microsoft Copilot

Thematic analysis of students' qualitative responses from the mixed-format questionnaire revealed four key factors influencing their competency in constructing scientific explanations. An overview of these themes, along with their descriptions and illustrative quotes, is presented in Table 2. Each theme is discussed in detail below.

Table 2. Summary of Key Themes, Descriptions, and Illustrative Quotes Regarding Factors Influencing Competency in Constructing Scientific Explanations

Theme	Description	Illustrative Quotes	Example Codes
1. The Role of Microsoft Copilot in Enhancing Deep Understanding	Students perceived Microsoft Copilot as a valuable tool that helped them comprehend complex scientific concepts, synthesize information, access in-depth knowledge, and provide them with examples of good scientific explanations.	“Copilot can quickly summarize what I have learned in a way that makes sense and includes examples that improve my understanding.”	Summarizing informationLearning from examplesEnhancing understanding
		“When looking for answers, Copilot provides clear and easy-to-understand explanations.”	Ease of understandingSimplifying complex concepts
2. Connecting Theories to Real-World Phenomena Through Learning Media	Learning activities that connected scientific theories with real-world phenomena, particularly through visual media and initial hypothesis formulation, significantly aided understanding and explanation. This included viewing videos/images and attempting to explain phenomena from their own perspective.	“I generally remember images or videos better than textual content.”	Preference for visual mediaVisuals enhance memory
		“Watching videos or images and attempting to explain the phenomena from my own perspective—whether right or wrong—helped me learn and understand things more deeply.”	Using visuals as stimulusFormulating initial explanationDeepening understanding through practice
3. Collaborative Learning Activities	Group activities, such as class presentations and collaborative discussions, were seen as important for developing communication skills and the competency to construct scientific explanations through idea sharing and mutual understanding.	“…Because I had the opportunity to express my thoughts and exchange ideas with my friends.”	Sharing opinions in groupsPeer dialogue
		“…Working in groups allowed us to share our opinions with one another.”	Sharing opinions in groupsGroup work facilitates sharing
4. Enjoyable Learning Experiences and Student Engagement	The use of Microsoft Copilot and collaborative peer work contributed to enjoyable learning experiences and increased student engagement, which motivated them to acquire knowledge and develop their competency.	“I think using Copilot to search for information is quite engaging—it’s both educational and enjoyable.”	Enjoyment from technology useAI as an engagement tool
		“Working in a group lets us share our opinions with friends, making the experience even more enjoyable.”	Enjoyment from collaboration

Theme 1: The Role of Microsoft Copilot in Enhancing Deep Understanding

Most students indicated that Microsoft Copilot helped them find the complex information andpresent it in a simplified form so that it was easier for them to understand the deep scientific concepts in easy language. It helped them to effectively synthesize information from multiple sources and allowed for effective comprehension of difficult content. Furthermore, Microsoft Copilot provided examples of good scientific explanations that were particularly useful in enhancing students’ competency in constructing scientific explanations more effectively. Some of the student responses are

S5: "Copilot can quickly summarize what I have learned in a way that makes sense and includes examples that improve my understanding."

S10: "When looking for answers, Copilot provides clear and easy-to-understand explanations."

Theme 2: Connecting Theories to Real-World Phenomena Through Learning Media

The analysis revealed that specific activities within the phenomenon-based learning framework were instrumental in enhancing students' competency in constructing scientific explanations.These activities included:

First, most students identified viewing visual media—such as videos, images, and experiments related to phenomena like earthquakes or floods—as the activity that most significantly enhanced their ability to explain scientific phenomena. This activity helped them visualize actual phenomena. This point was clearly articulated by a student who emphasized the role of visuals in memory:

S6: “...I generally remember images or videos better than textual content.”

Second, formulating preliminary explanations and hypotheses before using Microsoft Copilot was also seen as beneficial. Many students reported that expressing their initial thoughts and analyses before conducting research helped them clarify their understanding and refine their explanations more effectively. An example supporting these findings is

S7: “Watching videos or images and attempting to explain the phenomena from my own perspective—whether right or wrong—helped me learn and understand things more deeply.”

Theme 3: Collaborative Learning Activities

The third key theme was the impact of collaborative learning activities. Students mentioned that group activities, such as class presentations and collaborative discussions, helped them develop their communication andcompetency in constructing scientific explanations. Moreover, working in groups and collaboratively explaining the researched information were pointed out by many students as activities that assist in sharing ideas and enhancing mutual understanding. In other words, students learned from their peers and co-created meaning. Some examples of student opinions that support these findings include

S2: “…Because I had the opportunity to express my thoughts and exchange ideas with my friends.”

S8: “…Working in groups allowed us to share our opinions with one another.”

Theme 4: Enjoyable Learning Experiences and Student Engagement

Finally, the analysis indicated that enjoyable learning experiences and high levels of student engagement were key contributing factors. This sense of enjoyment was derived from two primary sources: the use of technology and peer collaboration.

First, students reported that using Microsoft Copilot made the learning process more engaging and enjoyable. They perceived the tool not just as a way to simplify finding information but as an inherently interesting way to learn, which in turn increased their motivation. This sentiment was captured by S17:

S17: “I think using Copilot to search for information is quite engaging—it’s both educational and enjoyable.”

Additionally, students highlighted that collaborative work with their peers was a significant source of enjoyment. The ability to share ideas and opinions in a group setting was seen as making the learning process more interesting. As S8 explained:

S8: “Working in a group lets us share our opinions with friends, making the experience even more enjoyable.”

The thematic analysis showed that there are four key themes influencing the development of students’ competency in constructing scientific explanations: (a) the role of Microsoft Copilot in enhancing deep understanding, (b) connecting theories to real-world phenomena through learning media, (c) collaborative learning activities, and (d) enjoyable learning experiences and student engagement.

Discussion

The findings of this study demonstrate a significant improvement in students' competency in constructing scientific explanations following their participation in phenomenon-based learning with Microsoft Copilot. The substantial increase in competency scores, as indicated by the Wilcoxon Signed-Ranks Test (Z = 4.213, p < .001), suggests that this integrated approach is an effective pedagogical strategy.The magnitude of this improvement should be contextualized by the students' low baseline scores. At the pre-test stage, students had limited prior exposure to the specific format of constructing scientific explanations, and this unfamiliarity with the constructed-response task likely contributed to their very low initial scores. This outcome aligns with previous research highlighting the benefits of phenomenon-based learning for enhancing students' scientific explanation skills (Islakhiyah et al., 2017). Notably, our study extends this understanding by demonstrating how this competency can be further enhanced through the integration of a Gen AI tool. The enhanced competency observed in this study can be explained through four key supporting factors identified in the thematic analysis:

First Factor: The Role of Microsoft Copilot in Enhancing Deep Understanding

One of the primary drivers behind the students' improved competency was the role of Microsoft Copilot as a cognitive scaffolding tool. The qualitative data suggest that students did not merely use Microsoft Copilot as a search engine but as a partner that facilitated a deeper understanding of complex scientific content. For instance, students highlighted its ability to quickly summarize information and provide clear and easy-to-understand explanations (S5, S10). This function is critical because, by simplifying complex concepts and providing structured examples—a known capability of modern Gen AI to tailor content to individual user needs (López-Meneses et al., 2025; Rawatet al., 2024)—the AI tool likely reduced the students' extraneous cognitive load. This aligns with Cognitive Load Theory, which posits that reducing the cognitive burden associated with information processing allows learners to allocate more cognitive resources toward developing a deeper, more meaningful understanding (Chandler & Sweller, 1991; Hill & Hannafin, 2001; Sweller et al., 2011). This deeper understanding is essential, as students who are more knowledgeable about a phenomenon are better equipped to provide reasonable explanations(Laliyo et al., 2023; McCain, 2015).

In addition to reducing cognitive burden, Microsoft Copilot also served as a modeling tool by exposing students to well-structured scientific explanations. This process was facilitated by the teacher, who provided guidance on how to critically evaluate the explanations students received from the Gen AI (Cooper, 2023). This combination of exposure to models and guided evaluation is what helps students develop their competency in constructing scientific explanations.Crucially, our study suggests that Gen AI provides a dynamic and on-demand source of these examples, which is a distinct advantage over the static examples typically found in textbooks.This process, where learners engage with high-quality models, enables students to internalize the key components and structure of effective scientific reasoning, a learning mechanism supported by expert-novice studies (Chi et al., 1981), which demonstrated that observing experts' thinking and explanation methods enables novices to develop expertise more rapidly, and research on explanation-building scaffolds (Hsu et al., 2015; Wood et al., 1976; Zhang et al., 2006).

SecondFactor: Connecting Theories to Real-World Phenomena Through Learning Media

A second critical factor involved a two-step process designed to connect abstract theories with real-world phenomena: first, by presenting phenomena through engaging multimedia, and second, by prompting students to construct an initial explanation before conducting formal research. This approach, central to the initial phase of phenomenon-based learning design, appeared to be highly effective. Students’ qualitative responses indicated that observing phenomena through videos and images was a powerful learning stimulus. As one student noted, “I generally remember images or videos better than textual content” (S6), suggesting that visual media can enhance memory and engagement. By visualizing complex events like earthquakes or floods, students could bridge the gap between theoretical principles and their tangible manifestations. This practice aligns with established pedagogical frameworks that use real-world observations as a critical starting point (Cano & Lomibao, 2022; McLure, 2023). This initial exposure is crucial for developing students' ability to construct scientific explanations (Yao & Guo, 2018).

The second part of this process, Composing an Initial Explanation (step 2), was equally crucial for fostering metacognitive awareness. This step, where students articulated their initial understanding before conducting research, was more than just a preliminary exercise; it was a critical moment for them to externalize their initial thoughts and confront their own understanding. As one student articulated, the act of "attempting to explain the phenomena from my own perspective—whether right or wrong—helped me learn and understand things more deeply” (S7).This powerful statement suggests that by externalizing their initial ideas, students were able to identify gaps in their knowledge and potential misconceptions. This process of active self-assessment created a need to know, making the subsequent research phase with Microsoft Copilot more purposeful and effective. This finding supports the notion that student-led questioning and argumentation are vital for enhancing the ability to construct scientific explanations (Laliyo et al., 2023).

Third Factor: Collaborative Learning Activities

The third factor highlights the power of collaborative learning in developing students' competency in constructing scientific explanations. The data show that peer-to-peer interaction was not merely a supplementary activity but a core component of the learning process. Students explicitly stated that group work was beneficial because it provided an “opportunity to express my thoughts and exchange ideas with my friends” and “allowed us to share our opinions with one another” (S2, S8).This act of verbalizing and debating ideas with peers likely compelled students to organize their thoughts, justify their reasoning, and refine their understanding in real-time (Chusinkunawut et al., 2021).This finding is consistent with Vygotsky’s social constructivism theory (Vygotsky, 1978), which postulates that learning takes place through social interaction. Our study provides a contemporary example of this theory in a technology-enhanced setting, where collaborative discussions served as a crucial mechanism for co-constructing knowledge and moving from individual understanding to a shared, more robust explanation. The social interaction did not just facilitate a deeper understanding of the content; it also scaffolded the very practice of constructing and communicating scientific arguments, ultimately enhancing students' overall competency in constructing scientific explanations(Baker, 2015; Gerard et al., 2019; Schreiber & Valle, 2013).

Fourth Factor: Enjoyable Learning Experiences and Student Engagement

Finally, the affective dimension of the learning experience emerged as a significant factor. The data indicate that both the use of technology and collaborative work contributed to a more enjoyable and engaging learning environment. Students expressed that using Microsoft Copilot was perceived as “both educational and enjoyable” (S17),while working in groups made the experience “even more enjoyable” (S8). These findings corroborate existing literature by demonstrating that the integration of technologies such as Gen AI within a collaborative, phenomenon-based framework can significantly enhance student motivation and engagement (Almasri, 2024; Mauris De la Ossa et al., 2024; Ng et al., 2024). These emotional factors play a significant role in encouraging the level of students’ commitment to the development of competency in constructing scientific explanations. When students have positive attitudes toward learning, they tend to invest greater effort and dedication to understanding and explaining various phenomena. This result aligns with the concept of learning engagement, which posits that positive affect and fun are related to academic achievement (Fredricks et al., 2004; Marrone et al., 2024; Moore, 2019). Therefore, Microsoft Copilot in phenomenon-based learning not only improves learning effectiveness but also increases student motivation and satisfaction, ultimately contributing to educational success.

Conclusion

This study investigated the effect of Gen AI-assisted phenomenon-based learning on students' competency in constructing scientific explanations and identified the factors influencing this development. The findings clearly indicate that this integrated approach leads to statistically significant improvements in students' competency. This success stems from four key factors: Gen AI’s role as a cognitive scaffold and modeling tool, the connection of theory to real-world phenomena via multimedia, collaborative learning dynamics, and enhanced student engagement.

This research makes several important contributions to the field. First, it provides early empirical evidence of the synergistic effect that emerges from combining phenomenon-based learning with Gen AI for developing students' competency in constructing scientific explanations. More significantly, it moves beyond simply confirming efficacy by identifying the specific interplay of cognitive, pedagogical, and affective factors, offering a clearer understanding of how and why this technology-enhanced learning environment works.

The implications of these findings are significant for both practice and theory. For educational practitioners, this study demonstrates that Gen AI can function effectively as a cognitive partner within structured pedagogical frameworks, particularly when combined with teacher-facilitated critical evaluation. From a theoretical perspective, our results extend established frameworks such as Cognitive Load Theory and Social Constructivism by demonstrating how Gen AI can serve as a dynamic, more knowledgeable other while effectively reducing extraneous cognitive load.

By providing robust empirical evidence of Gen AI's effectiveness in enhancing constructing scientific explanations competency, this study establishes a foundational understanding that will inform evidence-based AI integration in science classrooms, moving the field beyond speculation toward systematic, research-driven implementation.

Recommendations

Based on the findings of this study, we offer the following precise recommendations for educational practice and future research.

Recommendations for Educational Practice

The findings underscore the pivotal role of teachers in AI-assisted learning environments. We recommend that educators develop two critical competencies: (a) effective AI prompt design techniques and (b) strategies for guiding students to critically evaluate AI-generated content.

Beyond teacher development, ensuring inclusive implementation requires addressing digital access barriers through both immediate and long-term strategies. At the classroom level, teachers can design collaborative activities that accommodate device sharing. Institutionally, schools should assess technology access and establish device lending programs where needed.

Recommendations for Future Research

Several research directions emerge from these findings. Comparative studies should examine the effectiveness of different Gen AI tools, such as ChatGPT and Gemini, alongside Microsoft Copilot to determine which platforms best support scientific explanation development. Additionally, controlled studies comparing Gen AI-assisted phenomenon-based learning with traditional approaches would help isolate the specific contributions of AI integration.

The generalizability of these findings needs further exploration across diverse student populations and educational contexts. Research should examine how this pedagogical model performs with students of different ages, backgrounds, and prior knowledge levels, as well as its effectiveness in subject areas beyond science.

Longitudinal research represents another critical avenue, particularly studies that track the retention of competency in constructing scientific explanations over extended periods. To support such long-term studies effectively, developing and validating automated assessment systems that can evaluate the quality of student-generated scientific explanations would provide valuable tools for both educators and researchers, enabling scalable feedback and more precise measurement of learning outcomes.

Finally, methodological considerations for research with young learners deserve attention. Our experience with eighth-grade students revealed two important insights for future research. During pilot testing, we observed that students often provided brief responses to open-ended questions, requiring us to allocate sufficient time and provide clear confidentiality assurances to elicit richer data. Additionally, instrument wording proved critical—using student-friendly language such as “performing better on the written test” rather than technical terms like “competency in constructing scientific explanations” enhanced comprehension and response validity(see the note in Appendix B). Researchers working with similar populations should incorporate these design considerations to improve data quality and ensure meaningful participation from young learners.

Limitations

While this study highlights the effectiveness of phenomenon-based learning with Microsoft Copilot in enhancing students' competency in constructing scientific explanations, several limitations should be considered.

First, the study's context and sample limit the generalizability of the findings. This research was conducted with a small, specific sample of 23 female eighth-grade students within a single science unit on Earth and Its Changes. Consequently, the results may not be directly transferable to male or co-educational populations, different grade levels, or other subject areas. While the small sample size was suitable for an in-depth, exploratory qualitative analysis, larger and more diverse samples are needed to validate the quantitative findings and enhance generalizability.

Second, the effectiveness of the intervention may be contingent on teacher quality, which was not controlled for in this study. The teacher facilitating the sessions possessed a strong understanding of the pedagogical framework and demonstrated crucial skills in guiding student inquiry, particularly in helping students critically evaluate information from Microsoft Copilot given the variable reliability of Gen AI-generated sources. The specific instructional style, enthusiasm, and ability to mediate student-AI interactions could have significantly contributed to the positive outcomes. Consequently, these findings may not be directly generalizable to settings where teachers have different levels of experience or training in technology-enhanced pedagogical skills. Additionally, this study did not systematically account for variations in teacher scaffolding, as the nature and extent of guidance provided may have differed between groups based on their immediate needs during activities, potentially influencing the learning process.

Third, the study did not analyze the internal dynamics within each student team, such as peer interactions, work distribution, or emergent leadership. The positive learning outcomes may have been influenced not only by the pedagogical model itself but also by the benefits of collaborative learning within supportive groups. Therefore, the impact of specific group dynamics on competency development remains an important area for future investigation.

Finally, technological factors presented several limitations. At the systemic level, while this study ensured all participants had device access, broader implementation must address digital equity, including disparities in personal device ownership and stable internet connections. At the individual level, this study did not control for student-device variance. Although all participants used a device, differences in hardware specifications (processing speed, screen size), software performance, and individual students' digital literacy could have led to different user experiences with Microsoft Copilot, potentially affecting engagement and learning outcomes. These technological factors should be considered when interpreting the findings.

Ethics Statements

This study was reviewed and approved by the Human Research Ethics Committee of SuanSunandha Rajabhat University, with the research project approval number COE. 2-054-2025. Written informed consent was obtained from the parents of all research participants before their involvement in the study. Furthermore, informed assent was actively sought from each student participant after the study's objectives, procedures, voluntary nature of participation, and their right to withdraw at any time without penalty were explained in age-appropriate terms bythe research team. Participants were informed of their right to withdraw from the study at any time if they felt uncomfortable. Additionally, all personally identifiable information of the participants was kept confidential and protected to ensure compliance with ethical research standards.

The authors would like to express their sincere gratitude to all the students who participated in this study for their time and valuable contributions. We are also deeply grateful to Asst. Prof. Dr. SuttipongBoonphadung from the Faculty of Education, SuanSunandha Rajabhat University, for his expert guidance and invaluable advice on the statistical analysis.

Conflict of Interest

The authors declare that there is no conflict of interest.

Generative AI Statement

To ensure clarity and accuracy in conveying the intended message, the author initially drafted the entire content in Thai. The manuscript was then translated into English by the author and subsequently refined using AI tools. ChatGPT was employed for linguistic enhancement, while QuillBot was used for grammar checking.

No AI was used to generate new content or modify academic concepts. After utilizing these AI tools, we thoroughly reviewed and verified the final version of our work. Additionally, the manuscript underwent further proofreading by a professional English language expert. We affirm that AI was used exclusively for language enhancement and had no involvement in data analysis, conclusions, or academic content development. As the authors, we take full responsibility for the content and accuracy of this published research.

Authorship Contribution Statement

Ratniyom: Conceptualization, design, analysis, writing, supervision, concept and design, data acquisition, data analysis, interpretation, drafting manuscript, critical revision of manuscript, final approval. Panmas: Concept and design, data acquisition, data analysis. Rattanakorn: Data analysis, statistical analysis. Tientongdee:Data analysis, critical revision of manuscript, editing/reviewing.

References

Adipat, S. (2024). Transcending traditional paradigms: The multifaceted realm of phenomenon-based learning. Frontiers in Education, 9, Article1346403. https://doi.org/10.3389/feduc.2024.1346403

Akkas, E., & Eker, C. (2021). The effect of phenomenon-based learning approach on students’ metacognitive awareness. Educational Research and Reviews, 16(5), 181-188. https://doi.org/10.5897/ERR2021.4139

Almasri, F. (2024). Exploring the impact of artificial intelligence in teaching and learning of science: A systematic review of empirical research. Research in Science Education, 54, 977-997. https://doi.org/10.1007/s11165-024-10176-3

Baker, M. J. (2015). Collaboration in collaborative learning. Interaction Studies, 16(3), 451-473. https://doi.org/10.1075/is.16.3.05bak

Balasubramaniam, S., Chirchi, V., Kadry, S., Agoramoorthy, M., Gururama Senthilvel, P., Satheesh Kumar, K., & Sivakumar, T. A. (2024). The road ahead: Emerging trends, unresolved issues, and concluding remarks in generative AI—A comprehensive review. International Journal of Intelligent Systems, 2024(1),Article 4013195. https://doi.org/10.1155/2024/4013195

Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77-101. https://doi.org/10.1191/1478088706qp063oa

Cano, J. C., & Lomibao, L. S. (2022). Design, development, and validation of phenomenon-based learning video for asynchronous remote instruction. American Journal of Educational Research, 10(4), 194-200.

Chandler, P., & Sweller, J. (1991). Cognitive load theory and the format of instruction. Cognition and Instruction, 8(4), 293-332. https://doi.org/10.1207/s1532690xci0804_2

Chi, M. T. H., Feltovich, P. J., & Glaser, R. (1981). Categorization and representation of physics problems by experts and novices. Cognitive Science, 5(2), 121-152. https://doi.org/10.1207/s15516709cog0502_2

Chusinkunawut, K., Henderson, C., Nugultham, K., Wannagatesiri, T., & Fakcharoenphol, W. (2021). Design-based science with communication scaffolding results in productive conversations and improved learning for secondary students. Research in Science Education, 51, 1123-1140. https://doi.org/10.1007/s11165-020-09926-w

Clarke, V., & Braun, V. (2013). Teaching thematic analysis: Overcoming challenges and developing strategies for effective learning. The Psychologist, 26(2), 120-123. https://uwe-repository.worktribe.com/output/937596

Cooper, G. (2023). Examining Science Education in ChatGPT: An Exploratory Study of Generative Artificial Intelligence. Journal of Science Education and Technology, 32, 444-452. https://doi.org/10.1007/s10956-023-10039-y

Deng, R., Jiang, M., Yu, X., Lu, Y., & Liu, S. (2025). Does ChatGPT enhance student learning? A systematic review and meta-analysis of experimental studies. Computers and Education, 227, Article 105224. https://doi.org/10.1016/j.compedu.2024.105224

Fredricks, J. A., Blumenfeld, P. C., & Paris, A. H. (2004). School engagement: Potential of the concept, state of the evidence. Review of Educational Research, 74(1), 59-109. https://doi.org/10.3102/00346543074001059

Gerard, L., Kidron, A., & Linn, M. C. (2019). Guiding collaborative revision of science explanations. International Journal of Computer-Supported Collaborative Learning, 14, 291-324. https://doi.org/10.1007/s11412-019-09298-y

Gupta, S., Tarapore, R., Haislup, B., & Fillar, A. (2025). Microsoft Copilot provides more accurate and reliable information about anterior cruciate ligament injury and repair than ChatGPT and Google Gemini; however, no resource was overall the best. Arthroscopy, Sports Medicine, and Rehabilitation, 7(2), Article 101043. https://doi.org/10.1016/j.asmr.2024.101043

Hill, J. R., & Hannafin, M. J. (2001). Teaching and learning in digital environments: The resurgence of resource-based learning. Educational Technology Research and Development, 49, 37-52. https://doi.org/10.1007/BF02504914

Hsu, C.-C., Chiu, C.-H., Lin, C.-H., & Wang, T.-I. (2015). Enhancing skill in constructing scientific explanations using a structured argumentation scaffold in scientific inquiry. Computers and Education, 91, 46-59. https://doi.org/10.1016/j.compedu.2015.09.009

Hundelhausen, M. E. M., Duran, O. P., & Barriosnuevo, M. C. J. (2024). Design of a school butterfly garden as an experiential learning strategy to strengthen students’ scientific and environmental competencies. International Journal of Professional Business Review, 9(12), Article e05199. https://doi.org/10.26668/businessreview/2024.v9i12.5199

Islakhiyah, K., Sutopo, S., & Yuliati, L. (2017). Scientific explanation of light through phenomenon-based learning on junior high school student. In Proceedings of the 1st Annual International Conference on Mathematics, Science, and Education (ICoMSE 2017) (pp. 141-153). Atlantis Press. https://doi.org/10.2991/icomse-17.2018.31

Jauhiainen, J. S., & Guerra, A. G. (2023). Generative AI and ChatGPT in school children’s education: Evidence from a school lesson. Sustainability, 15(18), Article 14025. https://doi.org/10.3390/su151814025

Jauhiainen, J. S., & Guerra, A. G. (2024). Generative AI and education: Dynamic personalization of pupils’ school learning material with ChatGPT. Frontiers in Education, 9, Article 1288723. https://doi.org/10.3389/feduc.2024.1288723

Khamcheensri, S., Chittrikhan, P., & Boonyaratklin, W. (2023). Nangsuerianraiwichaphunthanwitthayasatlaetheknoloyi, Matthayomsueksa pi thi 2, lem 2[Basic education textbook: Science and technology, grade 8, volume 2] (7th ed.). AksornCharoentat ACT Co., Ltd.

Laliyo, L. A. R., Utina, R., Husain, R., Umar, M. K., Katili, M. R., & Panigoro, C. (2023). Evaluating students’ ability in constructing scientific explanations on chemical phenomena. Eurasia Journal of Mathematics, Science and Technology Education, 19(9), Article em2328. https://doi.org/10.29333/ejmste/13524

Lee, A. V. Y., Tan, S. C., & Teo, C. L. (2023). Designs and practices using generative AI for sustainable student discourse and knowledge creation. Smart Learning Environments, 10, Article 59. https://doi.org/10.1186/s40561-023-00279-1

López-Meneses, E., López-Catalán, L., Pelícano-Piris, N., & Mellado-Moreno, P. C. (2025). Artificial intelligence in educational data mining and human-in-the-loop machine learning and machine teaching: Analysis of scientific knowledge. Applied Sciences, 15(2), Article 772. https://doi.org/10.3390/app15020772

Marrone, R., Lam, B., Barthakur, A., Brinkman, S., Dawson, S., & Gabriel, F. (2024). The relationship between wellbeing and academic achievement: A comprehensive cross-sectional analysis of system-wide data from 2016-2019. Journal of Learning Analytics, 11(3), 123-141. https://doi.org/10.18608/jla.2024.8357

Mauris De la Ossa, L. A., Garzón, N. N., Pérez Pulido, C., García Herrán, B. E., & Hoyos, J. H. (2024). Evaluation of artificial intelligence as a source of motivation in the teaching of chemistry: A study with tenth-grade students at the Adventist School of Ibagué. Evolutionary Studies in Imaginative Culture, 8(1), 2193-2208. https://doi.org/10.70082/esiculture.vi.1721

McCain, K. (2015). Explanation and the nature of scientific knowledge. Science and Education, 24, 827-854. https://doi.org/10.1007/s11191-015-9775-5

McLure, F. (2023). The thinking frames approach: Improving high school students’ written explanations of phenomena in science. Research in Science Education, 53, 173-191. https://doi.org/10.1007/s11165-022-10052-y

McNeill, K. L., & Krajcik, J. (2006, April 7-11). Supporting students’ construction of scientific explanation through generic versus context-specific written scaffolds [Paper presentation]. American Educational Research Association Annual Meeting, San Francisco, CA, United States.

McNeill, K. L., & Krajcik, J. S. (2008). Inquiry and scientific explanations: Helping students use evidence and reasoning. In J. Luft, R. L. Bell, & J. Gess-Newsome (Eds.), Science as inquiry in the secondary setting(pp. 121-134). National Science Teachers Association Press. http://bit.ly/46G9SYK

Ministry of Education Thailand. (2017). Basic education core curriculum B.E. 2551 (A.D. 2008).http://bit.ly/3U71sCi

Mittal, U., Sai, S., Chamola, V., & Sangwan, D. (2024). A comprehensive review on generative AI for education. IEEE Access, 12, 142733-142759. https://doi.org/10.1109/ACCESS.2024.3468368

Moore, P. J. (2019). Academic achievement. Educational Psychology, 39(8), 981-983. https://doi.org/10.1080/01443410.2019.1643971

Ng, D. T. K., Tan, C. W., & Leung, J. K. L. (2024). Empowering student self-regulated learning and science education through ChatGPT: A pioneering pilot study. British Journal of Educational Technology, 55(4), 1328-1353. https://doi.org/10.1111/bjet.13454

Organisation for Economic Co-operation and Development. (2023). PISA 2022 results (Volume I and II) - Country notes: Thailand. https://bit.ly/4kXGMrA

Organisation for Economic Co-operation and Development. (2025). PISA 2025 science framework. https://pisa-framework.oecd.org/science-2025/

Penuel, W. R., Turner, M. L., Jacobs, J. K., Van Horne, K., & Sumner, T. (2019). Developing tasks to assess phenomenon-based science learning: Challenges and lessons learned from building proximal transfer tasks. Science Education, 103(6), 1367-1395. https://doi.org/10.1002/sce.21544

Rasul, T., Nair, S., Kalendra, D., Robin, M., de Oliveira Santini, F., Ladeira, W. J., Sun, M., Day, I., Rather, R. A., & Heathcote, L. (2023). The role of ChatGPT in higher education: Benefits, challenges, and future research directions.Journal of Applied Learning and Teaching, 6(1),41-56. https://doi.org/10.37074/jalt.2023.6.1.29

Rawat, B., Joshi, Y., & Bisht, N. S. (2024). Investigating the impact of machine learning in personalized education systems. In Proceedings of the 2024 First International Conference on Technological Innovations and Advance Computing (TIACOMP) (pp. 221-227). IEEE. https://doi.org/10.1109/TIACOMP64125.2024.00045

Schreiber, L. M., & Valle, B. E. (2013). Social constructivist teaching strategies in the small group classroom. Small Group Research, 44(4), 395-411. https://doi.org/10.1177/1046496413488422

Sukumal, P. (2024). Assessing learning outcomes in Thai secondary schools: A comparative study of traditional and alternative evaluation methods. Insights into Modern Education, 1(6), 53-66.

Sweller, J., Ayres, P., & Kalyuga, S. (2011). Cognitive load theory. Springer. https://doi.org/10.1007/978-1-4419-8126-4

Symeonidis, V., & Schwarz, J. F. (2016). Phenomenon-based teaching and learning through the pedagogical lenses of phenomenology: The recent curriculum reform in Finland. Forum Oświatowe, 28(2), 31-47. https://bit.ly/4jM1ucV

Vygotsky, L. S. (1978). Mind in society: Development of higher psychological processes (M. Cole, V. John-Steiner, S. Scribner, & E. Souberman, Eds.). Harvard University Press. https://doi.org/10.2307/j.ctvjf9vz4

Wannathai, P., & Pruekpramool, C. (2024). Investigating the ability in constructing scientific explanations of Thai grade 10 students: Insights from learning achievement, attitude, and school size. Science Education International, 35(2), 143-153. https://doi.org/10.33828/sei.v35.i2.8

Wood, D., Bruner, J. S., & Ross, G. (1976). The role of tutoring in problem solving. Journal of Child Psychology and Psychiatry, 17(2), 89-100. https://doi.org/10.1111/j.1469-7610.1976.tb00381.x

Yao, J.-X., & Guo, Y.-Y. (2018). Validity evidence for a learning progression of scientific explanation. Journal of Research in Science Teaching, 55(2), 299-317. https://doi.org/10.1002/tea.21420

Zapata-Rivera, D., Torre, I., Lee, C.-S., Sarasa-Cabezuelo, A., Ghergulescu, I., & Libbrecht, P. (2024). Editorial: Generative AI in education. Frontiers in Artificial Intelligence, 7, Article 1532896. https://doi.org/10.3389/frai.2024.1532896

Zhang, B., Liu, X., & Krajcik, J. S. (2006). Expert models and modeling processes associated with a computer-modeling tool. Science Education, 90(4), 579-604. https://doi.org/10.1002/sce.20129

Zhou, E., & Lee, D. (2024). Generative artificial intelligence, human creativity, and art. PNAS Nexus, 3(3), Article pgae052. https://doi.org/10.1093/pnasnexus/pgae052