Bulletin of the Technical Committee on Learning Technology (ISSN: 2306-0212) |
Authors:Stephanie Karmini Emmanuel Gerard1, Mas Nida Md. Khambari*
1 , and Ahmad Fauzi Mohd Ayub1,
1: Universiti Putra Malaysia, Serdang, Selangor, Malaysia
Abstract:
Generative Artificial Intelligence (GenAI) is increasingly being adopted in education, yet there remains limited understanding of how secondary STEM teachers perceive its performance value in everyday classroom practice. Guided by the Unified Theory of Acceptance and Use of Technology (UTAUT), this qualitative case study examined how secondary STEM teachers perceived the role of performance expectancy in their use of GenAI for teaching and learning. Semi-structured interviews were conducted with eight STEM teachers from a public secondary school in Kuala Lumpur, Malaysia, and the data were analyzed using thematic analysis. The findings revealed that teachers perceived the performance value of GenAI in reducing lesson planning time, supporting instructional design, and assisting differentiated instruction. However, these benefits were not perceived as automatic. Teachers consistently emphasized the need to verify, adapt, and align GenAI outputs with curriculum requirements and students’ learning needs before using it in classroom practice. The study proposes the concept of conditional performance expectancy, whereby the perceived usefulness of GenAI depends on teachers’ professional judgement rather than the technology alone. These findings contribute to understanding teachers’ use of GenAI in secondary STEM education and highlight the importance of professional development that supports the critical and responsible integration of GenAI into classroom practice.
Keywords: Generative Artificial Intelligence (GenAI), STEM education, performance expectancy, Unified Theory of Acceptance and Use of Technology (UTAUT).
I. INTRODUCTION
Generative Artificial Intelligence (GenAI) has emerged as a major technological development with growing relevance for education. GenAI can produce new content, including text, images, code, audio, and video, by identifying patterns in large datasets [1]. Recent studies show that GenAI is used for lesson planning, tutoring, content generation, peer evaluation, assessment, and the development of higher-order thinking skills [2], [3]. [4] suggested that GenAI has the potential to support instructional improvement by assisting teachers in designing engaging learning experiences and adapting instruction to students’ needs. Studies also caution that teachers must remain involved in evaluating and adapting GenAI-generated content because effective use depends on classroom context and teacher judgement [5]. The instructional value of GenAI depends on whether teachers use it to the realities of their classrooms, subject requirements and students’ learning needs.
Existing research on GenAI has largely focused on higher education while studies in secondary education have examined technology acceptance and adoption [6]. In Malaysia, national policies such as the Malaysia Education Blueprint Annual Report 2023, the National Fourth Industrial Revolution (4IR) Policy, and the Education Development Roadmap (RPM) 2026–2035 emphasize strengthening STEM education, digital transformation, AI readiness and the integration of emerging technologies into teaching. These national policy directions require secondary STEM teachers to prepare students with future-ready competencies while also ensuring that teaching and learning is aligned with the national curriculum requirements. The integration of GenAI requires context-sensitive pedagogical decisions that consider learners, curriculum and local educational environments [5]. Empirical evidence show that GenAI has the potential to support lesson planning, materials generation and differentiated instruction but findings have emphasized that teachers must evaluate the quality of AI-generated outputs before integrating them into classroom teaching [6]. Despite these developments, there remains limited understanding of how secondary school STEM teachers perceive the role of performance expectancy in their use of GenAI in teaching and learning. Given the benefits of GenAI, fewer than half of school teachers integrated GenAI into educational practice where it was used for lesson preparation, assessment, and administrative tasks than using it for classroom teaching and learning activities [7]. Thus, this study aims to address the following research question: How do secondary STEM teachers perceive the role of performance expectancy in their use of GenAI in teaching and learning?
Although the foundational capabilities of GenAI are recognized globally, this study highlights that in emerging economies, its perceived value is heavily dependent on overcoming localized systemic constraints. This is in congruent with [8] and [9] who mentioned the importance of understanding the systemic contexts and constrains of users. For developing nations striving to meet advanced educational standards without the robust support systems often found in developed countries, teachers’ ‘conditional performance expectancy’ acts as a vital bridge between technological potential and on-the-ground classroom realities.
Performance expectancy is one of the most important concepts in the technology acceptance literature. Originating from the UTAUT, it integrates constructs such as perceived usefulness from the Technology Acceptance Model (TAM), relative advantage from the Innovation Diffusion Theory and outcome expectations from Social Cognitive Theory [8]. It refers to the extent to which individuals believe that using a technology will improve their job performance [8]. Previous studies have examined performance expectancy of GenAI using quantitative approaches grounded theories like UTAUT or TAM and examined behavioral intention or technology adoption [9], [10]. These quantitative models operationalize performance expectancy as perceived usefulness but cannot explain how teachers negotiate conflicting demands between efficiency and pedagogical responsibility. Within these studies, performance expectancy has been conceptualized through perceived usefulness, productivity benefits and teachers’ intentions to adopt GenAI. As such a qualitative case study offers the opportunity to explore how secondary STEM teachers construct the meaning of performance expectancy through their everyday experiences of planning lessons, designing instruction, adapting GenAI-generated materials and responding to diverse student needs.
Across educational contexts, teachers have described GenAI as improving instructional preparation and classroom support. In Germany, [6] found that STEM teachers perceived ChatGPT as beneficial for lesson planning, assessment and personalized learning, and these perceived benefits influenced them to use ChatGPT for their classroom practices. Similarly, K–12 teachers used AI-tools for preparing instructional materials, designing differentiated activities and supporting personalized learning [11]. Likewise, [12] found that science teachers perceived AI-tools integrated lessons as effective for promoting inquiry-based learning and improving student engagement in classroom practice. Taken together, these findings provide evidence that teachers perceived positive performance expectancy with GenAI for improved efficiency in lesson preparation and instructional support. However, these findings still lack to explore how teachers judge the conceptual accuracy, curriculum alignment or pedagogical appropriateness of GenAI-generated materials [13]. As, STEM subjects involve high accuracy, STEM teachers must evaluate whether AI-generated explanations are scientifically accurate, consistent with curriculum standards and appropriate for students with different learning needs before integrating them into classroom practice [5], [14]. These previous studies lack insights on how secondary STEM teachers determine whether GenAI-generated materials are suitable for effective classroom implementation. This reinforces the need to explore how teachers perceive GenAI as contributing to teaching quality and lesson delivery within their respective STEM subject.
GenAI was perceived as useful in assisting teachers in designing innovative lesson plans, developing engaging learning activities, and providing instructional suggestions that enhance classroom practice [4]. Similarly, [11] found that teachers perceived that the use of AI- tools to prepare instructional materials, develop lesson activities, and support personalized learning, indicating that AI- tools functions as a resource for generating instructional ideas rather than merely automating routine tasks. In STEM education, these capabilities may enable teachers to explore alternative teaching approaches that promote inquiry-based learning and student engagement [12]. These studies emphasize the availability of AI-generated ideas while providing limited understanding of how secondary STEM teachers evaluate, adapt, and apply these ideas within their own classroom contexts.
II. THEORETICAL FRAMEWORK
This study is guided by the Unified Theory of Acceptance and Use of Technology (UTAUT), developed by [15]), has four constructs which are performance expectancy, effort expectancy, social influence, and facilitating conditions as key determinants of technology adoption. The UTAUT identifies performance expectancy as the strongest predictor of technology use because it reflects users’ beliefs that a technology will improve job performance [15]. As this study examines how secondary STEM teachers perceive the role of performance expectancy in their use of GenAI, performance expectancy serves as the primary theoretical lens. The perceived usefulness of GenAI is understood as both the capabilities of the technology and also on teachers’ professional judgement in evaluating, refining and contextualizing AI-generated outputs before classroom use. This study conceptualizes performance expectancy as conditional performance expectancy, whereby usefulness depends on teachers’ evaluation, curriculum alignment, verification and professional judgement. Thus, this study adopts a contextualized interpretation of performance expectancy that recognizes teachers as active evaluators of GenAI technology.
III. METHODS
This study was conducted in a secondary school located in Kuala Lumpur, Malaysia. The school maintains the national 60:40 STEM to non-STEM student ratio and records a one of the highest STEM-stream enrolments in the Sentul district. Second, there was an adequate pool of possible participants because the school has many STEM teachers who are currently teaching in the school. This school is also prepared to adopt cutting-edge teaching strategies in line with Industry 4.0 requirements compared to other urban public schools due to its well-equipped advanced facilities, which include projectors, smart TVs, interactive whiteboards, and 3D printing technology. Focusing on a single school allowed the researcher to gain a deeper understanding of the phenomenon within its real-life context, which is consistent with qualitative case study research [16].
A. Participants and Sampling
In terms of selection method, participants were required to have at least two years of teaching experience, currently teach at least one STEM subject, and have experience using GenAI in their teaching and learning practices. The participants represented Mathematics, Additional Mathematics, Science, Biology, Chemistry, and Physics, allowing different STEM perspectives to be explored. As participants were purposively selected based on their prior experience using GenAI, the findings reflect the perspectives of experienced users within a single school context. Therefore, the findings should be interpreted within the scope of this bounded qualitative case study rather than as representative of all secondary school teachers.
The pilot study was conducted at the same school and involved two participants who were not included in the main study. This step was instrumental to seek for any discrepancies in the interview protocol. Once this was done, the study progressed with the actual data collection. Participants were secondary school STEM teachers who had experience using GenAI tools in their teaching practices. Purposive sampling was used to recruit participants who could provide rich information relevant to the research objectives. Eight secondary school STEM teachers from one public secondary school in Kuala Lumpur participated in the study. The final sample of eight teachers was considered sufficient because data saturation had been achieved. During the analysis of the seventh and eighth interviews, no new codes or sub-themes emerged, indicating that the existing themes had been sufficiently developed and additional interviews were unlikely to provide substantially new insights [16].
TABLE I OVERVIEW OF PARTICIPANTS
| Pseudonym | Gender | Teaching Experience | Subject Area(s) | GenAI Tools Used |
|---|---|---|---|---|
| Raul | Male | 15 years | Mathematics, Additional Mathematics | ChatGPT, Gemini |
| Mayuran | Male | 15 years | Science, Biology | ChatGPT, Gemini, QuillBot, Grammarly |
| Shikin | Female | 18 years | Mathematics | ChatGPT, Gemini |
| Khadijah | Female | 15 years | Physics | ChatGPT, Gemini |
| Chong Kah Ting | Female | 8 years | Chemistry | ChatGPT, Gemini |
| Angel | Female | 15 years | Mathematics | ChatGPT, Gemini |
| Dhalia | Female | 15 years | Mathematics, Science | ChatGPT, Gemini |
| Shumitra | Female | 3 years | Chemistry | ChatGPT, Gemini, Canva AI |
B. Data Collection Methods
Data for this study were collected through semi-structured interviews with STEM teachers. Semi-structured interviews allow researchers to explore participants’ experiences, perceptions, and perspectives while providing flexibility to probe deeper into relevant topics [17].
An interview protocol was developed based on the research question and the UTAUT framework. The interview questions focused on teachers’ perceptions of performance expectancy when using GenAI tools in teaching and learning. Each interview lasted approximately 30 to 45 minutes and was conducted face-to-face. With participants’ consent, the interviews were audio-recorded and later transcribed for analysis.
C. Data Analysis
Data analysis followed Creswell’s data analysis spiral. The five steps of data analysis involved (i) managing and organizing the data; (ii) reading and recording ideas; (iii) coding and classifying data into themes; (iv) interpreting the data; and (v) presenting the findings [18]. Data was analyzed using the analytic circles approach as the process of each step in the data analysis is rather interrelated and these steps could happen simultaneously [18].
1) Managing and Organizing the Data
To manage and organize data accordingly, three files were created for each participant containing the consent form, demographic information, and interview transcript. Each interview was recorded and the audio recordings were transcribed and checked against the recordings to ensure transcription accuracy. Hardcopy documents were converted into digital form. Backup copies were saved in external storage and in Google Drive to ensure data security and prevent data loss.
2) Reading and Recording Ideas
Each interview transcript was read several times to gain a comprehensive understanding of the participants’ experiences and perceptions regarding their use of GenAI in teaching and learning. Repeated reading enabled the researcher to become familiar with the content before commencing the coding process and to identify recurring ideas and meaningful statements across participants.
3) Coding and Classifying Data into Themes
The interview transcripts were imported into NVivo for coding and a two-cycle coding method was adopted [19]. Coding was conducted by the first author and discussed regularly with the supervisory team. For the first cycle, Descriptive Coding was used to assign concise labels to meaningful segments of text that summarized the topic or issue discussed by the participants. Each transcript was coded line by line in NVivo, and similar responses were assigned the same descriptive code. For the second cycle, Pattern Coding was used to examine relationships among the descriptive codes and group similar codes into broader categories. Related codes were compared across all participants and merged to develop sub-themes. This process was conducted inductively, allowing codes, categories and subthemes to emerge from the interview transcripts. Next, sub-themes were organized deductively under the broader theory-driven themes of PE and ensured that the sub-themes reflected the participants’ experiences.
4) Interpreting the Data
The themes and sub-themes were reviewed repeatedly against the interview transcripts to ensure they accurately represented the participants’ responses. Similarities and differences of experiences across participants were compared, and quotes were selected to support each sub-theme. The findings were interpreted in relation to the research questions, literature and theoretical framework of the study.
5) Presenting the Findings
Finally, the findings of the study were presented according to the research questions. Each main theme and the sub-themes were described and supported by quotes from the participants as in the transcript. NVivo was used throughout the analysis to facilitate data organization, retrieval, comparison, and management of coded data.
Validity and trustworthiness were addressed throughout this study to ensure meaningful and useful findings. The semi structured protocol interview underwent an expert review process to establish content validity. Two lecturers with expertise in educational technology and qualitative research reviewed the interview questions for clarity, relevance, and alignment with the study objectives. Feedback from the expert review was incorporated to refine the wording and sequencing of the questions before pilot testing. Credibility was strengthened through member checking. Participants reviewed their interview transcripts to confirm accuracy. They were also asked to review selected interpretations of the findings to ensure that their views were not misrepresented. To confirm whether they agreed with the interpretation of their interviews, they had to sign a respondent validation form. This approach is known as member checking and it is used to avoid misinterpretation of the participants’ views [17]. Additionally, any discrepant cases or outlier opinions were actively sought and discussed among the research team to ensure the themes accurately reflected the full range of participant experiences. Early assumptions on the findings were avoided by making interpretations based on recorded data and theoretical framework of this study. An audit trail was also kept to document the research process and key decisions to ensure that there was no bias regarding the study conducted as well as the findings derived from it [20]. To achieve this, details of the procedures involved in this study was thoroughly explained to the participants.
Ethical procedures were followed before data collection. Informed consent forms were given to all participants. The participants were informed about the purpose of the study, their role, and their right to withdraw at any time. The possible risks and benefits of participation were also explained. Permission to conduct the study was obtained from the Universiti Putra Malaysia Human Research Ethics Committee (JKEUPM) and the Educational Planning and Research Division of the Ministry of Education. Permission was also obtained from the school principal before inviting teachers to participate. To protect confidentiality, pseudonyms were used for the school and participants. Interview recordings and transcripts were kept confidential. In order to minimize harm towards the participants of this study the questions used in the instrument were non-invasive and designed to avoid emotional distress or discomfort.
IV. FINDINGS
This section presents findings related to Performance Expectancy, referring to teachers’ perceptions of how the use of GenAI enhances their teaching effectiveness and work performance. The analysis is organized into three subthemes: (i) reducing planning time while increasing professional judgement, (ii) supporting instructional design, and (iii) supporting differentiated instruction.
TABLE II FINDINGS OVERVIEW
| Subtheme | Representative evidence | Interpretation |
|---|---|---|
| Reducing planning time while increasing professional judgement | Teachers perceived that GenAI reduced lesson preparation time by generating lesson plans, activities and draft materials, outputs still required verification and adaptation. | Teachers perceived GenAI as improving planning efficiency while maintaining responsibility for evaluating and contextualising AI-generated outputs. |
| Supporting instructional design | Teachers perceived GenAI as a brainstorming partner for alternative teaching ideas, lesson structures and classroom activities, particularly when they experienced creative fatigue. | Teachers perceived GenAI as enhancing instructional creativity by expanding teaching possibilities rather than replacing lesson planning. |
| Supporting differentiated instruction | Teachers perceived GenAI as assisting the preparation of differentiated materials for students with different ability levels, while recognising the continued need to review the outputs before classroom use. | Teachers perceived GenAI as supporting differentiated instruction while maintaining professional responsibility for verifying instructional quality and suitability. |
A. Reducing Planning Time while Increasing Professional Judgement
Across participants, GenAI was frequently used as an aid for planning lesson preparation. Teachers used GenAI to generate lesson structures, activities, and draft materials for their respective subjects’ lesson plans. Teachers perceived the value of GenAI in supporting the early stages of lesson preparation than to generate materials rapidly and maintained professional responsibility for instructional quality.
Angel explained how GenAI supported faster lesson preparation by providing immediate ideas that could be selected and refined:
“In terms of getting ideas, it really saves time. GenAI can give us activity ideas. Once we have the idea, we can quickly complete the RPH. Overall, it saves a lot of time, because it gives us ideas and we just need to choose which ones to use.”
Similarly, Shikin described how GenAI reduced the time but the spent was spent validating its work:
“For example, lesson plans used to take a lot of time because we needed to think of suitable activities. Now, GenAI generates suggested activities but I still need to validate whether the lesson plan is suitable for my students.”
These accounts indicate that GenAI strengthened performance expectancy by reducing the time and cognitive effort required for lesson planning and material preparation, especially at the initial stage of instructional design. Teachers consistently perceived that GenAI reduced the time required to generate lesson plans, teaching materials and classroom activities. These findings support previous studies showing that teachers use GenAI to reduce lesson preparation time and generate instructional materials more efficiently [6], [11].
However, the perceived time saving was conditional as participants described producing initial teaching materials more quickly, they also emphasized that these outputs required careful checking, adaptation and curriculum alignment before classroom use. GenAI shifted their work from creating instructional materials to evaluating, refining and contextualizing output from GenAI. The findings extend this understanding by showing that efficiency alone did not determine teachers’ perceptions of performance expectancy. Although participants valued the speed with which GenAI generated lesson plans and activities, they consistently explained that these outputs required verification, adaptation, and curriculum alignment before classroom implementation. This finding supports [5]), who argued that teachers remain responsible for evaluating output from GenAI before using it in their respective subjects. Thus, GenAI is perceived as improving teachers’ performance only when its outputs could be reviewed and adapted using their professional judgement.
B. Supporting Instructional Design
Teachers perceived GenAI as a source of ideas. Participants repeatedly described GenAI as useful for brainstorming, generating instructional possibilities, and helping them think beyond routine approaches. In this sense, GenAI was valued for saving time and also for supporting teachers’ creativity in planning lessons and activities.
Mayuran explained:
“When you get ideas directly and quickly, it reduces the time needed to think through everything from scratch.’’
Dhalia also connected GenAI to idea generation during moments of mental fatigue:
“I feel it improves the quality, especially when I’m tired or out of ideas. As teachers, our workload is heavy. Even if we are creative, sometimes our brains feel ‘tepu’ (saturated). GenAI helps spark new ideas.”
Teachers perceived GenAI as supporting instructional design by providing alternative teaching ideas, classroom activities and lesson structures. GenAI is described as a source of new instructional possibilities because it expanded their instructional ideas by suggesting alternative classroom activities, lesson structures and approaches that they may not have considered independently. This is similar to previous studies where teachers perceived GenAI as a useful resource for generating instructional ideas than replacing lesson planning [4], [12]. Thus, teachers perceived GenAI as stimulating new instructional possibilities especially when they experienced time constraints or during creative fatigue.
However, this study’s’ findings provide a deeper explanation of how teachers judged these ideas. Teachers did not accept GenAI outputs directly rather they selected, modified, and contextualized ideas from GenAI according to their teaching objectives and students’ needs. These findings show that teachers perceived GenAI as for instructional creativity. A collaborative process between GenAI and teacher expertise existed rather than as an automated function of the technology. Consequently, performance expectancy reflected teachers’ confidence in their ability to transform GenAI-generated suggestions into meaningful classroom activities instead of relying on the technology itself.
C. Supporting Differentiated Instruction
Teachers also perceived GenAI as supporting differentiated instruction by enabling them to generate learning materials for students with different ability levels more efficiently. While idea generation focused on producing possibilities, teaching quality referred more directly to how those ideas translated into classroom practice and supported better learning experiences.
Raul also highlighted the usefulness of GenAI for differentiated mathematics teaching:
“In Maths, I can easily generate scaffolded explanations for students who struggle with basic algebra, and complex challenge problems for those who can handle advanced mathematics like calculus.”
Chong Kah Ting acknowledged the value of GenAI for differentiated learning, but also emphasised the limitations:
“For that purpose, I asked GenAI to generate different levels of questions. During the lesson, I could see students answering questions according to their levels, and it did improve their learning experience.”
However, participants also cautioned the use of GenAI for differentiated learning.
Raul also acknowledged the limitations of using GenAI-generated differentiated materials:
“In terms of marking worksheets, GenAI is not very helpful for me. I still have to check students’ work manually to see whether their methods and answers are correct.”
Teachers perceived that GenAI made differentiated instruction easier because it generated learning materials for students with different ability levels more efficiently than preparing them manually. Previous studies also have reported that GenAI assists teachers in preparing differentiated learning materials and supporting personalized instruction [6], [11]. The present findings support these observations but also show that teachers evaluated differentiated materials according to their suitability for different groups of learners. Across all participants, they also recognized that producing differentiated materials quickly did not remove the need for professional judgement. They explained that every worksheet, activity and explanation still required verification for factual accuracy, syllabus alignment and suitability for their students. These findings suggest that teachers perceived GenAI as a useful support tool for differentiated instruction because it reduced the burden of preparing multiple sets of materials and made it easier to tailor content to different proficiency levels.
However, as teachers carefully reviewed the GenAI outputs for accuracy, suitability, and alignment with syllabus expectations it uses in classrooms remained selective. Thus, teachers perceived differentiated instruction as depending on both the capabilities of GenAI and their own professional judgement. Therefore, the value of GenAI was determined by its ability to generate multiple versions of learning materials and also by teachers’ responsibility for ensuring instructional quality.
V. DISCUSSION
The findings revealed a consistent tension between the efficiency gained during content generation and the additional professional effort required for verification, adaptation and curriculum alignment across all participants. The findings indicate that the perceived value of GenAI was conditional rather than automatic. Participants consistently described benefits such as reducing planning time, generating instructional ideas, and supporting differentiated instruction. However, these benefits depended on teachers’ willingness to verify, adapt, and contextualize the GenAI outputs before using it. This extends the application of performance expectancy within the UTAUT framework where performance expectancy is understood as the belief that technology improves job performance [15]. This study proposes that performance expectancy in secondary STEM teaching should be understood as conditional performance expectancy, where perceived usefulness emerges only after teachers evaluate, adapt, and contextualize GenAI outputs.
The findings suggest that secondary STEM teachers find GenAI to be useful only when it supported their professional responsibilities indicating that teachers did not evaluate GenAI according to the speed with which it generated teaching materials alone but on the outputs that would support accurate teaching, meet curriculum expectations, and address students’ different learning needs before they considered the technology valuable. In this study, teachers remained responsible for ensuring factual accuracy, curriculum alignment, and suitability for students with different learning needs. Thus, performance expectancy reflected teachers’ evaluation of whether GenAI could assist professional decision-making and not whether it generated teaching materials quickly. Some variation in the use of GenAI among different STEM disciplines was observed. For example, mathematics teachers described using GenAI primarily to generate scaffolded explanations and differentiated mathematical problems, whereas teachers from other STEM subjects more frequently referred to lesson planning, classroom activities and instructional ideas. While these observations cannot be generalized, they suggest that teachers may evaluate the usefulness of GenAI according to the instructional demands of their subject.
This study also reflects on the broader expectations placed on Malaysian secondary school teachers under the Rancangan Pendidikan Malaysia (RPM) 2026–2035. The policy emphasizes high-quality teaching, continuous professional development, ethical digital pedagogy and the preparation of future-ready learners [21]. Within this context, teachers are expected to integrate emerging technologies and ensure that instructional materials remain accurate, pedagogically appropriate and aligned with curriculum requirements. The RPM also highlights more flexible and competency-based learning experiences that address diverse learner needs. Although the RPM 2026–2035 represents a future policy direction rather than fully implemented classroom practice, it signals Malaysia’s continued shift towards AI integration, digital transformation, teacher professional development and future-ready STEM education. In this regard, the findings suggest that teachers are already demonstrating professional practices that will be important as these reforms are implemented. Participants consistently emphasized verifying AI-generated outputs, adapting instructional materials and ensuring curriculum alignment, indicating a readiness to engage with GenAI responsibly within evolving educational expectations.
These findings also have practical implications for the implementation of GenAI in Malaysian secondary schools. As Malaysia continues strengthening STEM education and integrating digital technologies through the RPM 2026–2035, professional development should extend beyond introducing teachers to GenAI tools. Instead, training should emphasize how teachers can critically evaluate AI-generated content, verify factual accuracy, ensure curriculum alignment and adapt instructional materials to students’ learning needs. Schools should also provide guidance that supports responsible classroom use of GenAI, while developers of educational GenAI tools should consider designing features that better support curriculum alignment, differentiated instruction and teacher review.
VI. CONCLUSION
This study examined how secondary STEM teachers perceived the role of performance expectancy in their use of GenAI in teaching and learning. The findings showed that teachers perceived performance value across three main dimensions: reducing planning time while increasing professional judgement, supporting instructional design, and supporting differentiated instruction. However, these perceived benefits were not automatic. Teachers consistently explained that AI-generated outputs required verification, adaptation, and curriculum alignment before classroom implementation. Consequently, performance expectancy was shaped not only by the efficiency of GenAI but also by teachers’ professional judgement in determining whether AI-generated materials were accurate, appropriate, and suitable for their students.
This study has several limitations. First, the findings were derived from eight secondary STEM teachers from a single public secondary school and therefore the findings represent the experiences of a bounded qualitative case rather than all Malaysian secondary school teachers. All participants were existing users of GenAI, meaning the perspectives of non-users or infrequent users were not captured. The findings were also based on teachers’ self-reported experiences and did not include classroom observations, usage data, or student learning outcomes. Therefore, the study cannot determine whether the perceived benefits resulted in improved classroom practice or student achievement. Future research could include teachers from multiple schools and different educational settings, compare experienced and inexperienced GenAI users, and incorporate classroom observations, usage logs, and student learning data to provide a more comprehensive understanding of how GenAI supports performance expectancy in secondary school STEM teaching practices.
In summary, performance expectancy contributed to teachers’ use of GenAI because they perceived it as a useful tool for saving time, generating ideas, improving lesson delivery, and supporting differentiated learning with the teachers’ judgment, adaptation, and verification. The use of GenAI in STEM teaching was conditional on teachers’ ability to review, adapt, and align its outputs with pedagogical and curriculum demands.
Reference
[1] J. Kang and J. Ahn, “Technologies, opportunities, challenges, and future directions for integrating generative artificial intelligence into medical education: a narrative review,” Ewha Med. J., vol. 48, no. 4, p. e53, Oct. 2025, doi: 10.12771/emj.2025.00787.
[2] L. Ding, T. Li, S. Jiang, and A. Gapud, “Students’ perceptions of using ChatGPT in a physics class as a virtual tutor,” Int. J. Educ. Technol. High. Educ., vol. 20, no. 1, p. 63, Dec. 2023, doi: 10.1186/s41239-023-00434-1.
[3] T. Wu, H. Lee, P. Chen, C. Lin, and Y. Huang, “Integrating peer assessment cycle into ChatGPT for STEM education: A randomised controlled trial on knowledge, skills, and attitudes enhancement,” J. Comput. Assist. Learn., vol. 41, no. 1, p. e13085, Feb. 2025, doi: 10.1111/jcal.13085.
[4] D. T. K. Ng, E. K. C. Chan, and C. K. Lo, “Opportunities, challenges and school strategies for integrating generative AI in education,” Comput. Educ. Artif. Intell., vol. 8, p. 100373, Jun. 2025, doi: 10.1016/j.caeai.2025.100373.
[5] S. K. Shankar, G. Pothancheri, D. Sasi, and S. Mishra, “Bringing Teachers in the Loop: Exploring Perspectives on Integrating Generative AI in Technology-Enhanced Learning,” Int. J. Artif. Intell. Educ., vol. 35, no. 1, pp. 155–180, 2024, doi: 10.1007/s40593-024-00428-8.
[6] M. Beege, C. Hug, and J. Nerb, “AI in STEM education: The relationship between teacher perceptions and ChatGPT use,” Comput. Hum. Behav. Rep., vol. 16, p. 100494, Dec. 2024, doi: 10.1016/j.chbr.2024.100494.
[7] Y. H. Cheah, J. Lu, and J. Kim, “Integrating generative artificial intelligence in K-12 education: Examining teachers’ preparedness, practices, and barriers,” Comput. Educ. Artif. Intell., vol. 8, p. 100363, Jun. 2025, doi: 10.1016/j.caeai.2025.100363.
[8] A. Granić, “Educational Technology Adoption: A systematic review,” Educ. Inf. Technol., vol. 27, no. 7, pp. 9725–9744, Aug. 2022, doi: 10.1007/s10639-022-10951-7.
[9] R. E. Sewandono, A. Thoyib, D. Hadiwidjojo, and A. Rofiq, “Performance expectancy of E-learning on higher institutions of education under uncertain conditions: Indonesia context,” Educ. Inf. Technol., vol. 28, no. 4, pp. 4041–4068, Apr. 2023, doi: 10.1007/s10639-022-11074-9.
[10] J.-P. Cabrera-Sánchez, Á. F. Villarejo-Ramos, F. Liébana-Cabanillas, and A. A. Shaikh, “Identifying relevant segments of AI applications adopters – Expanding the UTAUT2’s variables,” Telemat. Inform., vol. 58, p. 101529, May 2021, doi: 10.1016/j.tele.2020.101529.
[11] S. C. Kong, Y. Yang, and C. Hou, “Examining teachers’ behavioural intention of using generative artificial intelligence tools for teaching and learning based on the extended technology acceptance model,” Comput. Educ. Artif. Intell., vol. 7, p. 100328, Dec. 2024, doi: 10.1016/j.caeai.2024.100328.
[12] S. Nikolic et al., “A systematic literature review of attitudes, intentions and behaviours of teaching academics pertaining to AI and generative AI (GenAI) in higher education: An analysis of GenAI adoption using the UTAUT framework,” Australas. J. Educ. Technol., Dec. 2024, doi: 10.14742/ajet.9643.
[13] O. Filiz, M. H. Kaya, and T. Adiguzel, “Teachers and AI: Understanding the factors influencing AI integration in K-12 education,” Educ. Inf. Technol., vol. 30, no. 13, pp. 17931–17967, Aug. 2025, doi: 10.1007/s10639-025-13463-2.
[14] J. Park, T. W. Teo, A. Teo, J. Chang, J. S. Huang, and S. Koo, “Integrating artificial intelligence into science lessons: teachers’ experiences and views,” Int. J. STEM Educ., vol. 10, no. 1, p. 61, Oct. 2023, doi: 10.1186/s40594-023-00454-3.
[15] C. N. Prilop, D.-K. Mah, L. J. Jacobsen, R. R. Hansen, K. E. Weber, and F. Hoya, “Generative AI in teacher education: Educators’ perceptions of transformative potentials and the triadic nature of AI literacy explored through AI-enhanced methods,” Comput. Educ. Artif. Intell., vol. 9, p. 100471, Dec. 2025, doi: 10.1016/j.caeai.2025.100471.
[16] J. F. Price and S. Grover, “Generative AI in STEM Teaching: Opportunities and Tradeoffs,” Community for Advancing Discovery Research in Education (CADRE), 2025. [Online]. Available: https://cadrek12.org/
[17] V. Venkatesh, M. G. Morris, G. B. Davis, and F. D. Davis, “User Acceptance of Information Technology: Toward A Unified View,” MIS Q., vol. 27, no. 3, pp. 425–478, Sep. 2003, doi: 10.2307/30036540.
[18] R. K. Yin, Case study research and applications: design and methods, 6th ed. Los Angeles London New Delhi Singapore Washington DC Melbourne: Sage, 2018.
[19] S. B. Merriam and E. J. Tisdell, Qualitative Research : A guide to design and implementation, 4th ed. Jossey-Bass., 2016.
[20] J. W. Creswell and C. N. Poth, Qualitative inquiry and research design: Choosing among five approaches, 4th ed. Sage Publications, 2018.
[21] J. Saldaña, The coding manual for qualitative researchers, 4e ed. Los Angeles London New Delhi Singapore Washington DC Melbourne: SAGE, 2021.
[22] M. Q. Patton, Qualitative research & evaluation methods: integrating theory and practice, 4th ed. Thousand Oaks, California: SAGE Publications, Inc, 2015.
[23] Ministry of Education Malaysia, “Rancangan Pendidikan Malaysia 2026–2035,” Ministry of Education Malaysia, 2026. [Online]. Available: https://www.moe.gov.my/dokumen-penuh-rancangan-pendidikan-malaysia-rpm-2026-2035
[24] Economic Planning Unit, “National Fourth Industrial Revolution (4IR) Policy,” Prime Minister’s Department, 2021. [Online]. Available: https://www.ekonomi.gov.my/sites/default/files/2021-07/National-4IR-Policy.pdf
[25] Ministry of Education Malaysia, “Malaysia Education Blueprint 2013 – 2025 Annual Report 2023,” Ministry of Education Malaysia, Putrajaya, Malaysia, Annual Report, 2024. [Online]. Available: https://www.moe.gov.m
Authors
Stephanie Karmini Emmanuel Gerard
is a current student pursuing her Masters in Instructional Technology and Innovation at Universiti Putra Malaysia. She holds a Bachelor’s degree in Education (Mathematics) from Universiti Pendidikan Sultan Idris. Her research focuses on the acceptance to use Generative Artificial Intelligence among STEM teachers. She is a recipient of the Higher Education Scholarship Program (HLP) awarded by the Ministry of Education, Malaysia. She can be reached via email: ministeph86@gmail.com.
Mas Nida Md. Khambari
has a Ph.D. in Curriculum and Instruction (Educational Communications & Technology) from the University of Wisconsin-Madison. She is presently an Associate Professor at the Faculty of Educational Studies in Universiti Putra Malaysia. Her current research interest includes Information Technology, Educational Technology, Digital Agility, Teachers Professional Development, GenAi in Education, and Interest-Driven Challenge Based Learning. She can be reached via email: khamasnida@upm.edu.my.
Ahmad Fauzi Mohd Ayub
is a Professor at the Department of Foundation of Education, Faculty of Educational Studies, Universiti Putra Malaysia, Malaysia. His research interests included projects in mathematics education, educational technology, learning management systems, and multimedia education. He can be contacted at email: afmy@upm.edu.my.