Longitudinal analysis of teacher self-efficacy evolution during a STEAM professional development program: a qualitative case study

The study addresses the global push to integrate Arts and Humanities (A&H) into STEM education (STEAM) to make learning more holistic and engaging, but highlights that teachers struggle to operationalize integration in practice. To respond, the authors developed a five-stage framework for teacher professional development (TPD) focused on integrating A&H into STEM teaching (IAT), and examine its impact through a qualitative longitudinal case study. The research focuses on a pre-service mathematics teacher in China and uses teacher self-efficacy—conceptualized as beliefs about individual teaching performance and influence on student outcomes—as a key indicator of TPD effectiveness. The purpose is to document how self-efficacy evolves across staged professional learning and to inform models of teacher learning within STEAM contexts.

Theoretical background outlines multiple conceptions of STEM/STEAM and frames STEAM as integrating one or more A&H elements (e.g., history, culture, arts) with one or more STEM school subjects under a common theme while maintaining disciplinary identities. Self-efficacy, based on Bandura, and teacher self-efficacy (TSE) are defined with two dimensions: beliefs about one’s teaching performance and influence on student outcomes. Prior work shows teachers’ lower confidence outside their disciplinary comfort zone and particularly in interdisciplinary settings (integrated STEM, IAT). The study operationalizes TSE in two dimensions (individual performance and student outcomes) to suit the performance-driven Chinese context. Regarding professional learning, the literature emphasizes interconnected domains (personal, practice, consequence, external) and calls for linking teacher learning to student outcomes via real classroom evidence (Clarke & Hollingsworth; Sancar et al.). Cai et al.’s teacher professional learning model posits pathways from increased knowledge and belief change to improved instruction and student learning, but the authors argue for an update to reflect interplays between knowledge, beliefs, practice, and consequences, and to position student outcomes as feedback informing beliefs.

Design: Longitudinal qualitative case study of one pre-service mathematics teacher (pseudonym: Shuitao) over 18 months in China, within a broader multi-case project. The TPD is structured by a five-stage IAT framework aligned with policy and incorporating community of practice and lesson study. Five-stage TPD framework: Stage 1 Literature study (reading A&H materials, presenting, selecting topics); Stage 2 Case learning (co-design with in-service teachers, observation, feedback cycles, reflection, case report); Stage 3 Micro-teaching (simulated 40-min lessons, video review, self/peer/mentor feedback, iterative re-teaching; at least nine simulated sessions across three topics); Stage 4 Micro-course development (create 5-min videos, play to real students, collect feedback); Stage 5 Classroom teaching (independent IAT lesson implementation with real students, collect feedback from mentors, peers, and students). Participant: First-year master’s student in mathematics curriculum and instruction at a top Normal university in Shanghai; undergraduate major in applied mathematics; no prior IAT or teaching experience; engaged in nearly all activities. Worked in a group of four pre-service teachers. Timeline: Spanned across two academic years covering all five stages (approx. 18 months). Data collection: Seven interviews (one pre, five post-each-stage, one final); videos of micro-teaching and teaching; recordings of case development discussions; reflection notes after each stage (five), journals, and a final summary report; feedback data from students and in-service teachers. Analysis: Seven-stage inductive analysis (Merriam; Strauss & Corbin), guided by an a priori coding list aligned to two TSE dimensions (SE-IHT-IP and SE-IHT-SO). Data chronologically sorted, coded, and themes mapped to the two categories; attention to disentangling the often-intertwined dimensions. Reliability/Validity: Two independent coders (Cohen’s kappa = 0.959). Validation via participant confirmation across transcripts, replication of statements in separate interviews, and member-checking of conclusions. Multiple data sources reviewed by at least two researchers to consensus.

- Prior to TPD: Very low self-efficacy in both dimensions. The participant lacked knowledge of IAT approaches, struggled to find and organize historical materials, and doubted any positive effect on student outcomes. - After Stage 1 (Literature study): Slight increase in SE-IHT-IP. Improved ability to source/select/adapt history materials aligned to curriculum standards; modest confidence in initial steps of IAT. - After Stage 2 (Case learning): Further gains in SE-IHT-IP and emerging SE-IHT-SO. Through co-design with in-service teachers, the participant learned IAT approaches (e.g., genetic approach) and pedagogical content knowledge. Observing classroom implementations and analyzing student feedback strengthened belief in student impact. Data points: >93% of students reported increased interest in mathematics due to Napier’s story; >75% recognized that log(M+N) ≠ log_a M × log_a N, attributable to historical framing. - After Stage 3 (Micro-teaching): SE-IHT-IP improved steadily after an initial dip. Repeated simulated teaching (~10 sessions; requirement at least 9 across topics), video-based self-assessment, and mentor/peer feedback improved proficiency, fluency, classroom pacing, and specific skills (e.g., board work, language organization), enabling completion of curriculum-required teaching tasks. - After Stage 4 (Micro-course development): SE-IHT-SO strengthened. 5-min micro-courses shown to real students elicited feedback indicating improved understanding of the nature of mathematics (as evolving and human) and more positive attitudes/motivation toward learning mathematics. - After Stage 5 (Classroom teaching): Large increases in both SE-IHT-IP and SE-IHT-SO driven by mastery experiences in authentic classrooms and positive verbal persuasion from mentors/in-service teachers. The participant consistently met or exceeded curriculum objectives, witnessed students’ cognitive and affective changes, and reported substantial overall self-efficacy growth. - Overall: Self-efficacy evolved gradually and non-linearly, intertwined with developing knowledge, skills, and beliefs. Student feedback functioned as a critical source informing beliefs about impact. The five-stage framework supported cumulative, staged growth connecting theory to practice.

Findings directly address the research questions by showing that staged, practice-rich TPD for IAT produces sustained improvements in pre-service teachers’ self-efficacy in both individual performance and perceived influence on student outcomes. The growth was incremental with fluctuations, reflecting the complex, multi-level nature of TSE development interwoven with knowledge, skills, and beliefs. The study confirms the value of mastery experiences, structured feedback (from mentors, peers, and students), and iterative practice in building TSE. It extends teacher learning models by integrating Cai et al.’s pathways with Clarke & Hollingsworth’s interconnected domains, adding bidirectional links among knowledge, beliefs, practice, and student outcomes, and emphasizing student feedback as both an outcome and a feedback source shaping teacher beliefs. Practically, the results highlight effective TPD design features for STEAM: a clear content focus (IAT), emphasis on student learning evidence, strong theory-practice linkage via community of practice and lesson study, extended duration, and step-by-step scaffolding to protect and build self-efficacy.

An 18-month, five-stage TPD framework for integrating A&H into STEM teaching effectively enhanced a pre-service teacher’s self-efficacy in both individual performance and student outcomes. Growth unfolded progressively across stages through mastery experiences, iterative practice, and multi-source feedback, culminating in strong confidence after authentic classroom implementations. The study proposes an updated model of teacher professional learning in STEAM contexts, depicting continuous, cyclical interactions among teaching performance, teacher knowledge, teacher beliefs, and student learning where student outcomes feed back into teacher learning. It also synthesizes effective STEAM TPD design features: content focus, emphasis on student learning and feedback, close theory–practice connection, community of practice, extended duration, and staged guidance. Future work should quantitatively validate the framework (e.g., experimental or design-based studies), triangulate additional data sources, examine participant variability, and adapt the model across educational systems and cultures.

Single qualitative case limits generalizability; absence of pre–post or quasi-experimental design precludes quantitative confirmation of effectiveness. Participant variability was not fully reported; additional cases are needed. Although most participants improved, about 5% dropped out citing challenge/overload. Future studies should use experimental or design-based approaches, triangulate multiple data sources, and investigate contextual adaptation beyond the Chinese teacher education system.