According to the CBPL approach, high-quality curricular resources and active instruction experiences help teachers improve their knowledge, attitudes, beliefs, and teaching to improve their students’ learning outcomes (Short and Hirsh, 2020). Curriculum-driven professional development is more focused and intentional than otherwise (Wong and Tsui, 2007). Mutually supportive schools, districts, and professional development providers can further improve teachers’ teaching (Akinyemi and Nkonki, 2020).

Curriculum-based professional learning includes four design features: core, functional, structural and essential (Short and Hirsh, 2020). Following Mishra and Koehler’s (2006) technological pedagogical content knowledge (TPACK) framework for teacher knowledge, core design features (curriculum, transformative learning, equity) develop preschool teachers’ shared understanding of CT concepts and their relevance to daily life, pedagogical content teaching skills, and diverse lesson activities (both plugged and unplugged). Transformative learning experiences introduce new teaching materials (e.g., videos) that teachers use during the PDP and then with young learners in class, thereby helping teachers enhance their perceptions, beliefs, and self-efficacies. To promote equity, the PDP uses culturally and developmentally appropriate integration of CT into ECE.

Functional design features of PDP includes learning designs, reflection/feedback, beliefs, and change management. The learning designs introduce CT integrated curriculum with peer teachers, who implement and mirror pedagogical strategies with digital tools to teach CT. Reflection and feedback enhances teachers’ practical skills with these tools. As teachers’ competences grow, PDP cultivates change in their teaching beliefs. PDP also helps teachers become change agents and manage their changes to enhance their CT integrated classroom practices and scaffold student learning.

Structural design provide regularity and increase consolidation of learning. Grounded in constructionist learning theory (Csizmadia et al., 2019), teachers collaborate to plan, implement, and assess new CT teaching materials and lessons for effectiveness and sustainability. Over time, they adapt their teaching practices to their students’ needs.

Essential elements are leadership, coherence, and resources. PDP helps groups of leaders and teachers develop a shared, long-term vision for school-wide CT integration. Furthermore, they build networks with other schools implementing CT for mutual sharing and support. With sufficient resources, they can build sufficient material (e.g., Beebots) and staff capacity for CT implementation.

The CBPL training course includes six workshops across 6 months and a summer training institute for participants to apply their workshop ideas to teaching students. The first workshop helps teachers understand the critical dimensions of Brennan and Resnick’s (2012) CT framework: key ideas (e.g., iteration, parallelism; computational concepts), computer scientist actions (e.g., debugging projects; computational practices), and computer scientist views of the world (computational perspectives). In this workshop, teachers learned about CT components via daily life problem-solving activities and a related short video. They learned terms such as “loops,” “conditionals,” and “variables” within simple computer program structures. Then, they explored links between CT components and the ECE curriculum via brainstorming and discussion. Next, they reflected on CT’s educational purposes and made recommendations for excellent CT instruction. We encouraged their use of CT within a suitable educational philosophy and ECE learning environment. The first workshop promoted open discussion around questions such as these (rather than simply giving answers):

These teachers considered five CT uses.

The second workshop enhanced teachers’ pedagogical understanding of computational concepts, practices, and perspectives via hands-on practice. Teachers examined data practices, modeling, simulation practices, computational problem-solving practices, and systems thinking activities. To understand and apply these practices to ECE, they collaborated in groups to solve CT problems involving the programming of Beebots, Cubetto, Robo_Wunderland, and ScratchJr (artifact-based learning, Bairaktarova et al., 2012).

These teachers also developed “activities … to engage a variety of audiences with great ideas from computer science and CT, without learning coding, programming or even using a digital device” (unplugged CT, Bell and Vahrenhold, 2018, p. 497; T. Bell et al., 1998). These activities incorporate games, puzzles, stories, challenges, and rich sensory experiences (see examples at Hello Ruby; code.org; ACARA; CS Unplugged, Bell, 2021). Unplugged activities are inexpensive, accessible, and require few resources (Lee and Recker, 2018). Unplugged CT promotes its development by complementing programming, integrating it into other topics, and scaffolding teachers and students, serving as a “priming phase” before programming to increase students’ awareness of algorithmic stages (Huang and Looi, 2021). By using familiar materials and concepts, unplugged CT such as paper circuitry activities “provide multiple entry points for students less familiar with computational thinking concepts” to learn algorithmic thinking, conditional logic, debugging, etc. (Lee and Recker, 2018). Lastly, these teachers wrote reflections on how their understanding and teaching of CT changed.

The third workshop cultivates teacher beliefs to support their CT teaching via group exploration and application. As participating teachers discuss, understand, implement, and assess CT teaching practices, they appreciate their effectiveness and sustainability, which enhances their beliefs and attitudes about effective CT teaching (Bower et al., 2017; Yadav et al., 2017; Corradini et al., 2018). Notably, they increase their belief in their capacity to act and accomplish teaching goals (teaching self-efficacy, Bandura, 2000). They created a bank of design scenarios and resources for teaching and assessing the CT learning of young children. Challenged to develop suitable CT activities for ECE, these teachers created compelling practical application scenarios.

We encouraged these teachers to use a scientific approach to these activities that encourages students to identify a problem, present a hypothesis, conduct an experiment, analyze the results, discuss their perspectives, generalize the findings, and revise accordingly (Merrill, 2002). Below are examples for various age groups based on their usefulness, availability, affordability, and variety in the local market (not tied to any one robotic platform):

Rather than prescribe best practices, our course sought to equip teachers with tools to identify best resources and practices for each plugged or unplugged activity. For example, we revised Kemp’s model of instructional design (Kemp, 1985) to highlight nine key elements in each activity’s development:

In the fourth and fifth workshops, groups of teachers collaborated to iteratively design shared CT curricula and lessons for ECE based on their school’s available resources and culture; and the capabilities, dispositions, and predispositions of both teachers and students (Seow et al., 2019; Kotsopoulos et al., 2021). In this iterative design, they clarified their representative modeling, CT processes and errors, and students’ errors, and then used this information to improve their lessons and curricula (Fields et al., 2019). Based on Kemp’s (1985) framework, we designed the CT integration resource bank so that its structure emphasizes the critical CT components for a lesson, including suitable in-house plugged, unplugged and web resources. Teachers in the same grade level collaboratively created lesson plans for CT integration; they identified supportive resources for the CT content, designed a lesson plan, used available materials, and presented it to other teachers (with videos, PowerPoint presentations, documents, and photographs). These complex lesson plans:

The fourth and fifth sessions prepared teachers to design a CT instructional project following theoretical orientation (workshop 1), practical training (workshop 2) and exploration (workshop 3). The research team evaluated and analyzed each project, and gave the teachers feedback to help them reflect on possible improvements.

The sixth workshop focused on teacher feedback and a CT integrated curriculum. Specifically, teachers discussed key obstacles to creating and teaching an effective lesson plan. They shared CT teaching experiences, practical examples, and results for young children at various ages.

Using a bottom-up approach, teams of teacher representatives led the other participants to create a CT integrated curriculum (Lassonde and Israel, 2009) and institutionalize it. Teacher representatives developed further learning and teaching resources (plugged and unplugged) to enrich 2 to 6-year-old students’ learning activities, conduct reviews of activities associated with age-appropriate CT integration (Bers et al., 2014), shared best practices (Melasalmi and Husu, 2016), and enhanced professional development for colleagues (Kermani and Aldemir (2015). Also, they expanded their teacher networks within and across schools, and strengthened partnerships with community key players to attract support and additional resources for long-term planning (Palts and Pedaste, 2020).

As a result of these CBPL workshops, we expect teachers to improve their CT knowledge, beliefs, attitudes and self-efficacies.

We used a convenience sample. Of the eight invited kindergartens, five kindergartens consented to conduct this PDP. Due to COVID-19 delay, only two kindergartens participated in this 6-month, school-site training programme at different times (February–July 2021: 13 teachers, 1 principal, 1 curriculum coordinator; June–December 2021: 13 teachers, 1 principal, 1 curriculum coordinator). Both kindergartens served 2.2 to 6-year-old students. Figure 1 explains our participant recruitment strategy and study protocol.

All 25 teachers are qualified and registered ECE teachers in Hong Kong (see demographics in Table 1) and consented to participate in this research.

All 25 teachers participated in six, 2-h CT PDP workshop sessions, and then taught CT to students aged two to 6 years during either a 6-week summer school or a 2-week camp. All of them wrote reflections at the end of each workshop. All participants confirmed that they did not engage in other workshops during this time. All teachers responded to pre-workshop surveys, and 21 of them completed the post-workshop survey. All teachers were interviewed after the sixth workshop.

Statistical power differs across levels. For α = 0.05 and an effect size of 0.4, statistical power for 25 people is 0.52 and for 46 survey completions across time is 0.80 (Konstantopoulos, 2008). As this small sample has low statistical power, false negatives are likely, but significant results remain valid.

Accurate analyses of these data must address issues involving outcomes and explanatory variables (see Table 2).

Outcome issues include differences across teachers or across time, discrete outcomes and multiple outcomes. As completed surveys at the same time or by the same teacher likely resemble one another more than those completed at different times or by different teachers (cross-classified data), an ordinary least squares regression underestimates the standard errors, so we use a multilevel cross-classification analysis (Goldstein, 2011). For ordered discrete outcomes (e.g., ratings), ordinary least squares regressions can bias the standard errors, so we use an ordered Logit regression (Kennedy, 2008). To aid understanding of these results, we report the odds ratio of the regression coefficient, namely the percentage increase or decrease in the likelihood of the outcome (Kennedy, 2008). Multiple outcomes can have correlated residuals that underestimate standard errors, so we use multivariate outcome multilevel analysis (Hox et al., 2017).

Explanatory variable issues include omitted variable bias, indirect effects, cross-level interactions, many hypotheses’ false positives, effect size comparisons, and robustness. As unspecified, omitted explanatory variables can cause omitted variable bias (Kennedy, 2008), we use a difference-in-differences analysis of participants at different times to reduce this bias (Bertrand et al., 2004). Separate, single-level tests of indirect mediation effects on cross-classified data can bias results, so we test for simultaneous multi-level mediation effects with a multilevel M-test (MacKinnon et al., 2004).

With cross-classified data, incorrectly modeling interaction effects across levels (e.g., survey response × teacher) can bias the results, so we use a random effects model (Hox et al., 2017). If an explanatory variable’s regression coefficient (e.g., β_yvj = β_yv0 + g_yvj) differs across levels (g_yvj ≠ 0?), then we model the possible cross-level moderation with structural variables (e.g., teacher).

As testing many hypotheses can cause false positives, we reduce them via the two-stage linear step-up procedure; it outperformed 13 other methods in computer simulations (Benjamini et al., 2006). When testing whether effect sizes differ, Wald and likelihood ratio tests do not apply at boundary points, so we use Lagrange multiplier tests, which also have more statistical power for small deviations from the null hypothesis (Bertsekas, 2014).

Lastly, we test whether the results remain stable (robust) despite small data or analytic differences (Kennedy, 2008). As any mis-specified equation in a multiple outcome model can introduce errors into otherwise correct equations, we model each outcome variable separately. Then, we separately run subsets of the data.

We conducted thematic analysis of the written reflections, applying initial coding, axial coding and theoretical coding to identify major patterns and themes (Braun and Clarke, 2006). In the initial coding, data were coded by keywords or phrases (e.g., algorithm, if-then). In the axial coding, the data were coded into categories (e.g., beliefs) and sub-categories (all children should learn CT; only technically-inclined children should learn CT). In the theoretical coding, data were compared with pre-existing models of PDP effects on CT knowledge, beliefs, attitudes, and teaching self-efficacies (e.g., Rich et al., 2021b).

We analyzed the semi-structured interviews (Fraenkel et al., 2015) with Creswell’s (2012) six-step method. First, we anonymized and converted all interview transcripts to text. Then, the lead author read the interview transcripts several times to understand the interviewees’ ideas and tone. Initial codes were developed and then iteratively refined (Fraenkel et al., 2015). Next, the codes were evaluated, and some were merged while others were deleted. Then, themes were developed and organized into a hierarchy of codes. Tables and figures were created to illustrate the findings. Finally, findings were examined, and thick description and inter-coder reliability tested the trustworthiness of the interpretations.

These teachers’ survey responses regarding CT definition, beliefs, attitudes, knowledge, and CT teaching self-efficacies showed vast differences before vs. after the workshops. After the workshops, these teachers were far more likely) to define CT as including algorithmic solution of a problem, analytic ability, solving problems as a computer scientist, logical problem solving, or problem solving (+47% or higher odds ratio for each, see Table 4). After the workshops, these teachers were much more likely to know what CT is (+44%), know steps to teach CT effectively (+36%), know CT enough to teach it to young children (+48%), and know unplugged CT enough to teach it (+49%). After the workshops, these teachers were also far more likely to believe that all students should learn CT (+48%) and that effective teaching helps students learn CT (+35%). Their following attitudes also vastly improved after the workshops: interest in integrating CT into their teaching (+49%), continually find better ways to teach CT (+48%), and want to attend more training about teaching practices (+33%). After the workshops, teachers reported greater overall CT teaching self-efficacy (+44%), plugged CT teaching self-efficacy (+47%), and unplugged CT teaching self-efficacy (+42%).

CT teaching self-efficacy often differed more across teachers (teach with CTA overall: 62%; teach with plugged CTA: 64%; teach with unplugged CTA: 50%) than across time (overall CTA: 38%; plugged CTA: 36%; unplugged CTA: 50%; see Table 5). (See Appendix Table A1 for correlation-variance-covariance matrices). All results discussed below describe first entry into the regression, controlling for all previously included variables. Ancillary regressions and statistical tests are available upon request.

The result suggest that these workshops changed not only these teachers’ CT knowledge and CT attitudes but also their CT beliefs and CT teaching self-efficacies.

School, workshop completion and teacher’s CT knowledge were all linked to CT teaching self-efficacies. In the school with a greater sharing and teamwork culture or sufficient resources to support CT teaching, teachers reported greater overall and plugged CT teaching self-efficacies. These results suggest the importance of school culture and resources for supporting CT teaching. Also, workshop completion was linked to greater unplugged CT teaching self-efficacy. Unlike past studies of shorter PDPs (Rich et al., 2021b), this longer PDP increased teachers’ unplugged CT self-efficacy. Lastly, teachers who reported greater CT knowledge reported greater overall, plugged, and unplugged CT teaching self-efficacies, cohering with past studies regarding the importance of teacher CT knowledge for CT teaching self-efficacy (Rich et al., 2021b).