1 Introduction

The pervasiveness of computing technologies makes it imperative to train the future workforce to solve problems using computational skills. Hence, computer science (CS) education is important for K-12 schools starting from early grades (Avcı & Deniz, 2022; Bers et al., 2014). One target skill is computational thinking (CT), which entails leveraging CS principles to devise efficient, adaptable, and reusable solutions for problems (Weber et al., 2022; Wing, 2006, 2011). Exposure to CT starting from early childhood can generate foundational interest among students and invite underrepresented groups to join and diversify CS and STEM (science, technology, engineering, and mathematics) fields (Beyer, 2014; Gretter et al., 2019).

We argue that CT can be integrated into teacher education programs across multiple disciplines to promote STEM education (Vasconcelos, 2024; Yadav et al., 2016). CT works as a versatile toolkit whose subcomponents transcend disciplinary boundaries (Li et al., 2020; Yadav et al., 2016). Hence, teacher educators can foster CT and problem solving across disciplines in ways that promote decomposition of problems into smaller parts, recognition of patterns, abstraction of general principles, creation of step-by-step solutions, and iterative troubleshooting of prototypes (Shute et al., 2017; Wing, 2008, 2011). One approach is to adopt developmentally appropriate technologies that are suitable for young kids, such as robotics kits (Papadakis, 2020). Robots allow CT through physical and cognitive experiences, that is, the robot manipulation and the construction of pseudo programming languages that are designed for kids.

However, opportunities for CT instruction in early childhood teacher education programs are scarce. The issue arises in part because CT interventions for preservice teachers frequently emphasize higher grade levels, under the mistaken belief that CT is primarily a skill for older K-12 students. Scarce opportunities to practice CT may leave preservice teachers with negative attitudes and worries about their ability to teach CT (Gal-Ezer & Stephenson, 2010; Ling et al., 2017). This study sought to investigate the effect of a CT module on early childhood preservice teachers’ attitudes and intentions toward teaching CT.

2 Related literature

2.1 Computational thinking

CT has been defined as a problem-solving skill that involves leveraging computational practices to devise algorithmic solutions in ways that can be automated or implemented by an intelligent agent (Shute et al., 2017; Weber et al., 2022; Wing, 2006, 2008, 2011). The CT literature abounds, and it often mentions five key computational practices. First, CT includes decomposing problems into smaller tasks to be addressed individually and more easily. Problem decomposition is the process of providing structure to what is ill-structured, that is, examining a complex problem and figuring out ways to strategically and systematically tackle it in parts (Kwon & Cheon, 2019; Melro et al., 2023; Rich et al., 2019). Second, CT entails creating a map or a blueprint with sequentially organized algorithmic steps that can be implemented as the solution to a problem (Kirçali & Özdener, 2023; Wong et al., 2024). This is called algorithmic thinking, and the algorithms are structured in ways that show possible paths and decisions to be activated given a set of predetermined inputs. Third, CT involves abstraction and manipulation of information that is relevant to the problem at hand while ignoring irrelevant information (Cetin & Dubinsky, 2017; Gün-Tosik & Güyer, 2024; Qian & Choi, 2023). This is called abstraction, and it is essential as one switches focus to address different aspects of a problem or to address the problem at different levels (e.g., at the micro level by reviewing algorithms step by step or at the macro level by reviewing the entire algorithmic solution). Fourth, CT entails recognition of patterns across similar problems to apply similar algorithmic solutions to address them (Basawapatna et al., 2011; Chen et al., 2023; Yasin & Nusantara, 2023). Recognition of patterns occurs through activation of past memories and experiences to activate knowledge about strategies for problem solving. And fifth, CT entails debugging, that is, detecting and resolving errors in algorithmic solutions (Kim et al., 2018, 2022; Sun et al., 2024; Vasconcelos et al., 2020) to ensure effectiveness, to improve efficiency, and to optimize solution implementation.

While CT is often considered a computer programming skill, its application extends beyond programming (Barr & Stephenson, 2011; Yadav et al., 2022). We argue that CT is a cross-disciplinary set of skills that can be combined with content from different disciplines (Günbatar & Bakırcı, 2019) to show learners meaningful, interdisciplinary, and real-world applications (Alqahtani et al., 2022; Çiftçi & Topçu, 2023). CT skills are inherent to problem solving activities with scientific modeling experiments, structured engineering design projects, algebraic equations, and technology-enhanced experiences. Specifically, we propose using CT as a foundation for the design and implementation of STEM learning activities that follow the 5E Learning Cycle (Bybee, 2018). Each activity begins with a problem to engage students in a task, followed by opportunities for students to explore and test different problem-solving solutions, provide explanations linked to CT approaches, and evaluate CT implementation. At the culmination of each activity, a discussion promotes reflection about how each CT subskill is operationalized across the 5Es, how CT is connected to science and engineering practices, and how activities can be adapted and modified for students based on grades, resources and specific student needs.

2.2 CT with robots

Educational robots are tools to facilitate hands-on activities through active exploration, prototyping, iterative testing, and problem solving. These types of robots are often controlled with a pseudo programming language whose symbols (e.g., physical buttons, stackable digital blocks) control robot navigation (Misirli & Komis, 2023). Controlling the robot, stacking blocks of code to create complex commands, watching robot movements, and tweaking the code blocks encourages both hands-on and mental abilities (Fegely & Tang, 2022; Zhang et al., 2021). Particularly, the act of physically manipulating the robot and its pseudo programming language and the sensory stimuli that the robot movement generates assists learners in developing problem solving skills while creating notions of abstract ideas (Bers et al., 2019; Ching & Hsu, 2024). Further, engaging in discussions about building robots can enhance STEM learning by applying principles of engineering design (Yuan et al., 2022), utilizing computer science principles such as iterative design and debugging (Kim et al., 2018, 2022), facilitating scientific inquiry and discovery-based experiences (Vasconcelos et al., 2024), and promoting mathematical literacies (Zha et al., 2022).

2.3 CT for early childhood preservice teacher education

Lavigne et al. (2022) note the value of CT beginning in early childhood, stating that CT should build through children’s play, enabling young learners to engage in sequencing activities that involve breaking down problems into smaller, more manageable tasks, exploring patterns, and finding and fixing mistakes. CT instruction has been shown to increase executive function in young learners, a key indicator of later academic success (Robertson et al., 2020). Promoting CT skills with robots to young children can foster development of collaboration skills through peer-to-peer interactions (Burleson et al., 2018; Pugnali et al., 2017), embodied learning strategies during problem solving (Baek et al., 2019; Kopcha et al., 2021), and development of CT skills (Angeli & Valanides, 2020; Bers et al., 2014, 2019; Pugnali et al., 2017; Taylor & Baek, 2019).

Equipping preservice teachers with the skills to integrate CT into their future teaching is essential. Through the use of the 5E learning cycle, preservice teachers can explore ways to integrate CT into existing instruction and promote ongoing science and engineering practices (Gao & Hew, 2022). The first step in the 5E learning cycle is to engage students in relevant problem-solving tasks that require CT to solve them, while sparking interest in the topic. For example, what is the best route to get the required resources for the animals in the zoo? Students can then explore through hands-on activities to test different ideas, where they decompose the problem into smaller steps, determine different approaches, and test different solutions (algorithmic thinking). After exploration, students engage in explaining their approaches, looking at their findings to determine the best approach (argue from evidence), identify ways to revise (debug and remix), and present and communicate their ideas to others (evaluate). By exposing preservice teachers to robots suitable for various age groups, from early childhood options like Bluebot, Talebot, and Ozobot to more advanced choices like RoboRobo, Lego Mindstorms, and Spheros, integrated activities can help educators understand CT from a student’s viewpoint. To program a robot, preservice teachers break down the main objective into smaller tasks that are easier to achieve, concentrating on relevant programming functions needed for each step while disregarding unnecessary ones. They design step-by-step sequences to control the robot’s movement, apply similar code segments for repeated movements, and utilize robot movements to iteratively identify and correct block-based code errors until the robot executes the intended movement.

However, the gap between the demand for CS education in schools and the opportunities for preservice teacher preparation is substantial (Yadav et al., 2014; Zhu & Wang, 2024). Many teacher preparation programs lack introduction to CS courses. Consequently, preservice teachers may graduate feeling unprepared and lacking confidence to teach CT (Weber et al., 2022; Yadav et al., 2022). Studies on CT for preservice teachers are limited, especially those targeting early childhood grades (Avcı & Deniz, 2022; Çiftçi & Topçu, 2023). CT in computer science is not an isolated skill but rather a fundamental skill that enhances learning across all academic disciplines. Developing CT strategies can equip students with essential problem-solving skills that are crucial for navigating the digital world, fostering critical thinking, creativity, and adaptability, while also laying a foundation for future STEM learning.

2.4 Promoting positive CT attitudes and intentions

Promoting positive attitudes and intentions toward teaching CT among preservice teachers is critical. It is expected that those who hold positive attitudes are more likely to try to infuse CT into their classrooms (Fessakis & Prantsoudi, 2019). Research has shown that preservice teachers and K-12 teachers believe CT instruction is valuable (Barlow et al., 2023; Fessakis & Prantsoudi, 2019; Yadav et al., 2022) especially given its potential to promote critical dispositions for problem solving within and beyond CS such as “persistence, resilience, and the ability to collaborate with others” (McGinnis et al., 2020, p. 95). Further, preservice and K-12 teachers have reported being interested in integrating CT into the classroom across disciplines within and beyond STEM (Fessakis & Prantsoudi, 2019; Yadav et al., 2022). However, while many are willing to learn computational skills, they often feel unprepared to teach CT practices (Bower & Falkner, 2015; Fessakis & Prantsoudi, 2019; Yadav et al., 2022). Preservice teachers who feel unprepared may develop negative attitudes and resist teaching CT for various reasons that range from CT being too challenging for certain groups of students, to curriculum constraints, to limited time and support from schools (Cabrera et al., 2019).

Professional learning on CT within teacher education courses can promote positive attitudes toward CT instruction (Yadav et al., 2022) and reduce anxiety toward teaching CT in conjunction with other subjects through integrated STEM learning (Weber et al., 2022). In fact, CT was found to have the most significant effect on preservice teachers’ intentions to implement STEM learning experiences (Günbatar & Bakırcı, 2019). However, professional learning needs to be thoughtfully designed in order to promote positive attitudes. Specifically, teacher educators need to engage preservice teachers in experiences that promote a more refined comprehension of CT which is achieved through first-hand practice (Avcı & Deniz, 2022; Kaya et al., 2019; Mason & Rich, 2019; Rich et al., 2021). Further, professional learning should be designed with CT as an element that is already embedded in the curriculum—as opposed to an add-on or an adaptation yet to be performed to the curriculum – which can reduce preservice teachers’ perceived challenges in teaching CT (McGinnis et al., 2020). In addition, professional learning should offer CT-integrated STEM experiences, which was found to strengthen preservice teachers’ self-efficacy beliefs toward teaching CT (Çiftçi & Topçu, 2023). Finally, preservice teachers should be exposed to CT instruction starting in the first years of their teacher education, and have opportunities to ponder the impact of CT instruction for them as educators as well as their students (Akcaoglu et al., 2023).

2.5 The literature gap

Most studies on CT for preservice teachers focus on elementary and secondary grades. This is not surprising given that CT is often overlooked in early childhood education (Fessakis & Prantsoudi, 2019). This is concerning because unprepared early childhood preservice teachers miss opportunities to identify instances in which young children engage in CT practices during unstructured play (Avcı & Deniz, 2022), with or without technologies, and across different subject domains. This study addressed this literature gap by introducing CT to early childhood preservice science teachers.

3 Theoretical frameworks

Two theoretical frameworks served as the foundation for the intervention design and data analysis in this study: constructionism and the theory of planned behavior.

3.1 Constructionism

Constructionism, developed by Seymour Papert, builds on Piaget’s constructivism but places greater emphasis on learning through the creation of tangible artifacts. Piaget’s constructivism explains how individuals construct knowledge through experience as they build and refine mental models about the world. Individuals either integrate new information into existing schemas (assimilation) or modify schemas based on conflicting information (accommodation) (Piaget, 1952). Hence, Piaget’s theory focuses on cognitive and internal processes. Papert’s constructionism extends this theory by highlighting learning through externalization of knowledge during hands-on activities. Rather than passively absorbing information from a more knowledgeable entity, such as a teacher or a textbook, learners actively engage in constructing their own understanding through meaningful and creative processes (Harel & Papert, 1991; Holbert et al., 2020; Kafai, 2012; Papert, 1980). Specifically, constructionist learning involves self-driven inquiry, creation of tangible or virtual artifacts that are personally relevant, and collaboration with peers and/or stakeholders. The construction of artifacts induces creativity in the design and development of artifacts, as well as critical thinking and problem solving to learn from mistakes and iteratively improve artifacts (Broza et al., 2023; Yang & Chang, 2013). Constructionism is a suitable approach to support CS education for preservice teachers who engage in coding projects, design and implement projects that are relevant to them, and collaborate with others while using technologies (e.g., robots, simulations) that are controlled by a programming language (Vasconcelos, 2024).

3.2 Theory of planned behavior

The Theory of Planned Behavior was developed to explain human behavior using psychological principles. This theory considers that intention is the direct precursor to human behavior, and intention is affected by three key factors (Ajzen, 1991, 2012, 2020). First, attitudes refer to an individual’s overall assessment of a target behavior, encompassing both positive or negative perspectives on the behavior itself and its potential outcomes. Attitudes involve a cognitive component, such as beliefs about consequences of the behavior, and an affective component, such as emotional responses to the behavior. When an individual perceives the behavior as beneficial, enjoyable, or consistent with their values, they are more likely to perform it. Conversely, negative attitudes can reduce intentions to perform the behavior. For example, preservice teachers who hold positive attitudes toward learning to teach CT are more likely to integrate it into their lesson plans.

Second, subjective norms consist of social influence to perform a target behavior, and these norms are shaped by the expectations of significant others or broader societal standards. Specifically, subjective norms reflect perceptions of what others value or expect regarding the behavior, as well as the individual’s willingness to align with those expectations. In practice, an individual is more likely to adopt a behavior if they believe significant others approve of or expect it. Conversely, disapproval from influential people may weaken intentions toward behavior. A preservice teacher may have stronger intentions to teach CT if their course instructor or school culture endorses its importance.

Third, perceived behavioral control entails an individual’s perceived ability to perform the behavior, taking into account factors that may facilitate or hinder it. Perceived behavioral control encompasses perception of available resources, existing obstacles, and confidence in their own capability. When perceived behavioral control is high, both intention and the likelihood of performing the behavior increase. Conversely, when individuals face external barriers or lack self-confidence, their intention and ability to perform the behavior decrease. A preservice teacher who feels unprepared or lacks the technical skills or resources to teach CT may have weaker intentions to do it, even if their attitudes and subjective norms are favorable.

It is essential to recognize that these variables do not act independently. Rather, they interact to shape behavioral intention (Ajzen, 1991, 2012, 2020). For instance, a preservice teacher may have positive attitudes toward CT and perceive strong social approval (subjective norms) but they might still hesitate to implement it if they perceive a lack of control over the necessary skills and resources.

4 Purpose, hypothesis, and research questions

This study investigated changes in early childhood preservice teachers’ attitudes toward learning to teach CT and their intentions to teach CT in early childhood. We hypothesized that Robotics in Early Childhood STEM Education (REC-STEMEd) leads to statistically significant positive attitudes and more positive intentions toward teaching CT. The following were the research questions:

  • Do preservice teachers’ attitudes toward learning to teach CT in early childhood change after REC-STEMEd?

  • What are preservice teachers’ intentions toward teaching CT in early childhood after REC-STEMEd?

5 Methods

5.1 Research design

This study adopted a pre/post design that aimed at evaluating the effects of the intervention on participants’ attitudes and intentions without a control group or random assignment to groups (Campbell & Stanley, 1963; Thyer, 2012).

5.2 Setting and participants

Participants attended three sections of an in-person course on early childhood science teaching offered at a Southeastern United States university. This course prepares early childhood preservice teachers to become generalists who are often responsible for teaching multiple subjects. Therefore, providing experiences that foster cross-disciplinary learning is essential to help them move beyond disciplinary silos. Institutional research approval and informed consent were obtained before the study. Participants were 39 female preservice teachers in their Senior year of college. Due to smaller-than-usual cohort sizes, participants from all three course sections were treated as a single group. This participant sample included 27 White participants, four African American, three Asian, three Latinas, two multiracial, and one Native American. They were 21.82 years old (SD = 1.62) on average. No participant had robot programming experience.

5.3 The intervention: REC-STEMEd

One way to integrate CT into preservice teacher preparation is with robot programming. Developmentally appropriate robots for young children can promote sequencing skills (Kazakoff et al., 2013), fine motor skills and design thinking (Bers et al., 2014), as well as discipline-specific content knowledge (Alqahtani et al., 2022). Robots enable learners to engage in low-threshold tasks that simplify CS concepts (e.g., delays, loops) using programmable blocks (Kim et al., 2018).

This study implemented REC-STEMEd, which consisted of three 90-min sessions and one online module that spanned across 4 weeks. The module on CT featured synergistically integrated science and CT activities for early childhood preservice teachers (Table 1). The sequence of experiences was intentionally designed based on the cognitive theories of Jean Piaget (1952). First, participants were introduced to CT as a cross-cutting skill that entails problem decomposition, pattern recognition, abstraction, algorithm design, and debugging. Participants were prompted to reflect on how they apply CT in their daily lives, helping them connect the concept to their own pre-existing schemas. For example, they considered algorithmic steps as instructions to navigate the university campus.

Table 1 REC-STEMEd learning module
Full size table

Second, participants engaged in unplugged CT activities. In small groups, one participant provided peers with directions for drawing an unknown geometric shape. Afterwards, they reflected on how to make the step-by-step instructions shorter and more efficient. Beginning with unplugged activities enabled participants to practice CT in a concrete and familiar way before shifting to more abstract technology-based representations. Moreover, this activity enabled participants to reflect on the early childhood stages of cognitive development (pre-operational stage, ages 2–7), where young learners are still developing skills to communicate and understand abstract concepts.

Third, participants practiced CT with robots, which helped them transition from concrete, egocentric activities to more symbolic forms of communication. Before each robotics activity, participants were provided hands-on instruction on how to use the robots and how to teach with the robots in early childhood. Participants were presented two problems that integrated science content and robot programming:

  • They were tasked with guiding a Bluebot, a small programmable robot, representing a zookeeper moving on a zoo map to deliver the appropriate food to different animals (Fig. 1, left). To control the Bluebot, participants used its built-in directional buttons to input commands such as move forward, turn right, turn left, and move backward, constructing an algorithmic sequence. This process entailed planning step-by-step movements, predicting outcomes, and debugging errors when needed. Participants had the option to use directional cards before or during the algorithm design phase. These cards served as a visual aid, allowing participants to map out their planned sequences before programming the Bluebot. Many participants used the cards to strategize their movement paths, test ideas, and refine their algorithmic sequences. This activity enabled participants to practice CT with technology without requiring knowledge of a programming language.

  • Participants programmed an Ozobot, a small programmable robot, to practice principles of responsible recycling. This task required participants to use block-based coding to program and guide it through a map. The goal was to sort different types of waste – paper, plastic, metal, and organic material – into the appropriate bin (Fig. 1, right). To accomplish this, participants used Ozoblockly, a block-based coding platform that allows users to snap together blocks, each embodying a specific programming command. By assembling these blocks, participants created an algorithmic sequence of instructions for the Ozobot. They iteratively created, tested, and debugged their code to ensure the Ozobot followed the correct recycling path. This activity introduced participants to more advanced CT concepts using a pseudo programming language.

Fig. 1
figure 1

CT activities with Bluebot (left) and Ozobot (right)

Full size image

Finally, participants discussed their experiences with CT across all activities and brainstormed ways to integrate CT into early childhood science teaching. They reflected on the CT skills they had developed, comparing the successes and frustrations across the various types of CT activities. Further, they brainstormed ideas and strategies for teaching CT in their classrooms during the internship program.

5.4 Instruments

Two quantitative instruments were adopted in this study. The Attitudes toward Learning to Teach Computational Thinking Survey, adapted from Abun et al.’s (2019) Attitude toward Research Survey, used a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). This instrument was chosen because it accounts for both cognitive and affective factors as predictors of behavioral intentions. Particularly, the instrument contains 20 items divided into four subscales: positive affection, negative affection, positive cognition, and negative cognition. The items were minimally modified for this study by replacing the word research with learning to teach computational thinking. For example, the item “research makes me nervous” (Abun et al., 2019, p. 228) was modified to “learning to teach computational thinking makes me nervous.” These minor modifications did not change the test’s original structure and subscales. Moreover, one item from the positive cognition was removed since it was not applicable to the aim of this study.

The Intentions for Teaching Computational Thinking Survey, adapted from Moses et al.’s (2020) Theory of Planned Behavior Questionnaire for Students’ STEM Career Choices, used a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). The instrument contains 26 items divided into four subscales: attitudes, subjective norms, perceived behavioral control, and intentions. The survey items were slightly modified for this study by replacing the expression to choose a career in STEM by to teach computational thinking. For instance, the item “I intend to choose a career in STEM” (Moses et al., 2020, p. 300) was modified to “I intend to teach computational thinking.” Further, the subjective norms subscale in the original survey included 15 items related to the influence of teachers, parents, and friends on participants’ intentions. However, these populations were not relevant to our study about CT. So, we adapted the items to measure the influence of the course instructor and the study participants’ teaching partner instead. These minor modifications neither changed the structure nor the subscales of the instrument.

5.5 Strategies for validity and reliability

A few strategies were adopted for enhanced validity and reliability in research findings. First, both quantitative instruments were selected because their items align with the Theory of Planned Behavior. Hence, these instruments allowed theoretical and methodological alignment with latent variables in this study (Alwin, 2010). Second, the authors strived to make minimal modifications to the instruments to maintain reliability. Third, the complete modified instruments (Appendix 1 and 2) were reviewed by researchers with experience on CT and quantitative measurement for face validity (Mosier, 1947). Fourth, both instruments had been piloted before: the attitudes survey with college students and the intentions survey with teenagers. The Intentions for Teaching Computational Thinking Survey showed high reliability coefficients (Cronbach, 1951; DeVellis, 2017) across all subscales of at least (α = 0.73). Fifth, we ensured that key demographic characteristics which could potentially skew the study findings were considered to prevent participant selection bias (Baldwin, 2018; Dunbar-Jacob, 2012; Slack & Draugalis, 2001). Specifically, no study participant had prior experience in CT or robot programming to prevent competing factors or confounding variables. Sixth, we ensured strong alignment between the definition, representation, and relevance of presented constructs and instruments (Mosier, 1947; Sireci, 1998). Specifically, we designed learning activities that consistently reinforced the same core CT constructs—problem decomposition, algorithm design, abstraction, pattern recognition, and debugging—within the context of early childhood science learning. Along these lines, the data collection instruments were adapted to align them with those target constructs. Seventh, we ensured that the intervention activities were conducted as close as possible to the original research plan across the different participant cohorts to maintain fidelity of implementation (O’Donnell, 2008; Stains & Vickrey, 2017). To achieve this goal, all course sections were taught by the same instructor, all activities were aligned with predetermined learning standards, and at least two research team members were present in every session to ensure consistency in methodological procedures across all cohorts.

5.6 Data analysis

After checking the normality of the subscales and identifying that they were normally distributed, separate paired samples t-tests were run to detect if there were statistically significant differences (Field, 2013; Nolan & Heinzen, 2012) between the participants’ pre- and post-attitudes toward learning to teach CT in early childhood for each subscale. Only 29 participants’ data were included in the paired samples t-tests since 10 participants did not complete either the pre-survey or the post-survey. Also, descriptive statistics, specifically mean and standard deviation, were calculated to represent the differences noted for each subscale of the attitudes after the intervention and to determine the preservice teacher’s intentions for teaching CT survey. Furthermore, the mean and standard deviation of each item within the subscales of the Intentions for Teaching Computational Thinking Survey were reported to highlight key trends within each subscale. This analysis provides a deeper understanding of the specific factors influencing preservice teachers’ intentions to teach CT. These measures of central tendency and variation, respectively, helped summarize the typical value and the extent of variability within the data set (Field, 2013).

6 Findings

6.1 Attitudes toward learning to teach CT

Paired samples t-tests showed that there is a statistically significant difference between the participants’ positive affection (t(28) = −2.73, p = 0.005), negative affection (t(28) = 2.06, p = 0.025), and positive cognition (t(28) = −1.78, p = 0.043) toward learning to teach CT after REC-STEMEd. The effect size, Cohen’s d, was 0.64 for positive affection, 0.75 for negative affection, and 0.75 for negative cognition indicating a moderate effect of the REC-STEMEd intervention. However, the REC-STEMEd has a small effect on the preservice teacher’s positive cognition (d = 0.45). While the preservice teachers’ positive affection significantly increased, their negative affection toward learning to teach CT decreased after the REC-STEMEd.

No statistically significant difference was found between participants’ pre-negative cognition (M = 2.85, SD = 0.49) and post-negative cognition (M = 2.71, SD = 0.73) toward teaching CT, (t(28) = 0.99, p = 0.165). However, their negative cognition decreased while their positive cognition significantly increased. Table 2 represents the descriptive statistics of each subscale of the Teacher’s Attitudes toward Learning to Teach CT survey.

Table 2 Descriptive statistics and paired samples t-tests results for preservice teachers’ attitudes
Full size table

6.2 Intentions to teach CT

Overall, the preservice teachers’ attitudes toward teaching CT, subjective norms, perceived behavioral control and intentions to teach CT are positive. The mean scores for all subsections are close to 4 (Agree). The intentions subscale has the highest mean score among all subscales (Table 3). The findings for each of the subscales are discussed below.

Table 3 Descriptive statistics for intentions to teach CT
Full size table

6.2.1 Attitude toward teaching CT

The overall mean score for the attitudes toward teaching CT was 3.50 with a standard deviation of 0.52. This shows that preservice teachers had moderately positive attitudes toward teaching CT after the intervention. The mean scores for all the items are above 3 (Neutral). They expressed the strongest agreement on statement 1 “Teaching computational thinking is important.” (M = 3.92, SD = 0.55) while statement 5 “Teaching computational thinking will bring me respect” was rated the lowest among other statements (M = 3.06, SD = 0.67). Table 4 presents descriptive statistics for each item.

Table 4 Descriptive statistics for attitudes toward teaching CT (n = 39)
Full size table

6.2.2 Subjective norms

Overall, preservice teachers perceived the course instructors’ and their teaching partners’ expectations and judgments as important in their choice to teach CT (M = 3.69, SD = 0.49). While the means for the statements related to the course instructor were between 3.69 and 4.00, the teacher partners’ statements had means ranging between 3.22 and 3.58. This shows that they gave more weight to the course instructor’s expectations and judgments. Table 5 reports the descriptives of each statement in the subjective norms subscale.

Table 5 Descriptive statistics for subjective norms (n = 39)
Full size table

6.2.3 Perceived behavioral control

The mean score for this section falls between (3) Neutral and (4) Agree (M = 3.53, SD = 0.53), that is, preservice teachers believed that they had the ability, confidence and sense of control to teach CT. One notable finding was the lowest rating of item 2, “I think it will be easy for me to teach computational thinking” (M = 2.97, SD = 0.81). On the other hand, they rated the statement that the teaching of CT is under their control with highest agreement (M = 3.75, SD = 0.60). Table 6 provides the mean and standard deviation of each statement for perceived behavioral control.

Table 6 Descriptive statistics for perceived behavioral control (n = 39)
Full size table

6.2.4 Intentions

The intention to teach CT subscale has the highest mean score (M = 3.87, SD = 0.53) compared to other subscales, which indicates that preservice teachers would like to teach CT in their future classrooms. The mean scores of all statements are very close to 4 (Agree). Descriptive statistics for each statement are presented in Table 7.

Table 7 Descriptive statistics for intentions (n = 39)
Full size table

7 Discussion

Despite the growing demand for computer science education in K-12 (Avcı & Deniz, 2022), preservice teachers have limited opportunities to learn and practice foundational concepts like CT (Broza et al., 2023; Çiftçi & Topçu, 2023; Yadav et al. 2022). This gap is especially pronounced in early childhood education, where research and training often focus on older grades, potentially leading to negative attitudes toward teaching CT (Avcı & Deniz, 2022; Çiftçi & Topçu, 2023; Weber et al., 2022). To fill this gap, this study implemented REC-STEMEd, a month-long module featuring integrated CT and science into an early childhood teacher education course. The module included unplugged activities and hands-on experiences with early childhood-friendly robots (Bluebots and Ozobots). Grounded in constructionism and the theory of planned behavior, the study examined the module’s impact on preservice teachers’ attitudes and intentions to teach CT.

Findings from the Attitudes toward Learning to Teach CT survey revealed a statistically significant improvement in participants’ positive affection and positive cognition about learning to teach CT. For instance, participants reported that learning to teach CT is exciting and indispensable for their career. Findings also showed a statistically significant decrease in participants’ negative affection about learning to teach CT, that is, their fears and concerns greatly decreased after participating in REC-STEMEd. Participants’ negative cognition, which relates to perceived difficulties and challenges, slightly decreased although the difference was not statistically significant. These findings align with several studies showing that CT interventions improve attitudes, self-efficacy, and beliefs among preservice teachers (e.g., Chang & Peterson, 2018; Jaipal-Jamani, 2018; Jaipal-Jamani & Angeli, 2017; Jeon & Kim, 2017; Kim et al., 2015). However, some studies in the literature revealed no statistically significant differences in attitudes toward programming (Cetin, 2016; Cetin & Andrews-Larson, 2016).

We argue that the promising positive impact of REC-STEMEd on early childhood preservice teachers’ attitudes is connected to two aspects of its design. First, the transition from unplugged to plugged activities aided in making abstract CT concepts more concrete. By initially using their own bodies and peers (concrete) for directional inputs (unplugged) before progressing to symbolic languages (abstract) for programming robots (plugged), participants could better conceptualize CT in a developmentally appropriate way. Existing research supports this approach. Jaipal-Jamani and Angeli (2017) found that robotics activities significantly increased preservice teachers’ interest and self-efficacy, while Alqahtani et al. (2022) reported that robotics activities led to positive intentions toward robotics-enhanced teaching. These findings highlight the importance of incorporating educational robots in preservice teacher training to facilitate CT application.

Second, integrating CT with science may have reinforced its application through real-world examples, helping participants recognize its practical value (Günbatar & Bakırcı, 2019; Weber et al., 2022). This aligns with Çiftçi and Topçu’s (2023) study, which highlighted the benefits of interdisciplinary CT instruction for early childhood preservice teachers. Similarly, Gleasman and Kim (2020) found that embedding CT in mathematics education moderately improved preservice teachers’ attitudes toward CT and significantly improved their attitudes toward programming. Addressing preservice teachers’ reluctance to teach CT, as noted by Gal-Ezer and Stephenson (2010), REC-STEMEd’s interdisciplinary approach seems to be a promising solution.

In contrast, results from some studies encourage further inquiry into the impact of robotics and CT training on preservice teachers’ interest in STEM subjects. Kim et al. (2015) found that after participating in robotics activities, interest in science and engineering increased significantly, while interest in technology showed a non-significant increase, and interest in mathematics experienced a non-significant decrease. Yadav et al. (2014) reported that implementing CT modules led to no significant differences between the experimental and control groups. Notably, neither study embedded CT within interdisciplinary scenarios. REC-STEMEd’s interdisciplinary design, which ties CT to authentic scenarios involving animal nutrition and environmental responsibility, likely helped participants connect CT to meaningful, practical situations. This approach not only reinforced the relevance of coding and robotics in everyday problem solving but may have also reduced anxiety and stress about learning to teach CT, though participants remained cautious about its perceived difficulty.

Findings from the Intentions to Teach CT survey showed that participants exhibited moderately positive attitudes toward teaching CT, with most agreeing on the value of CT instruction for K-12 students. In terms of subjective norms, participants agreed that their instructor and teaching partner(s) influenced their decision to teach CT in the future, with a slightly stronger focus on instructor influence. Specifically, they pointed out that the teacher educator’s expectations and teaching strategies affect their interest to teach CT. Regarding perceived behavioral control, participants reported a slightly positive perception of their ability, confidence, and sense of control in teaching CT. Interestingly, while participants agreed that it is under their control to teach CT, most thought that teaching CT is no easy task. Finally, participants reported positive intentions toward teaching CT as they plan to teach CT to their future students.

Prior research on CT instruction for preservice teachers has generally reported positive effects on their perceptions of CT and coding/programming as essential components of K-12 education (Avcı & Deniz, 2022; Broza et al., 2023; Vasconcelos & Kim, 2020; Yadav et al., 2022). For example, Alqahtani et al (2022) designed and implemented activities in which preservice teachers engaged with CT by solving math problems using robots. This experience led to increased intentions to incorporate robotics into their future teaching. Similarly, Tankiz and Atman Uslu (2023) found that integrating CT with other subjects significantly improved preservice teachers’ self-efficacy for teaching CT, and Ogegbo (2022) found positive attitudes and intentions to integrate CT into science classrooms among in-service teachers. However, not all studies report uniformly positive outcomes. Zha et al.’s (2020) study showed that their intervention did not significantly impact preservice teachers’ interest in CT, although qualitative findings suggested a more favorable attitude toward CT integration among preservice teachers from STEM majors compared to non-STEM majors.

A recurring theme in the literature is preservice teachers’ concerns about potential barriers and challenges in teaching CT, such as insufficient training, lack of resources, and a lack of confidence in their own abilities (Gudmundsdottir et al., 2020; Lachner et. al, 2021; Ling et al., 2017; Yadav et al., 2022). For example, Broza et al. (2023) noticed a disparity between preservice teachers’ improved CT skills and a reduction in their self-efficacy in teaching CT. Similar findings on low self-efficacy have been reported in studies of in-service teachers (Kaya et al., 2019; Mason & Rich, 2019). Given these challenges, we argue that beyond developing CT skills and content knowledge, teacher education programs should intentionally support preservice teachers’ pedagogical knowledge development, which may enhance their confidence and preparedness to teach CT (Yadav et al., 2022).

Participants in the present study mentioned the impact of the teacher educator and peers on intentions to teach CT. Perhaps this is because REC-STEMEd closely involved both stakeholders in the module. First, the teacher educator constantly checked in and provided formative feedback to participants as they worked on CT tasks. In addition, working in pairs or small groups may have contributed to a sense of joint effort in which group members keep each other accountable during the CT tasks. These positive experiences may have contributed to participants’ higher intentions to teach CT. Lin and William’s (2016) study found a strong relationship between preservice teachers’ positive attitudes and their intentions toward STEM teaching. Grounded on the premise that technology-enhanced learning experiences are predictors of preservice teachers’ use of those technologies in their future teaching (Alqahtani et al., 2022), we argue that experiences with developmentally appropriate CT activities as part of science lessons helped participants envision ways to do it in their future classrooms, which resulted in stronger intentions to teach CT.

8 Implications for STEM teacher education

Several key implications for infusing CT into teacher education can be drawn from this study’s findings. First, when introducing the concept of CT, instructors should address preservice teachers’ preconceptions and misconceptions by providing targeted support that fosters positive attitudes and dispositions toward CT. Designing activities that directly confront these misconceptions can help preservice teachers develop a clearer and more confident understanding of CT. Second, connecting CT to preservice teachers’ prior schemas and experiences can make it more tangible and accessible. Helping them see how CT applies to their daily lives and instructional practices can shape more achievable and realistic perceptions of its role in education. Integrating CT with at least one other discipline can further reinforce its cross-curricular relevance and encourage preservice teachers to implement across various subjects. Additionally, embedding CT within an existing curriculum can help preservice teachers see its application in developing critical thinking, creativity, and adaptability – skills that are critical for both educators and students. Rather than viewing CT as a separate subject, preservice teacher education programs should see ways to integrate it within existing teaching methods courses, helping educators in tailoring CT to different grade levels. Third, gradually scaffolding CT instruction—starting with foundational concepts and progressively increasing the complexity of CT tasks—can help preservice teachers become comfortable with ill-structured problem-solving and more abstract applications. Finally, offering opportunities to develop both content knowledge, pedagogical knowledge, and integrated instructional approaches can boost preservice teacher’s self-efficacy and intentions toward teaching CT.

9 Study limitations

It is critical to acknowledge study limitations to foster scientific integrity, transparency, and contextualization of findings. A few methodological limitations affected this study. First, the REC-STEMEd module was implemented for a limited time due to scheduling constraints in the teacher education course. The condensed timeframe may have restricted participants’ opportunities for deeper engagement with the module’s content and hands-on activities. Future research should explore the benefits of extending the implementation period, potentially integrating the module across multiple teacher education courses or spanning an entire academic term. This could allow for a more comprehensive assessment of the module’s impact. Second, the sample size was limited because of smaller-than-usual early childhood education cohorts. Additionally, participant attrition further reduced the final sample, as some participants did not complete either the pre- or post-attitudes survey. The smaller sample size may have impacted the statistical power of analyses and the generalizability of findings. Future research should aim to include larger and more diverse cohorts, along with strategies to reduce participant attrition. Third, the data collection and analysis relied exclusively on quantitative data about participants’ attitudes and intentions. While surveys are effective data sources to assess those constructs, qualitative data sources like interviews or open-ended questions would have provided a more comprehensive understanding about participants’ experiences. Follow-up studies should consider mixed methods approaches. Fourth, it was not possible to measure participants’ intentions toward teaching CT before the REC-STEMEd module, as most were introduced to CT for the first time during this study. Without prior exposure, they may not have had a meaningful baseline understanding or relevant intentions to report. Consequently, pre-intervention data could not be collected, limiting the ability to compare changes before and after the intervention. Future research can include preliminary instruction of CT before formal data collection to gauge changes in participants’ intentions. Fifth, the reliability analysis of the surveys was not conducted after modifications were made due to the small sample size. Without this, the internal consistency of the revised instruments remains uncertain. Future research should aim to collect data from a larger sample to enable a reliability analysis. Sixth, the study did not include a control group, which prevented comparisons regarding the impact of robotics-enhanced CT instruction vs. unplugged CT instruction. Future research should include a control group to facilitate examining the effects of different instructional approaches. Finally, the study did not include a delayed follow-up to measure long-term effects of the intervention on participants’ instructional practices after graduation. Without longitudinal data, it remains unclear if the intervention results in sustained improvements in instructional practices. Future research should adopt a longitudinal approach by following preservice teachers beyond their teacher education program to examine the lasting effects of the intervention.