Productive Failure to Promote Deeper Self-Directed Learning in Coding and Robotics Education

Abstract

In a world characterized by unpredictable change, students in Computer Science education must be deeper self-directed learners who can take ownership of their learning and transfer knowledge and skills to new contexts. This article reports on how productive failure was incorporated into an introductory coding and robotics course to enhance deeper self-directed learning. The population was 42 fourth-year pre-service teachers from two different campuses of a South African University. All students were invited to participate in the research, and 37 students consented to participate. A basic interpretative qualitative research design was followed. Guided self-reflection documents were used as data-gathering methods, and data were analyzed by applying thematic data analysis. The research concluded that productive failure, incorporated with cooperative pair programming and self-reflection, in introductory coding and robotics education, shows promising results for developing deeper self-directed learning. Furthermore, it is suggested that solvable problems should initially be introduced, because the new coding and robotics environment already contributes to the complexity of tasks. It was secondly concluded that participants’ self-reflections deepened after engaging with unsolvable problems. Follow-up research is required to determine if the transfer of knowledge and skills to new contexts occurred.

1. Introduction

Computer Science education pre-service teachers find themselves amid a world disrupted by artificial intelligence, digital innovations, and educational dilemmas. On the one hand, they need to keep up with rapid changes in technology, and on the other hand, they need to master teaching and learning strategies that will address the needs of their future learners and prepare them for a world characterized by disruptive technologies (Li et al., 2022). The focus of Computer Science education is therefore twofold. Firstly, it is not only to acquire new knowledge but also to be deeper self-directed learners that can adapt to change, navigate failures, and reflect on problem-solving attempts (Breed & Bailey, 2018; Prather et al., 2020). The knowledge and skills that are gained must then be transferred to solve problems in new contexts (Cheah, 2020). Secondly, as education students, they need to transfer their pedagogical knowledge and skills to facilitate relevant subjects in their classes one day. Against this background, promoting deeper self-directed learning (DSDL), the process of being self-directed learners and transferring knowledge and skills to solve problems in new contexts, becomes imperative (Van Zyl & Mentz, 2020).

Computer Applications Technology (CAT) is an elective school subject for learners in Grades 10 to 12 in South Africa. CAT entails the study of computer hardware and application software to solve real-world problems (Department of Basic Education [DBE], 2011). In line with the global movement to infuse coding and robotics (C&R) in education (Fegely et al., 2024), C&R is planned to be implemented as a compulsory subject in all public schools in South Africa (Department of Basic Education [DBE], 2021). Although C&R is not currently the focus of CAT education, pre-service CAT teachers will most probably be part of the cohort of educators who are tasked to facilitate C&R in their classes one day. Subsequently, an introductory course in C&R has been incorporated into the training of pre-service CAT teachers.

Learning coding, or computer programming, presents a considerable challenge, and even entry-level programming courses have high failure rates (Cheah, 2020). C&R involves a plethora of skills necessary for designing and implementing programming solutions (Yildiz & Seferoğlu, 2021). Students need to persist in solving complex problems, test practical solutions, reflect on how robots execute programming code, and optimize robot performance (Ioannou & Makridou, 2018). Consequently, experiencing failure and managing one’s thought processes are integral when learning how to program (Prather et al., 2020).

Traditional teaching and learning strategies often fall short of adequately addressing the multifaceted challenges inherent in learning to program (Cheah, 2020). This shortcoming is particularly noted in pre-service teacher education, where a gap persists in understanding how pre-service teachers acquire coding skills (Fegely et al., 2024). Furthermore, Nagy et al. (2025) elaborate that engaging in hands-on, active learning fosters greater retention and comprehension compared to traditional lectures. There is subsequently a pressing need for innovative pedagogical approaches that not only accommodate the complexities of coding but also actively embrace failure as a learning opportunity and promote reflective thinking.

Productive failure is a teaching and learning strategy that involves students in collaborative problem-solving within complex contexts, intentionally allowing for potential failures and minimizing support structures (Kapur, 2008; Steenhof et al., 2020). This strategy contradicts Cheah’s (2020) recommendation of consistently offering support during the initial stages of learning to program. Although solving complex problems is part of life, the disconnect between failure and success can lead students to give up on challenges or multifaceted problems, resulting in negative learning experiences. Beghetto (2018) therefore suggests that failure must be complemented by metacognitive strategies to enable reflection on setbacks and foster deeper understanding. Moreover, Denner et al. (2021) argue that collaborating in pair programming teams can foster resilience in the face of setbacks (Denner et al., 2021).

Deliberate strategies need to be incorporated to enhance students’ self-reflection and to promote DSDL (Breed & Bailey, 2018; Pedrosa et al., 2021; Van Zyl & Mentz, 2020). A gap in the literature exists regarding the incorporation of failure to promote DSDL and its subsequent effect on self-reflection. The primary research objective for this study, therefore, was to explore how productive failure could be applied in an introductory C&R course to promote DSDL among CAT pre-service teachers. The secondary research objective was to determine if incorporating productive failure would deepen students’ self-reflections. The research questions were the following:

• How should a learning environment be designed when incorporating productive failure in C&R education to promote DSDL?
• What do participants’ self-reflections reveal when incorporating productive failure in C&R education?

2. Theoretical and Conceptual Framework

This section addresses the following aspects: DSDL, productive failure, C&R education, and the significance of cooperative pair programming.

2.1. Deeper Self-Directed Learning

DSDL is an integration of the concepts of deeper learning (Hilton & Pellegrino, 2012) and self-directed learning (Knowles, 1975). While self-directed learning is the process in which students take ownership to fulfill their learning needs, deeper learning focuses on developing transferable knowledge and skills. Deeper self-directed learning is thus defined as the process where students intentionally take ownership of the learning process that results in transferable cognitive, intrapersonal and interpersonal competencies (Van Zyl & Mentz, 2020).

Within the intrapersonal domain that drives the DSDL process lie the primary processes of self-regulation, metacognition, and motivation and the secondary processes of choice, control, competence, and confidence. When analyzing, reasoning, planning, and determining goals, cognitive competencies such as problem solving, critical thinking, and creativity are developed. The realization of learning goals is enhanced within social learning environments where peers function as a cooperative learning team, communicating with one another, sharing knowledge and responsibilities, and building knowledge and skills while assisting each other. Ultimately, the aim of DSDL is for students to take charge and ownership of their learning, to achieve far transfer—the transfer of competencies to new contexts (Van Zyl & Mentz, 2020).

The theoretical foundations of DSDL are rooted in social constructivist theory (Vygotsky, 1978) and cognitive load theory (Sweller et al., 2019). According to social constructivist theory, learning is enhanced by social interaction, where knowledge construction is facilitated through reflective thinking and active engagement with peers (Vygotsky, 1978). Cognitive load theory is particularly relevant in the context of learning complex tasks and requires that learning environments be structured to effectively balance students’ cognitive load (Sweller et al., 2019). Students who are collaborating effectively share the cognitive load and subsequently reduce the cognitive load on the individual student (Sweller et al., 2019; Skuballa et al., 2019).

However, Johnson and Johnson (2019) argue that merely requiring students to work collaboratively will not result in effective cooperation. In cooperative learning, a structured and supportive environment is provided where small groups of students are positively interdependent on one another and work together to solve a problem and reach a common goal (Johnson & Johnson, 2019). Cooperative learning is distinguished from traditional group work by incorporating five principles, namely positive interdependence, individual accountability, face-to-face promotive interaction, social and group skills, and group processing, as highlighted by Johnson and Johnson (2019). Groups are subsequently structured cooperatively, with a focus on mastering the group task rather than competition between group members or individualistic efforts (Johnson & Johnson, 2019).

Research has indicated several positive outcomes of cooperative learning, such as higher achievement and increased productivity, increased frequency in generating new ideas and solutions, knowledge transfer, high-level reasoning and critical thinking (Johnson & Johnson, 2019). However, the value of cooperative learning becomes clear with challenging tasks. As the need for problem solving increases, elevated levels of reasoning and critical thinking are observed, along with a heightened demand for the application of previously acquired knowledge (Johnson & Johnson, 2019).

Managing failure and demonstrating resilience in the face of setbacks are characteristics of a self-directed learner (Guglielmino, 2013). These qualities are also essential for cooperative groups to address and reflect on during group processing (Johnson & Johnson, 2019). Deliberate planning for failure within cooperative learning environments and the use of metacognitive strategies to enhance reflection, therefore becomes essential, with the aim of developing DSDL.

2.2. Productive Failure

Planning for deliberate failure in teaching and learning has the potential to profoundly influence conceptual understanding and promote deeper learning (Kapur, 2015; Steenhof et al., 2020). When aiming to develop transferable knowledge and skills, such as problem solving, a deep understanding of concepts is required (Kapur, 2016). Research indicates that incorporating failure and allowing students to first engage in problem solving, can develop transferable problem-solving skills (Kapur, 2016).

The essence of productive failure is twofold. Firstly, it is to limit instructional support and structured teaching and learning methods during initial problem-solving processes (Kapur & Bielaczyc, 2012), and secondly, it is to design tasks that will lead to failure, despite students’ efforts to find solutions (Steenhof et al., 2020). Productive failure subsequently differs from traditional teaching and learning strategies which usually focus on obtaining success and providing scaffolding and guidance when students engage in complex tasks (Cheah, 2020).

In productive failure, students attempt to solve complex problems independently, without guidance from the educator in the form of direct instruction or scaffolding. Kapur and Bielaczyc (2012) describe two distinct phases of productive failure. Firstly, a generation and exploration phase (in which students attempt to collaboratively solve the complex problem), and secondly, a consolidation and knowledge assembly phase. During the second phase, the educator leads the class discussion and compares the solutions presented by learners with canonical solutions (Kapur & Bielaczyc, 2012).

In research done by Kapur and Bielaczyc (2012), 7th-grade Mathematics students first had to solve complex problems collaboratively without any scaffolding or teacher support. Students were grouped in triads that the teacher envisaged would work well together (Kapur & Bielaczyc, 2012). Students received complex problems to solve, followed by what-if scenarios. The role of the teacher in the first phase was to manage the classroom and to provide affective support. Students were informed that the goal of the assignment was to develop and explore multiple solutions to the problem, rather than finding a single correct solution. During the second phase, the teacher led a class discussion where groups presented their solutions. The teacher then consolidated and discussed the canonical solutions. Students who engaged in productive failure then later outperformed other students who had to solve the same complex problems in a traditional teacher-centered and supported environment (Kapur & Bielaczyc, 2012).

According to Sinha and Kapur (2019), specific criteria must be applied for productive failure to be effective and successful. The main aim when applying productive failure is not to successfully solve the problem but to compile as many possible solutions to solve the problem at hand. According to Sweller et al. (2019, p. 265), such a “goal-free effect” reduces the cognitive load of complex tasks. While attempting various solutions, multiple solutions and methods can thus be explored (Kapur, 2016), which encourages the construction of multiple schemas to promote learning and transfer (Nokes-Malach & Mestre, 2013). In such learning conditions, planning for failure results in a “beautiful risk” that allows students to experience “unexpected twists, turns and setbacks” to “fail forward” and to reflect on solutions (Beghetto, 2018, p. 20).

A balance must be found between keeping students engaged and interested in the problem-solving process, without being frustrated to the point of giving up (Kapur & Bielaczyc, 2012). Metacognitive and reflective thinking should be incorporated to assist students in their problem-solving attempts and decision-making (Sinha & Kapur, 2019) to capitalize on failure as a learning endeavor (Beghetto, 2018). Pedrosa et al. (2021) accordingly suggest that guided reflection questions be applied to promote self-reflection, trigger self-awareness, and elicit learning experiences that can inform subsequent problem-solving attempts.

The application of collaborative, student-centered learning in the generation and exploration phase provides students with the opportunity to take responsibility for their learning in a safe social learning space (Hod et al., 2018; Kapur & Bielaczyc, 2012). A collaborative problem-solving process enables students to work on the problem while interacting with one another and negotiating potential solutions (Buseyne et al., 2023). The collective processing capacity of groups that collaborate effectively subsequently becomes a scaffold for complex problem solving (Sweller et al., 2019).

Productive failure can be described as a failure paradox, where achievement is the other side of failure (Wooditch, 2019). Subsequently, requiring students to solve problems that seem to be “beyond their abilities” and which will result in failed attempts, will foster future learning and transfer (Steenhof et al., 2019, p. 739). Subjects such as C&R, which are typically characterized by challenges and failure (Cheah, 2020), can therefore provide the ideal context for applying productive failure.

2.3. Coding and Robotics Education

This section outlines theoretical lenses and thinking processes involved in C&R as well as cooperative pair programming.

2.3.1. Theoretical Lenses for Active Learning in Coding and Robotics

Social constructivism and constructionism are both learning theories that emphasize active engagement in challenging contexts. While both approaches share similarities, they have different foci. Constructed learning is embodied in “objects-to-think-with” (Papert, 1980, p. 11) and incorporates the affective and social domains of learning (Lodi & Martini, 2021). For example, objects associated with microworld environments (e.g., micro:bit, Scratch and Turtle Blocks) support conceptual and computational thinking (Dhakulkar & Olivier, 2021; Havenga & van Zyl, 2023; Wing, 2006).

2.3.2. Thinking Processes in Coding and Robotics

Computational thinking is considered a foundational competency in computer programming and in C&R (Lodi & Martini, 2021). Computational thinking necessitates mental processes such as stepwise or algorithmic thinking, pattern recognition, abstraction, and decomposition of problems into more manageable sections to be solved (Papert, 1980; Wing, 2006). Moreover, Lodi and Martini (2021) propose that communication and collaboration, reflection and meta-reflection, tolerance for ambiguity, and persistence when dealing with complex problems be incorporated in frameworks for developing computational thinking. The use of abstract or physical objects prompts an individual to “reflect on one’s actions and thinking” (Papert, 1980, p. 28). Accordingly, several scholars highlight problem solving and metacognitive thinking as interrelated when dealing with C&R and collaborative computational thinking (Socratous & Ioannou, 2022; Stewart et al., 2021). Metacognitive and reflective thinking can thus assist students in navigating their programming endeavors effectively, such as employing failure as a learning strategy to persist.

Self-reflection is acknowledged as a critical competency in problem solving (Falon et al., 2021). During self-reflection, students analyze their thoughts, behaviors, emotions and experiences to gain insight while solving problems (Negi et al., 2022). By reflecting on their previous experiences and behaviors, students gain insights into their problem-solving approaches and recognize strategies that work best for them (Falon et al., 2021). Incorporating self-reflection thus allows for in-depth analysis of new problems and facilitates the discovery of recurring patterns and connections with previous experiences (Breed & Bailey, 2018).

2.3.3. Cooperative Pair Programming in Coding and Robotics

As stated by Beghetto (2018, p. 19), creating a learning environment where peers seek help from each other is “taking a good risk” because the long-term benefits outweigh the risk of appearing incompetent in front of others. Pair programming involves two computer programmers collaborating closely to solve problems together, typically by working side by side on the same computer. Originally described by Williams and Kessler (2003), distinct roles are assigned to each group member: the “driver” is responsible for operating the computer and executing commands or designs, while the “navigator” serves as a strategic thinker, asking questions and ensuring the driver stays on track. The navigator also checks for errors and engages in brainstorming activities with the driver. Periodically, the driver and navigator swap roles during the problem-solving process.

In education, pair programming has been identified as an effective strategy for teaching and learning programming skills (Denner et al., 2021). This can be attributed to the “rubber-plant effect” (Sturdy, 2005, p. 213), hence conflicts in thoughts are often resolved through verbal expression. The ongoing dialogue between the driver and navigator continually stimulates each other’s thinking and aids in problem solving. Mentz et al. (2008) suggest that embedding the five elements of cooperative learning enhances the effectiveness of pair programming, leading to improved group performance and critical thinking skills (Bailey & Mentz, 2017).

Henceforth, in this article, the term “cooperative pair programming” will be used when referring to pair programming that incorporates the principles of cooperative learning.

3. Materials and Methods

3.1. Methodology and Participants

This study employed an interpretivist paradigm, and a basic interpretative qualitative research design (Merriam & Tisdell, 2015) was followed. The population involved 42 fourth-year CAT education students from two campuses of a South African university. As part of their teacher training, students were required to complete a two-week C&R course. Due to the small population, all students were invited to participate in the research, and no random sampling was conducted.

At the end of each week, all students were required to submit a self-reflection document electronically on the learning management system. A template containing guided self-reflection questions (see Appendix A) was provided. Students were assured that they would receive grades for only submitting the documents, not for the content, and were encouraged to provide honest reflections. The document included guided questions such as the following: “What did I learn today?” and “What challenges did I face?” (Pedrosa et al., 2021). No further guidance on self-reflection was given to participants.

Although the submission of these self-reflection documents contributed to the course assessment, self-reflection is widely regarded as a critical element of the learning process (Breed & Bailey, 2018; Pedrosa et al., 2021) and was therefore integrated into the productive failure intervention (see Table 1). Students completed a self-reflection following the solvable problems in week 1 (pre-PF), and once more after attempting the unsolvable problems in week 2 (post-PF) (see Appendix B.1 and Appendix B.2). Only the self-reflections of willing participants who gave consent to participate in the research were used for data-gathering. Self-reflections were anonymized, with participants’ names removed prior to data analysis. Consequently, the pre- and post-PF self-reflections of individual participants were not compared, and the absence of either reflection did not affect the analysis.

Open coding was employed to remain open to emerging theoretical insights (Saldaña, 2013). The pre- and post-PF self-reflections were coded sequentially to capture potential shifts in participants’ experiences. Codes were collated into categories based on recurring patterns, from which overarching themes were generated through an inductive data-driven process (Braun & Clarke, 2021). To enhance credibility, two researchers coded the data independently before comparing codes and categories. One researcher conducted the analysis using Atlas.ti version 25.0.1.32924 qualitative data analysis software, while the other manually coded the data in Excel spreadsheets. Following comparison and discussion to obtain inter-coder reliability, the codes and categories were refined in Atlas.ti, and the final themes were identified.

Ethical approval for the research was obtained from the relevant ethical committee of the university, and gatekeeper permission was also obtained. Two of the researchers held dual roles as lecturers of the course. Delayed consent was applied to mitigate power relations and ensure that participants’ self-reflections were not influenced by their awareness of participating in research. At the end of the course, an independent person visited the class to explain the purpose of the research and distributed informed consent forms to all students. Students indicated on the consent form whether or not they consented to participate. This procedure was implemented to protect the anonymity of consenting students and to avoid exposing them to potential pressure or bias. A total of 37 students consented to participate in the research.

3.2. Adapted Productive Failure Intervention

Two weeks, with two 90 min periods per week, were allocated for the C&R course. The Microsoft MakeCode environment and physical micro:bits were utilized. An adapted productive failure intervention (see Table 1) was implemented for all students enrolled in the course. Students had a very limited prior knowledge of C&R and the micro:bit environment. To prepare for class, participants received a link to the official Microsoft MakeCode for micro:bit website, to acquaint themselves with the C&R environment.

Table 1. C&R adapted productive failure intervention compared with Kapur and Bielaczyc (2012).

Week	Phase	C&R Adapted Productive Failure Intervention	Productive Failure (Kapur & Bielaczyc, 2012)
Week 1	Group composition	Cooperative pair programming groups, assigned randomly by the lecturer	Triads that the teacher envisages would cooperate reasonably together
	Generation and exploration	Solvable problems, because C&R is challenging for novices	Collaborate to solve a complex problem, followed by a what-if scenario
	Consolidation and knowledge assembly	Self-assessment, peer-assessment and individual self-reflection	Teacher consolidates and discusses solutions
Week 2	Generation and exploration	Unsolvable problems	Complex problem, followed by a what-if scenario
	Consolidation and knowledge assembly	Self-assessment, peer-assessment and individual self-reflection	Whole-group discussion led by the teacher. Groups share their representations and solution methods with the class. The teacher discusses solutions to well-structured problems.

Table 1. C&R adapted productive failure intervention compared with Kapur and Bielaczyc (2012).

Week	Phase	C&R Adapted Productive Failure Intervention	Productive Failure (Kapur & Bielaczyc, 2012)
Week 1	Group composition	Cooperative pair programming groups, assigned randomly by the lecturer	Triads that the teacher envisages would cooperate reasonably together
	Generation and exploration	Solvable problems, because C&R is challenging for novices	Collaborate to solve a complex problem, followed by a what-if scenario
	Consolidation and knowledge assembly	Self-assessment, peer-assessment and individual self-reflection	Teacher consolidates and discusses solutions
Week 2	Generation and exploration	Unsolvable problems	Complex problem, followed by a what-if scenario
	Consolidation and knowledge assembly	Self-assessment, peer-assessment and individual self-reflection	Whole-group discussion led by the teacher. Groups share their representations and solution methods with the class. The teacher discusses solutions to well-structured problems.

In class, the lecturers randomly assigned students to cooperative pair programming groups through a fun activity. Students received minimal guidance from the lecturers and no direct instruction, or explanations were provided. When students sought advice, lecturers responded with guiding questions to facilitate independent problem-solving.

Table 1 presents a comparison between the adapted productive failure intervention and the productive failure design described by Kapur and Bielaczyc (2012).

The adapted productive failure intervention incorporated the conditions proposed by Sinha and Kapur (2019) but differed in certain aspects from the design outlined by Kapur and Bielaczyc (2012) (see Table 1). Firstly, the method of group assignment differed. Lecturers randomly assigned students to pairs, rather than being placed in triads that the lecturers believed would work well together. Secondly, because C&R is traditionally challenging for novices (Cheah, 2020; Ioannou & Makridou, 2018), it was argued that even exploring the solvable C&R problems would present a significant challenge for participants. Therefore, as suggested by Kapur and Bielaczyc (2012), to keep students engaged and interested in the problem-solving process, without being frustrated to the point of giving up, the generation and exploration phase in week 1 involved solvable problems. Students received a list of problems to choose from and were also allowed to create their own problem scenarios (see Appendix B.1). Additionally, students were encouraged to explore online resources.

Thirdly, in the consolidation and knowledge assembly phase, participants could choose their best solutions for peer- and self-assessment. Fourthly, no direct instruction was provided by lecturers, nor was there a whole-group discussion led by them. Instead, pairs demonstrated and explained their solution to other groups and were assessed by using a rubric handed out by the lecturers. Ample time was allowed for assessment, and groups rotated to assess as many groups’ solutions as possible. Lastly, participants individually completed a self-reflection document, consisting of open-ended questions to guide their reflections (see Appendix A).

In the second week, students received a list of unsolvable problems compiled by the lecturers (see Appendix B.2). Again, to avoid discouraging students, the problems were intentionally presented as enjoyable tasks, although they could not be solved. The unsolvable problems were the only change from the intervention design of week 1. Students stayed in the same cooperative pair programming groups as in week 1, and the same procedure was followed in the consolidation and knowledge assembly phase (see Table 1).

4. Findings

In this section, the self-reflections conducted after the first week will be referred to as pre-productive failure (pre-PF) reflections. In addition, self-reflections that were conducted after implementing the unsolvable problem during the second week will be referred to as post-productive failure (post-PF) reflections. Data were anonymized, and individual participant reflections were not juxtaposed. The themes identified through data analysis will be presented, with efforts made to categorize them into cognitive, intrapersonal, and interpersonal aspects of DSDL. The discussion will commence with an examination of the pre-PF findings, followed by the post-PF findings.

4.1. Findings Based on Participants’ Pre-PF Reflections

Figure 1 depicts the themes and categories identified through the analysis of the pre-PF reflections. Participants identified their learning needs and searched for online resources to enhance their understanding and overcome challenges. Group support served as an important scaffold to enhance understanding of concepts and to overcome challenges. Despite having no prior knowledge of C&R, participants were intrinsically motivated, engaged, focused, and enjoyed the new challenges.

The next section provides a discussion of the themes that emerged from the data presented in Figure 1.

4.1.1. Reflection on Initial Challenges and Learning Needs

Initially, participants faced challenges in determining what was required to understand the micro:bit programming environment: “… figuring out what was expected of us” (P26); “Explored the basic functionality of micro bits” (P30); and “Understanding some of the code … and sourcing resources” (P22). “Today, I learnt how to make use of the micro:bit grid (5x5) to plot points such as to create a smiley face or number” (P27).

Participants reflected on their “aha-moments”: “I finally understood and was able to identify how the various points were plotted. The ‘aha’ moment was getting clarity, or a better understanding on the instructions” (P32). Long-term goals were mentioned: “to understand coding so that I can teach it to the learners in my class one day” (P2); and “aiming to become a master to an extent that I have alternative ways of teaching coding to learners without running out of approaches” (P37).

Participants highlighted the importance of finding relevant resources: “watching multiple YouTube videos” (P6); and “search[ing] online for any guidance to code in MakeCode to overcome this challenge” (P8). Participants also indicated specific skills and strategies they used when introduced to micro:bit. For instance, some used trial and error, while others employed decomposition, associated with computational thinking: “if we failed then we tried again, using the knowledge gained from the first attempt” (P24); and played around with the loops … till I eventually found the correct block” (P4). “[it] challenged me to think critically, break down complex issues, and find efficient solutions” (P9).

Participants identified several learning needs, which are prominent characteristics of self-directed learners: “there is so much more information I need to learn about” (P6); “explore additional resource … be able to add extension blocks, make variables and create images and add games” (P23); and “learn how to insert and play an external sound on the micro:bit.” (P24).

4.1.2. Group Cooperation Enhanced Knowledge and Skills

Effective group cooperation was integral to problem solving and also shaped the way in which participants engaged in C&R activities. Participants collaborated with peers, searched for online resources, and supported one another’s learning: “receiving the correct advice … enabled me to think critically and implement [C&R] effectively” (P3); and “We also consulted the lecturer and fellow classmates” (P5).

Group support encouraged persistence, which resulted in successful problem solving and completion of tasks: “By working efficiently my team member and I were able to successfully complete the task” (P3); “helped me to stay intrigued and engaged in the activity at all times” (P37); and “taught me new concepts and ways of thinking… stay focused, and it worked perfectly” (P32). One participant highlighted the value of support and emphasized awareness of his/her own limitations: “I get very distracted, asking classmates for help and helping them at times” (P33).

Participants highlighted the value of having a peer to review code for errors and ensure timely task completion. “Before running the code, I asked my partner and other classmates to take a look at the code” (P6). “I worked with my partner which made things look easier and finish our task on time” (P9).

Working in groups further contributed to enjoyment: “Seeing our peers cheering for their predictions was great to witness” (P25). “I was very focused, and we learned so much. I loved the group work” (P7).

4.1.3. Fostering Intrinsic Motivation and Intrapersonal Competencies

Although participants found the new micro:bit environment and initial activities challenging, their enjoyment, excitement and curiosity fostered intrinsic motivation to remain focused and engaged. “The excitement I had … was amazing. I am interested in this coding journey; I enjoy it a lot” (P36). “I was extremely focused on solving the problem. Difficult, but I am really enjoying it.” (P2).

Enjoyment and intrinsic motivation furthermore encouraged persistence to accomplish the task at hand. “I was really determined to get it right… determined to solve the problem.” (P4). Participants explored new functionality and, in the process, gained more confidence in their coding abilities. “We added an extra element to our project that was not even required” (P35); “experimented with different LED plotting patterns to explore variations” (P5); and “explored other components of MakeCode building blocks” (P8). “add an additional few blocks of code to allow the user to reset and clear the values of the micro:bit” (P33).

Participants demonstrated an awareness of their learning strategies and proposed ways to enhance their learning. “To be more open to adjusting your learning approach. If something isn’t working, don’t hesitate to try a different method or resource.” (P9). “Familiarize myself better with how micro:bit works beforehand. Practice more before coming to class” (P30).

Participants envisioned teaching C&R in the future, even though it was not the primary focus of their current teacher training. “Coding is fun and I want to create a lesson one day at a school that HAS robotics” (P7). “One of my long-term goals regarding coding is to teach coding not only to learners but to share my insights through blogs and podcasts or even a YouTube channel that will cater to people of all ages and in the long run even run workshops on coding” (P6).

4.2. Findings Based on Participants’ Post-PF Reflections

Figure 2 depicts the themes and categories identified through the analysis of the post-PF reflections. The following highlights the post-PF reflections: “it is my mission to instill in my students the valuable skills of computation and programming” (P19).

The next section provides a discussion of the themes that emerged from the data presented in Figure 2.

4.2.1. Persistence in Fostering Problem-Solving

Participants persisted in their attempts to solve problems and developed skills such as creativity, resilience, and finding resources. As mentioned by P30, “we persisted and experimented until finally achieving success” (P30). Creativity emerged along with curiosity and enjoyment, for example: “combining musical instruments and making them play at the press of a button, when shaking it” (P23). “no matter how hard a problem can become to solve, one must just keep trying…and keep thinking outside the box” (P27). Participants were focused on finding solutions and extensively searched for various online resources to solve their problems: “We used all the video solutions we had and used the solutions provided to make our own” (P25); and “use of Google, YouTube, the micro:bit program as well and the booklet provided in order to figure out what is required and how to do it” (P27).

Understanding of concepts moved to a deeper level and improved problem-solving skills emerged. “with all odds I was able to successfully complete my project with a deep understanding of how the micro:bit work” (P23); “I’m anxious to research these concepts further and put the techniques into action in order to deepen my comprehension of the subject” (P19). Subsequently, participants followed a structured approach to solving problems and focused less on trial and error. “We tried a new approach by building a song code using an accelerometer in order to control the strength and control where the ball must fall in the maze” (P27). “every time we did one coding, we tested the bot and see if it reacted. And if it reacted, we went on with the coding” (P19).

Decomposition and algorithmic thinking were applied as computational thinking elements to solve problems. “I broke the problem into smaller problems in order to solve the main problem” (P12). “First, the surface area of a matchstick needed to be worked out and then the surface area of the Earth’s surface. The Earth’s surface needed to be divided by the surface area of the matchstick” (P16).

4.2.2. Effective Cooperation Enhanced Learning

Participants acknowledged the value of working in cooperative pair programming groups. They mentioned that group support helped them maintain focus and engagement. “I was highly focused…having a working peer partner to do some learning together influences one to be focused throughout the activities” (P29). “I loved the group work since I was so focused and we learnt so much (P19). “…we realized that when we work together we work better” (P13). “The interactive components of the practice, group activities and hands-on activities, helped maintain my attention and curiosity” (P19).

Group support was often mentioned as an aspect that motivated participants to overcome challenges. “I struggled with customizing the notes and duration but I managed to get through working with my partner” (P31). “Collaborating with my peers made navigating the activity easier” (P25). One participant mentioned that they tried various options to solve a problem, but eventually, assistance from peers was the most useful resource. “At first, we changed the problem for another. Later, we went to redo [it]… asked our peers to assist us, and that proved fruitful” (P25). “Collaborated with my partner to brainstorm and implement solutions. Persisted…until finally achieving success.” (P30). Group support was extended to include other students who were willing to help each other. “Yes, the CAT class as a whole feels like a built-in family whereby everyone is willing to help one another” (P12). “I am proud that I did not know anything about coding and I learnt how to do it and how to do it correctly” (P13).

There were isolated instances where participants mentioned that they moved from group learning to individual learning to find their own resources. “His interactions contributed a lot to my learning however, due to the complexity of the problem we chose to solve, I did look more to my resources for help” (P27). “I need quiet time to go over all the blocks available and do some more tutorials on my own” (P15).

4.2.3. Deeper Self-Directed Learning Characteristics

Participants’ reports reflected DSDL characteristics across the interpersonal, intrapersonal and cognitive domains. Section 4.2.2 above highlighted competencies in the interpersonal domain. In addition, many reflections referred to the intrapersonal domain, suggesting that participants considered this dimension particularly significant in their learning. Within this domain, categories such as goal setting, intrinsic motivation, persistence, and self-efficacy were identified (see Figure 2). In the cognitive domain, creativity, computational thinking, enhanced problem-solving skills and deeper understanding emerged as prominent categories (see Figure 2 and Section 4.2.1).

Participants expressed pride in their learning journey with C&R. Overall, participants were not discouraged by the challenging, unsolvable problems but gained confidence in their C&R abilities and were determined to find solutions. They set personal goals to enhance their C&R skills and advance their ability to solve more challenging problems: “My goal…solve a more complex problem that I did the previous week” (P28); “The challenge motivated us to work diligently” (P30); “My personal growth and self-confidence as I was able to start a project from the beginning (set a goal), went through the challenges and overcome them (work towards it)” (P23); and “I learned to view every obstacle as a chance to grow and learn” (P19).

Small achievements led to intrinsic motivation, encouraged persistence, and contributed to improving self-efficacy: “…when I did not use the correct measuring units, I still got a solution and an answer” (P15); and “…explore beyond what was required because I looked at the other scenarios and how to better the task we choose” (P16). “…problems gave us an opportunity to explore more” (P33). “Overcoming the challenges faced demonstrated my problem-solving skills” (P23).

Participants reflected on strategies to improve engagement and learning: “I’m proud to have successfully code the micro bit with little problems… I had prepared for today’s lesson” (P29); “Balancing between trying different solutions and seeking external help” (P30); and “…focus on doing my research about the code I needed to do to reduce my mistake time” (P13).

Participants envisaged themselves as lifelong learners and future C&R educators: “Continue developing problem-solving skills and coding proficiency, applying them to more advanced projects” (P30); “I can now teach others about the micro:bit” (P23); “…using MakeCode to code instructions for Microbits” (P16); and “Thus squashing any doubt or even self-doubt…able to code complex games and solutions…use the knowledge I have to do it” (P27). Furthermore, participants indicated that they want to transfer their knowledge and enrich the lives of their learners in the future. “As a teacher, it is my mission to instill in my students the valuable skills of computation and programming. I want to create a generation that actively contributes to the digital world rather than merely being passive consumers because I recognize the growing relevance of technology in our lives” (P19).

5. Discussion

The discussion will first focus on the first research aim, to suggest a learning environment for incorporating productive failure in C&R to promote DSDL. Thereafter, the second research aim, to explore what participants’ self-reflections reveal regarding the implementation of productive failure, will be discussed.

5.1. Incorporating Productive Failure in Coding and Robotics

Figure 3 illustrates the adapted productive failure intervention implemented in the introductory C&R course to foster deeper self-directed learning.

Cooperative pair programming was employed as the teaching and learning strategy. In week 1, participants worked in cooperative pair programming groups to solve problems that were both solvable and challenging, as this was their first encounter with C&R. Thereafter, they individually completed guided reflections. In week 2, the cooperative pairs were given problems that appeared solvable but were in fact unsolvable, followed once again by the individual completion of guided reflections.

Participants had very limited prior knowledge of C&R. They first had to familiarise themselves with the MakeCode and micro:bit environment. Subsequently, the solvable problems that they initially received in the first week of the intervention can also be seen as complex problems, because of the new C&R environment that had to be mastered. From the pre-PF reflections, it was accordingly evident that learning initially focused on understanding the micro:bit environment and programming concepts. The challenges, accomplishments and long-term goals of participants could all initially be related to understanding C&R concepts.

The structured cooperative pair programming environment provided a safe environment for participants to explore and learn from one another. The group roles of navigator and driver ensured that each group member was accountable for contributing to the task at hand. Although time was limited, engaging in a social constructivist environment, where multiple working memories collaborated to explore the new environment and solve the C&R problem, accelerated learning (see Sweller et al., 2019; see Vygotsky, 1978).

From the outset, participants were curious, engaged, used online resources and were enthusiastic to experiment with the micro:bit. Initially, trial-and-error problem-solving strategies were employed, followed later by more structured problem-solving approaches.

The emergence of participants’ enjoyment, engagement, curiosity and perseverance towards C&R is noteworthy because CAT education students usually focus on solving problems by using word processors, spreadsheets and database software. Moreover, participants’ contentment with collaboration with a group member randomly allocated to them, and with limited guidance from lecturers, is also notable.

Post-PF reflections revealed that participants were not discouraged by the unsolvable problems but, surprisingly, even more motivated by those problems. They searched for a multitude of online resources and did their own research to prepare for class. It seems that not limiting participants in their efforts by giving them ill-structured unsolvable problems encouraged them to think creatively, to apply computational thinking processes and to follow a more structured approach to problem solving. Small successes became intrinsic motivators, encouraging continued efforts.

Both pre-PF and post-PF reflections revealed competencies in the cognitive, intrapersonal, and interpersonal domains (see Section 4.2.3), which suggest the development of DSDL (Van Zyl & Mentz, 2020). Post-PF reflections revealed that participants envisioned transferring their knowledge and skills to their future careers. Follow-up, longitudinal research is, however, required to determine if the transfer of knowledge and skills to new contexts occurred.

5.2. Increased Depth of Post-PF Reflections

A greater depth in cognitive processing was evident in the post-PF reflections. Figure 4 depicts the increased depth of reflections as reported by participants. Three parallel processes were identified. As indicated in Process A, the initial reflections focused on understanding concepts and finding resources while continuously identifying learning needs. Process B depicts group cooperation, which served as a scaffold for learning to deepen reflections. Process C depicts how participants developed confidence, persisted in problem solving and became aware of the problem-solving strategies they applied. Metacognitive processes became prominent as participants reflected on their thinking, assessed the efficacy of the strategies they applied, and offered suggestions for enhancing their approaches.

A heightened sense of group cohesion and teamwork emerged (See Figure 4, Process B). The dynamics of group interaction transitioned from primarily assisting each other in comprehending concepts, as evident in the pre-PF reflections, to engaging in collaborative brainstorming sessions aimed at generating potential solutions. The bolstered group support emerged as a valuable resource, serving as a catalyst for maintaining focus, sustained engagement, and resilience in the face of challenges. Instances where participants shifted from relying on group support to pursuing individual learning paths may suggest a transition toward assuming ownership of learning. The long-term goals of participants showed progression towards lifelong learning and competence in C&R. Despite complex tasks and challenges, participants’ enjoyment and interest in C&R were sustained and they expressed pride in their learning journey with C&R. Furthermore, as prospective teachers, they envisaged themselves as enriching their learners’ lives one day by facilitating C&R and problem-solving skills, thereby indicating the potential for developing DSDL characteristics (see Figure 4, Process C).

6. Conclusions

The findings provide preliminary evidence that the adapted productive failure intervention may foster DSDL characteristics. The improvement in learning over a brief period of two weeks is noteworthy, especially considering the challenging tasks undertaken by participants who lacked prior experience in C&R. It is therefore concluded that the effective structuring of groups, through the integration of cooperative pair programming, not only enables the mitigation of challenges inherent in complex tasks through group support mechanisms but also expedites the learning process.

It is further suggested that in C&R, the initial activities should consist of solvable problems, to allow students to explore the new environment and build their confidence (see Table 1). Students should also be allowed to compile their own problem scenarios and not be limited to the problems assigned by facilitators. Instead of following up on the first phase of productive failure with direct instruction, as suggested by Kapur and Bielaczyc (2012), self- and peer-assessment, as well as self-reflection, can be used in the second phase of productive failure in the context of C&R.

The findings indicate that incorporating productive failure with cooperative pair programming in an introductory C&R course in CAT education resulted in a “beautiful risk” (Beghetto, 2018). Although the findings exhibit promise, a stance of modesty regarding their implications is maintained, due to several limitations of the research. The research was applied in a limited period of two weeks, and a single self-report method was used as a data gathering method. Moreover, due to the small population and for ethical reasons not to exclude students from innovative teaching and learning strategies, no control group could be used. The dual role of lecturers as researchers could have potentially influenced participants’ responses. Although preliminary evidence of DSDL characteristics has been found, further longitudinal research is needed to confirm transfer and long-term development. More research on other populations and follow-up research is required to determine the transferability of the findings. We further acknowledge that the research cannot be generalized to other populations and encourage further research in other populations and other subjects.