Learning Performance of Different Genders’ Computational Thinking

Abstract

While the role of computational thinking (CT) has been widely reported in technology applications and further integrated into interdisciplinary learning, the integration of pedagogy-supported interdisciplinary activities for the empowerment of girls’ learning must not solely emphasise CT problem-solving skills. Rather, it must scaffold them with interactive learning that supports their characteristics while catering to gender equality. In this study, a gender-balanced interdisciplinary activity, integrating CT with Mandarin learning (ML), was designed for an elementary school in the Mandarin as a Second Language learning context using Social Robots (SRs). It sought to verify the results of the proposed method along with focused activities and interaction in an SR-integrated activity on the CT abilities and target-language learning of young learners. A total of 46 Grade 5 students, 26 boys and 20 girls, participated in the experiment. The study used a quasi-experimental method by examining the result of pre- and post-tests on language acquisition, programming self-efficacy, the educational robot attitude, and learning perceptions in the activity. The results indicated that there were no gender differences in terms of ML, self-efficacy in programming, or attitudes toward using SRs in the SR-integrated interdisciplinary activities. However, the boys and girls had different perceptions of learning. Suggestions for conducting SR-integrated interdisciplinary learning are given, along with pedagogical implications for the further promotion of women in technology.

1. Introduction

Increasing attention has been paid to children’s acquisition of computational thinking (CT) skills in order to learn to overcome potential difficulties that are to be faced in the 21st century [1]. In accordance with this need, many innovative approaches and methods of fostering such skills have emerged, such as the use of educational robots (ERs). Research has shown that ER activities have proceeded, such as Papert’s [2] constructionism introduced in the past few decades, which sees learning as active construction and supports hands-on, visualised, and tangible experiences [3]. It positively affects children’s learning from a low threshold of leveraging students’ interests to a kernel CT practice. Examples of such ER activities in CT are problem-solving development (Chevalier et al., 2020), pair programming [4], and unplugged CT board games [5]. Indeed, after engaging in ER activities, students are introduced to CT development; this includes but is not limited to developing CT thinkers, CT creators, and CT innovators.

Past studies have also presented that ERs can combine CT development and other cross-disciplinary learning. For example, Hsu et al. [6] used ERs to develop CT with cross-disciplinary English language learning. They extrapolated Papadaki’s [7] suggestion on CT effective communication for CT fluency. Others, such as Burhans and Dantu [8], used ERs in music learning, and Angeli and Valanides [9] used BeeBot to make ER coding attractive to girls. By tailoring ER learning beyond Science, Technology, Engineering, and Mathematics (STEM), children developed CT concepts—not only coding but also creative thinking, effective expression, and the integration of interdisciplinary learning in the ER activity (e.g., [10,11,12,13]). In essence, various advantages arise from designing ER activities based on constructionism; young learners are afforded the chance to boost their basic understanding of CT concepts, to be involved in an interdisciplinary path to transfer CT concepts, and to build positive learning attitudes and programming self-efficacy (e.g., [14,15,16]). The current study was grounded in constructionism to design hands-on ER activities for cross-disciplinary learning (CT and Mandarin language (ML) learning). As ERs can be visually programmed, this has the potential not only for CT development but also as a novel means of encouraging students’ ML when involved in the ER coding task.

It has been acknowledged that despite the growing attention and possibilities supported by institutions or environments to tailor constructionism-related CT activities for young learners [17], few studies have paid attention to gender differences in designing ER coding activities for children [17]. As Cheryan et al. [18] reported, there are unbalanced ratios of participation between male and female students in the field of computer science, and the number of women involved in computer science is surprisingly low. Even with the rapid development of technologies, the problem of gender discrepancy in STEM still exists on a global scale.

Indeed, it may not be easy to alter the results without attracting more women to computer science by carefully designing pedagogically meaningful ways to develop their interdisciplinary learning. Papavlasopoulou et al. [17] found that the gender gap in CT competence was almost non-existent as long as the curriculum developers recognised the gender differences in the strategies and approaches they adopted. However, merely assembling relevant elements (e.g., SRs, CT, ML, gender) together cannot guarantee the obtention of the anticipated results. Instead, ER interdisciplinary activities cannot only be concerned with cross-disciplinary learning (CT development and ML learning, in this case) when engaging in ER coding activities, but the tailored activity needs to consider gender differences catering to gender characteristics (particularly girls) while supporting their positive learning attitudes and programming self-efficacy.

One way to achieve anticipated goals is through the use of social robots (SRs), as they potentially present a promising approach to help students access CT concepts and engage in conversation about collaborative programming [19,20]. SRs also have the potential to reduce the gender gap in CT learning, along with interdisciplinary activities. They go beyond “an autonomous or semiautonomous robot that interacts and communicates with humans by following the behavioural norms expected by the people with whom the robot is intended to interact” ([21], p. 592). With advanced technology, SRs provide learners with the use of graphical coding blocks to access CT concepts and receive instant responses, along with social interaction, which aids in language development (e.g., ML), or other applications such as vocabulary acquisition (e.g., [22]). The SRs used in this study helped support instructors in the design of innovative interdisciplinary learning activities that developed students’ CT concepts and target-language learning, along with focused activities in an interactive manner catering toward a gender-balanced way.

Concisely, the goal of this study was to design and evaluate an interdisciplinary activity catered toward Grade 5 children aged 11 years with an instructional design that took gender differences into account. SR-integrated learning activities with a pedagogy-informed design were deployed in order to situate learners in the physical and embodied learning experience. These activities could be conducive to supporting boys’ and girls’ programming self-efficacy and ML in interdisciplinary learning while promoting their positive learning attitudes toward the use of SRs in the ML learning classroom. Thus, we wanted to answer the research question: Do SR-integrated interdisciplinary activities support learning? If so, are there gender differences in the SR-integrated interdisciplinary activities in terms of ML knowledge, programming self-efficacy, attitudes toward SRs, and learning perceptions? The following hypotheses were examined first: When integrating SRs into learning,

(1)

The girls had a similar achievement in CT to the boys.

(2)

The girls had a similar achievement in programming self-efficacy to the boys.

(3)

The girls had similar learning attitudes towards SR to the boys.

(4)

There was no difference in boys’ and girl’s learning perceptions of the instruction.

2. Related Work

2.1. CT Construction

While there has been a focus on children’s competence in CT skills, Papert’s [2] constructionism played a key role in advocating for young learners to think algorithmically when they used Logo to program. Papert [2] argued that the most powerful learning takes place when students are actively involved in the construction of a project or digital artefacts, since they can rely on their existing knowledge and explore new knowledge. While the term “object-to-think-with” was widely adopted by many teachers to promote CT skills after Papert’s promotion [17], Wing [23] made it popular by addressing that designing CT activities does not just involve coding or programming skills. Instead, activities consist of the skills of decomposing problems (from difficult issues to easier ones), constructing algorithms (problem-solving procedures in detail), abstraction, and adopting the attitudes that computer scientists generally hold to solve problems. Indeed, Papert’s and Wing’s efforts made CT a 21^st century literacy. A future-ready student can hardly solve problems in a wide range of contexts without applying CT techniques, since CT is a skill set that provides students with a transparent way to focus on the process and solution, supports their cognitive development, and helps them solve problems using technology.

2.2. ERs, CT, and Gender

In spite of the increased research and options offered by visual block programming learning to design CT learning, relatively few studies have focused on gender issues in designing activities for children [1,17]. Gender discrepancies in science, technology, engineering, and mathematics (STEM) exist, with women being less prevalent in the computer science field [18]. According to Ceci and Williams [24], the gender gap starts in elementary school, with imbalanced pedagogical approaches. A general lack of positive educational experiences in the learning process makes women unwilling to follow computer science paths [25], although many complex factors are also involved (e.g., fear of involvement in programming courses or stereotypes of career choices in computing [26]. The number of women in technology has significantly decreased over the past few decades [27], and there are relatively low ratios of females entering the field of computer science. While aiming to enhance girls’ empowerment in interdisciplinary learning, pedagogically informed activities that support gender equality must be taken into account.

Understanding the neuroscience of gender differences is crucial. Girls’ hippocampus, in charge of memory and language, develops more rapidly and is larger than that of boys [28]. Boys’ cerebral cortex, which manages spatial relationships, develops better than that of girls. These differences affect the skill development and ways of learning of both genders. Boys learn better through movement and visual experience but have trouble with long-term concentration, while girls perform better through collaborative activities but need more spatial skills training. Briefly, their ways of learning can be summarised as follows: girls are adapted to listening to others and sitting still to work on an artwork and like to share ideas with others, while boys prefer individualistic, kinaesthetic activities that they can control [29].

Papavlasopoulou et al. [17] and Angeli and Valanides [9] shared a similar view on the gender issue. Papavlasopoulou et al. [17] stated that there was a very big difference in the strategies and approaches that boys and girls used during CT activities. The gender gap in CT competence was almost non-existent if the teaching techniques design supported students’ understanding of CT principles and inspired both boys and girls to explore. Angeli and Valanides [9] suggested that instructional design should accommodate differences in the planning and implementation of coding activities, since girls have different strategies in CT learning compared to boys.

In terms of using ERs to promote CT, Bers et al. [30] provided a clear summary of the benefits of ER use, including engaging children in active, playable, and hands-on tasks via programming tangible ERs. Hsu et al. [6] further examined the influence of CT development with ERs, helped children to achieve attained goals, and reduced learning anxiety, while Tengler et al. [11] evidenced the effect of CT competency in the combination of storytelling, texts, and literature using ERs activity. Bers et al.’s [30] and Hsu et al.’s [6] studies showed the similarities of how effective ERs are for introducing CT skills via visualised programming; however, gender issues aligning ERs with CT development have not been fully explored. As we were interested in discovering gender issues when students were poised to target language interaction using the ML during the CT task, the experiment aimed to accommodate both genders, offering space and movement for boys and idea sharing (cooperative learning) for girls when involving them in ER coding activities.

2.3. SRs and Gender

Social robots (SRs), with their embodiment and ability to add social interaction to the learning context, are generally attractive to young learners [22]. With the rapid progress in technology, SRs can not only autonomously interact and communicate with humans but also allow learners to build, create, and program using graphical coding blocks to access CT skills in an interdisciplinary way (e.g., [31]). Although there has been no direct report exploring gender differences in interdisciplinary activities using SRs, SRs have been observed to increase communication skills [32] and collaboration skills [22], cognitive knowledge [33], and creativity [34]. SRs are an educational medium that facilitates social interaction among students themselves or students with the SRs (e.g., [22]) while also affecting CT access and any other disciplinary skills involved, such as language learning [35] or physics [36]. However, a constructionism-based CT acquisition in this disciplinary study may not be fully attained without reinforcing pedagogically informed gender-balanced instructional design.

Thus, the main concern in the current context was promoting the interdisciplinary learning of CT practices and ML in a gender-balanced way; meanwhile, the course objective was to foster students’ self-efficacy in CT concepts and linguistic gain in Mandarin learning. A guided instructional approach with theoretical support to take advantage of SRs is imperative when aiming to cope better with gender gaps.

First, SRs, with physical bodies that play the role of a human through vocal, gestural, and facial expressions, were used to help address the research questions. The SRs used in this study, called Kebbi, contain artificial intelligence and can be applied to real-world applications such as ambient intelligence, artificial life, game playing, and social interaction with verbal and nonverbal communication with teachers or students in classrooms (https://www.nuwarobotics.com/en/product (accessed on 1 December 2021).

Three roles of SRs (tutor, peer, and tool) in line with pedagogical support [37] were tailored to learning tasks in compliance with gender differences. As a tutor, the SR performed actions and gestures with animated expressions (e.g., acting out a story in the classroom) that increased children’s attention to the learning materials. The SRs’ gestures, eye gaze, and voice tone are powerful cues to supplement language production [35]; many studies have evidenced their effectiveness in terms of children’s attention, engagement, and motivation in language development (e.g., [22]). When tailored as peers, SRs can have active spontaneous participation and give feedback to learners. For example, while children are involved in programming robots, SRs can serve as peers that provide explicit feedback (e.g., “Good job!”) or can read the code out loud. When acting as tools, SRs are adaptive through sensors to adjust lesson levels according to learners’ needs, as has been widely used in previous studies (e.g., [37]). For example, SRs are affordable and offer interactive games that involve students in cooperative and problem-solving learning tasks (e.g., [36]).

Second, Brennan and Resnick’s [38] CT framework was adopted to support the tailored SR-integrated interdisciplinary activities, which supported students in actively constructing their learning [2]. Three dimensions of CT from Brennan and Resnick’s [38] CT framework were affordable and suggested: computational concepts (concepts that participants are involved in when they program, such as variables); computational practices (practices that participants perform, such as debugging); and computational perspectives (perspectives that participants develop for CT, themselves, and the world around them). Therefore, individual concepts, collaborative learning and interaction, and the environment they are in are all included in CT skills [39].

When adopted in this study, their framework enabled us to monitor the SR-integrated interdisciplinary activities and to understand how G5 students dealt with the CT concepts embedded in the Mandarin content, how their self-efficacy and learning attitudes affected their CT practices in the given task, and, finally, how boys’ and girls’ perspectives evolved in relation to themselves and the advanced SR world.

3. Research Methods

3.1. Participants

A total of 46 Grade 5 students, including 26 boys and 20 girls, who were taking Chinese as a Second Language in Singapore took part in this study. None of the students had ever accessed SRs before. A teacher with 10 years of teaching experience and who was familiar with technology-integrated lessons was appointed to the study. The students worked cooperatively to complete the tasks, where two students formed a group, similar to paired programming. With regard to ethics, the researchers adhered to the Ministry of Science and Technology’s rules (ROC). Participants were offered the option to withdraw from the experiment at any time without any grading effect.

3.2. Instructional Design

Kebbi, the social robot, was applied to the study as a story play actor with verbal and nonverbal interaction with the students, a companion for the pair of students during the coding activities, and a supplementary tool for gameplay between students in the classroom.

Two specific sessions were designed for the interdisciplinary activities: one for the Mandarin lecture demonstrations and one for the CT integration sessions (Figure 1). The content of both sessions was directly interwoven. The previous session of lecture demonstrations introduced a problem-solving task, derived from a unit of the textbook, ‘Stone Soup’, in the Mandarin lecture. The tale was a story of collaboration and sharing, wherein two hungry people convince a town that they can make soup from a stone. The result benefits from the individual ingredients that any involved villager is able to provide. The story is applicable to different disciplines and areas of expertise. Students then practised the sentence constructions of conditional compound sentences in Chinese. The aim was to create a desire to learn more about the CT workings shown by the SRs. Mandarin-medium instruction using PowerPoint in this section not only taught the vocabulary, idioms, and sentences in the story but also served as a threshold concept to explain CT. It enabled students to find similar activities during the computational sessions, during which problem-solving issues occurred on a daily basis.

The CT integration session (embedded in ML) gave students additional hands-on time with the physical and embodied SR access related to the Mandarin lecture.

Table 1. The CT integration session using SRs.

Phase	Focused Activity	SR’s Role	Explaining
1	Story play	Tutor	Phase 1 focused on story-play activities by reinforcing story content and related vocabulary items (Figure 2a,b). Relevant sentences were highlighted to students. Boys and girls (two students formed a group) were invited to act out the story with the SRs, aiming to inspire curiosity in order to generate later interest in the coding activities.
2	Coding activity	Peer	Phase 2 served as a coding activity requiring students to solve a problem (related to the Mandarin content, Stone Soup) using the conditional structure in the block-based programming embedded by SR itself and to control the SR’s speaking in line with the logic they designed with blocks. Students received feedback from SRs, such as “good job”, when giving a correct structure; they followed the SRs to practise the target language (Figure 2c,d). SR-facilitation in this stage not only gave them affirmative feedback but also made them feel that they were accompanied if debugging happened.
3	Reviewing gameplay	Tool	Phase 3 entailed a game-play activity after completing the coding activity. Two students cooperatively answered the questions; the instructor provided assistance to review the learnt CT concepts embedded in Mandarin (Figure 2e,f). This stage aimed to motivate both genders’ learning while improving girls’ confidence in their programming ability.

depicts the three phases of each session as a series of focused SR roles, as well as their estimated durations.

3.3. Research Process

This study adopted a quasi-experimental method. A 5-week experiment period was based on the two sessions, with two periods for Mandarin (two weekly E-book lectures and one weekly pre-test of 6 h in total) and two periods for CT integration (one weekly hands-on activity and a post-test of 3 h in total).

The previous two periods (the first session) were as follows: the students took the pre-Mandarin test, performed the self-efficacy pre-questionnaire for programming, and learnt the story in Mandarin and the basic CT concepts embedded in the target language. The last two periods (the second session) were a hands-on activity, in which students applied the learnt CT concepts and learnt to program using the SRs. After completing the three-phase activities, the post-test of Mandarin and the post-questionnaires (self-efficacy in programming, attitudes toward SRs, and learning perceptions) were administered in the ML classroom.

3.4. Instruments

This study used four instruments: (1) pre- and post-tests on the Mandarin linguistic knowledge; (2) programming self-efficacy scales; (3) the educational robot attitude scale; and (4) learning perceptions in the activity.

The pre-test and post-test of Mandarin proficiency comprised 32 multiple-choice questions, including vocabulary items (14 items worth 28 points), phonetic use (12 items worth 48 points), and word meanings (6 items worth 24 points), with a full score of 100. To validate both tests, one experienced Mandarin instructor and one technology education instructor were recruited. Two experts along with the researchers ensured the reliability of the tests.

Three questionnaires were applied. First, the programming self-efficacy scale, adapted from Tsai et al.’s [40] computer programming self-efficacy scale, with a 5-point Likert scale (1 = strongly disagree; 5 = strongly agree) was used to evaluate the students’ attitudes toward logical thinking, control, and debugging. This study used three dimensions of four items for the logical thinking dimension, three items for the control dimension, and three for the debugging dimension, with a Cronbach’s alpha value of 0.91, showing acceptable reliability.

Second, the educational robot attitude scale (ERAS) was adapted from Sisman et al.’s [41] ERAS, with three dimensions of engagement, anxiety, and intention, using a 5-point Likert scale (1 = strongly disagree; 5 = strongly agree). This study used five items for learning engagement, four for learning anxiety, and four for students’ intention to use SRs, with a Cronbach’s alpha value of 0.83, showing acceptable reliability. It was administered after the ERs with the CT and ML integration session.

Lastly, the learning perception scale was adapted from Rubio et al.’s [42] study on exploring the learning perceptions of genders in an introductory programming course, with three dimensions of story play activity, coding activity, and review gameplay activity, using a 5-point Likert scale (1 = strongly disagree; 5 = strongly agree). This study used four items for story play activity, four for coding activity, and six for reviewing gameplay, with a Cronbach’s alpha value of 0.78, showing acceptable reliability, administered after the ERs with the ML and CT integration session.

3.5. Data Analysis

Prior to the study, the Shapiro–Wilk Test was first conducted to ensure the normality of the data due to the small sample size. In light of the results, the values of the Shapiro–Wilk Test were 0.942 (p > 0.05) for boys and 0.913 (p > 0.05) for girls in Mandarin achievement; 0.932 (p > 0.05) for boys and 0.910 (p > 0.05) for girls in self-efficacy of programming activity; and 0.948 (p > 0.05) for boys and 0.939 (p > 0.05) for girls in the learning attitude toward ERs. The data in these three collections were normally distributed.

Therefore, to answer RQ1, the paired sample t test and the ANCOVA method were used to analyse the effects of Mandarin achievement (the independent variable as gender vs. the dependent variable as Mandarin achievement). With regard to the questionnaire on self-efficacy in programming activity in RQ2, paired sample t tests and the ANCOVA method were also used to analyse the effects of self-efficacy programming in ER interdisciplinary learning (the independent variable as gender vs. the dependent variable as self-efficacy in programming). For the learning attitude toward ERs in RQ3, the MANOVA test was adopted to examine the differences between boys and girls (the independent variable as gender vs. the dependent variables as learning engagement, learning anxiety, and intention to use SRs).

However, the values of the Shapiro–Wilk Test were 0.856 (p < 0.05) for boys and 0.800 (p < 0.05) for girls in students’ perception of the SR-focused activities and were not normally distributed, which indicated the choice of the nonparametric test suitable for the analysis. Therefore, a Mann–Whitney U test, a non-parametric hypothesis test, was applied to two independent samples. We used the Mann–Whitney U test, the equivalent of an independent samples t test, to examine the differences in the learning perceptions of boys and girls during the SR-focused learning activities (the independent variable as gender vs. the dependent variables as story play activity, coding activity, and reviewing gameplay activity).

4. Result

4.1. Mandarin Achievement

The purpose of this study was to examine if boys and girls had different learning outcomes and perceptions when taking part in the interdisciplinary activities of ML and CT integration. Descriptive statistics for the means and standard deviations relating to the two groups are shown in

Table 2. Descriptive statistics of the boys’ and girls’ ML tests.

	Boys			Girls
	N	Mean	SD	N	Mean	SD
Pre-test	26	51.04	20.68	20	59.56	23.27
Post-test	26	75.00	14.02	20	79.40	14.11

.

To measure students’ ML results in their ER interdisciplinary learning, the t test was used to ensure that both groups had equivalent ML knowledge before the activity. As no significant difference (t = 1.05, p > 0.05) in the pre-test of ML between the boys (M = 26; SD = 20.68) and the girls (M = 59.56; SD = 23.27) was found, paired sample t tests were first used to investigate the progress of ML (

Table 3. Paired sample t tests of the boys’ and girls’ ML tests.

	N	Pre-Test		Post-Test		t	Effect Size
		Mean	SD	Mean	SD
Boys’ ML	26	51.04	20.68	75.00	14.02	−8.06 ***	1.35
Girls’ ML	20	59.55	23.27	79.40	14.11	−4.83 ***	1.03

*** p < 0.001.

). Both genders significantly improved in their ML (Table 2)—boys: (t = −8.06 ***, p <0.001, ES = 1.35) and girls: (t = −4.83 ***, p < 0.001, ES = 1.03). Both genders made significant improvement in their linguistic knowledge of ML, showing that the interdisciplinary activities were beneficial for ML acquisition.

ANCOVA was carried out to find the differences in the ML of the two groups. As the Levene’s test of homogeneity was not violated (F = 0.06; p = 0.81 > 0.05), covariance analysis could be further performed to interpret the difference in ML between boys and girls. The result showed that no significant difference was found in the ANCOVA statistics (F = 0.07; p = 0.79 > 0.05) between boys’ ML (M = 75.00) and girls’ ML (M = 79.40). Boys (Adjusted mean = 76.54) had similar learning achievements in ML to those of girls (Adjusted mean = 77.39) (

Table 4. ANCOVA of the learning comprehension test.

Group	N	Mean	SD	Adjusted Mean	F	P
Boys	26	75.00	14.01	76.54	0.07	0.79
Gils	20	79.40	14.11	77.39

).

4.2. Self-Efficacy in Programming

To measure students’ self-efficacy of programming activity in ER interdisciplinary learning, descriptive statistics for the means and standard deviations relating to the two groups are first shown in

Table 5. Descriptive statistics of the boys’ and girl’s self-efficacy programming.

	Boys			Girls
	N	Mean	SD	N	Mean	SD
Pre-test	26	8.92	3.28	20	8.13	3.31
Post-test	26	12.04	2.71	20	11.66	2.20

.

The t test was used to ensure that both groups had equivalent self-efficacy programming before the activity. The result showed that no significant difference (t = 0.804, p > 0.05) in the pre-test of students’ self-efficacy of programming between the boys (M = 8.92; SD = 3.28) and the girls (M = 8.13; SD = 3.31) was found. Further, paired sample t tests were examined (

Table 6. Paired sample t tests of the boys’ and girls’ self-efficacy.

	N	Pre-Self-Efficacy		Post-Self-Efficacy		t	Effect Size
		Mean	SD	Mean	SD
Boys’ self-efficacy	26	8.92	3.28	12.04	2.71	−3.92 **	1.03
Girls’ self-efficacy	20	8.13	3.13	11.66	2.20	−5.42 ***	1.30

** p < 0.01, *** p < 0.001.

), and the results for both genders improved significantly—boys: (t = −3.92 **; p < 0.01, ES = 1.03) and girls: (t = 5.42 ***; p < 0.001, ES = 1.30). The activities did contribute to reducing the gender gap in the CL self-efficacy programming activities.

ANCOVA was carried out to find the differences in the self-efficacy programming of the two groups. As the Levene’s test of homogeneity was not violated (F = 1.29; p = 0.26 > 0.05), covariance analysis could be further performed to interpret the difference in ML between boys (M = 12.04) and girls (M = 11.66). The result showed that there were also no significant effects of gender differences in terms of the three dimensions of logical thinking (F = 0.14; p = 0.70 > 0.05), control (F = 0.007; p = 0.93 > 0.05), and debugging (F = 0.31; p = 0.58 > 0.05) on students’ self-efficacy in programming (

Table 7. Programming self-efficacy ANCOVA test.

Dimensions	Group	N	Mean	SD	Adjusted Mean	F	p
Logical thinking	Boys	26	4.07	0.95	4.05	0.14	0.70
	Girls	20	3.92	0.71	3.95
Control	Boys	26	3.93	1.00	3.91	0.007	0.93
	Girls	20	3.90	0.97	3.93
Debugging	Boys	26	4.02	0.90	4.00	0.31	0.58
	Girls	20	3.83	0.92	3.85

). The boys’ and girls’ programming self-efficacies were similar.

4.3. Attitudes toward Social Robots

One-way MANOVA was conducted to rule out the discrepancy of learning attitudes toward social robots between the two genders during the learning activities. The results showed that there was no significant multivariate effect on three items between genders: Wilks’ Lamba = 0.95; F = 42, p = 0.54 > 0.05. There were also no significant effects of gender differences in terms of the three dimensions of engagement (F = 1.09; p > 0.05), anxiety (F = 0.67; p > 0.05), intention to use SRs (F = 0.02; p > 0.05), and overall attitude (F = 0.83; p > 0.05) on students’ attitudes toward the educational robots (

Table 8. Educational robot attitude scale MANOVA test results.

Variable	Gender	N	M	SD	F	p
1. Learning engagement	Boys	26	4.32	0.63	1.09	0.30
	Girls	20	4.53	0.73
2. Learning anxiety	Boys	26	1.69	0.96	0.67	0.42
	Girls	20	1.48	0.69
3. Intention to use SRs	Boys	26	4.44	0.58	0.02	0.96
	Girls	20	4.43	0.79
4. Total	Boys	26	3.65	0.38	0.83	0.89
	Girls	20	3.67	0.43

). Boys’ attitudes toward SRs were similar to those of girls.

4.4. The Interdisciplinary Learning Perception

The Mann–Whitney U test was conducted to examine the perceptions of the two genders during the learning activities. The results of the Mann–Whitney U test analysis revealed that learning perceptions for the two genders were not similar. The learning perceptions of the story-play activity for girls (mean rank = 27.88) were statistically significantly higher than those for boys (mean rank = 20.13), with U = 172.50, z = −1.99, p < 0.05. The learning perceptions of the coding activity for girls (mean rank = 28.73) were statistically significantly higher than those for boys (mean rank = 19.48), with U = 155.50, z = −2.35, p < 0.05. The learning perceptions of reviewing the gameplay activity for girls (mean rank = 28.08) were statistically significantly higher than those for boys (mean rank = 19.98), with U = 168.50, z = −2.05, p < 0.05. Girls had more positive perceptions of the three tailored activities in terms of story-play activity, coding activity, and reviewing gameplay activity than boys (see

Table 9. Learning perception Mann–Whitney U test results for the two genders.

Variable	Gender	N	Mean Rank	Sum of Ranks	U	z
1. Story play activity	Boys	26	20.13	523.50	172.50	−1.99 *
	Girls	20	27.88	557.50
2. Coding activity	Boys	26	19.48	506.50	155.50	−2.35 *
	Girls	20	28.73	574.50
3. Reviewing gameplay activity	Boys	26	19.98	519.50	168.50	−2.05 *
	Girls	20	28.08	561.50

* p < 0.05.

).

5. Discussion

5.1. Summary

While ERs are increasingly used in CT knowledge promotion, many educators align them with interdisciplinary activities and focus on learning content beyond STEM in other cross-disciplinary studies (e.g., [5,6,22]), recognising that the ERs we access, touch, and make are often grounded by Papert’s constructionism, where learning is supported in a hands-on, practical, and tangible manner. ER activities with an interdisciplinary design should not be the only target of problem-solving skills and content acquisition. Rather, gender differences in interdisciplinary activities must be integrated into the teaching design when using ERs, such as SRs, to develop activities that successfully build CT concepts and embrace both boys and girls in a balanced way. By being rooted in constructionism-based learning in a hands-on learning manner, students are introduced to the cross-disciplinary ML learning, supporting their programming self-efficacy while enhancing their positive learning attitudes towards ERs. Coupled with existing research that has effectively used SRs to show positive learning results across subjects (e.g., [6,22,35]), instructional designs using critically developed activities with pedagogically informed approaches that examine their outcomes in terms of gender are lacking. More studies can be conducted that consider whether any available SRs with instructional support are engaging, hands-on, and gender-friendly for the particular context in which students are involved in cross-disciplinary study, as merely combining relevant elements (e.g., SRs, CT, ML, genders) cannot obtain the intended results.

5.2. Key Point: CT Skills and ML with Focused-Activity Design

Learning through a pedagogically tailored SR activity in a gender-friendly approach, both boys and girls improved significantly in terms of learning CT and ML. Children were introduced to interacting with SRs, which were especially designed to perform three roles in the activity, namely, as a tutor for a story-play, as a peer for a coding activity, and as a tool for a reviewing game. Our aim is not to claim which gender is better than the other when examining the differences in the learning results of the two genders. Instead, the results presented in the current study merit consideration as pioneering for SR interdisciplinary designers and CT practitioners, who should be aware of the conditions and demands placed on gender characteristics and used in cross-disciplinary activity design. It is a privilege to make children believe that everyone can learn to code, not only boys, while engaging in ML interaction. Indeed, the results showed that the integration of essential CT skills and ML acquisition in a gender-balanced manner could be jointly implemented using the ER-integrated three-role strategies with storytelling, coding, and reviewing gameplay activities, as proposed in this study. The SR-integrated strategy, along with focused activities, can be established using an immersive design providing a flexible setting for gender-friendly learning but a manageable level of effort for ML output acquisition. This subtle design is viewed as being highly vital to studies examining gender issues in cross-disciplinary learning in academic contexts, and we argue that if young learners do not find SR-integrated activities appealing, they will not participate in discussions with their peers regarding any prospective CT techniques [3], nor will they cooperate with their peers in coding activities while joining gameplay review and story act-out activities.

5.3. Critically Analysing Learning Outcomes

According to the findings, no gender differences were found concerning ML competence, self-efficacy programming for the coding activities, or learning attitudes toward SRs. These findings indicate that hypotheses H1, H2, and H3 were supported. The interdisciplinary activity using SRs proved to be helpful for scaffolding students’ learning during their engagement with the CT integration activities and met the need of aligning teaching activities with genders. The findings also provide more evidence that girls do not lack competence compared to boys, echoing Papavlasopoulou et al.’s [17] finding in a study investigating the gender gaps for young learners that females have similar competence in programming as males, but approaches and strategies during coding activities differ. They suggested that CT practitioners should focus on characteristics that will foster girls’ self-efficacies and avoid discriminating behaviours. In another study on gender-balanced performance in a Roberta project, Bredenfeld and Leimbach [43] found no gender differences and reported that males and females were equally capable of developing the CT skills that are required to be specifically tailored when using robots. The design, aligning with these two studies, proves the need to carefully design an instruction that caters to both genders’ characteristics.

To the best of our knowledge, no other studies have investigated gender differences that tailored CT concepts embedded in an ML context using SR-supported techniques and tailored SRs’ roles with theoretical support to improve classroom teaching. Çakır et al.’s [44] study successfully engaged students’ attitudes towards computing and enhanced girls’ self-reported confidence and competence with computers, but they did not integrate interdisciplinary learning into formal classes and involved girls in only a full-day workshop.

Meanwhile, it is acknowledged that students are able to write code and know how to program, which is in line with Brennan and Resnick’s [38] framework on CT design and the SRs’ roles. Serving as a tutor to act out the story pertaining to the ML lecture, SRs highlighted CT concepts from the story that children engaged with when they were programming. Acting as a peer to support coding activities, SRs facilitated the computational practices that children actually developed. For example, students abstracted the concepts from the story and started coding and debugging using the IF conditional structure. SRs gave feedback by either saying “good job” or speaking the sentences again if students wrote the code correctly. Lastly, playing the role of tools, SRs helped in focusing on computational perspectives (Brennan and Resnick’s [38] model) that boys and girls developed for computation, themselves, and the world around them. For example, SRs reinforced CT concepts embedded in MT content by reviewing the material in the previous session and refocusing on CT concepts on a daily basis (e.g., stone soup story). This suggests that acquiring a certain level of coding skills requires a pedagogy-informed design. Brennan and Resnick’s [38] framework enabled the authors to monitor scaffolding in the development of children’s CT skills and ML while learning using Kebbi.

Moreover, girls’ perceptions of the learning activities were more positive than those of boys, showing that gender differences exist in children’s access to activities, which allowed us to reject hypothesis H4. However, it corresponds to previous studies which found that using educational robot approaches can successfully improve women’s attitudes. McGill [45] supported this interesting difference, stating that robots make females more confident because they give a sense of empowerment to females, after examining the learning perceptions of boys and girls in the use of robots. Zviel-Girshin et al. [46] stated that young learners, especially girls, treat robots as friends and feel interested in participating in tasks [46]. Social robots have been reported to engage females’ learning participation [43]. In addition, although differences in perceptions were presented, the children found the interdisciplinary activities to be interesting and highly enjoyable, as the learning content (Stone Soup) was connected to the CT concepts in their ML classroom. From the teacher’s observation, the students all perceived the activities as valuable learning experiences.

Using these results, we can now answer our original research question: do SR-integrated interdisciplinary activities support learning? The answer is affirmative when incorporating SRs into cross-disciplinary learning, but the design needs to consider the roles that SRs play during the learning activity in order to have a gender-balanced domain. Then, regarding the question “Were there gender differences in the SR-integrated interdisciplinary activities in terms of ML knowledge, programming self-efficacy, attitudes toward SRs, and learning perceptions?”, the results indicated that no significant differences were found in the previous three dimensions. Using SRs can provide alignment strategies tailored to gender-balanced courses, which will affect girls and change their limited participation in the future field of computer science. Moreover, it is certain that there is a significant difference in children’s perceptions.

5.4. Limitations

Several limitations should be clarified. Both genders improved their learning achievement, increased their self-efficacies in programming, and had positive attitudes toward SRs. However, one must be aware of the likelihood that the findings in the present study were due to a flaw brought on by the limited sample size. Thus, it may be hard to generalise the results, since only a small sample was recruited in the current study. However, the results from the current study serve as pioneering results for gender instructors and CT practitioners. Including more participants could add more insights to the results in the future, and further examination of the influence of SR applications on learning is needed. Second, the lack of a control group to compare the gender differences in learning performance is apparent. Future studies addressing gender differences that use SR agent techniques and compare them with a non-SR (e.g., education roots) control group are recommended. The last limitation concerns technical problems. Over-enthusiastic young learners damaged the SRs’ hands when they won the game, or sometimes students occasionally over-touched the SRs’ faces that could exhibit facial expressions, causing unnecessary technical problems and distracting students from the task.

5.5. Future Research

The empirical SR findings led to two potential future research directions. One possibility would be to explore more detailed specific learning behavioural patterns of boys and girls. This would help to comprehend the techniques and approaches taken by children (boys and girls) to think computationally and approach problems analytically. Meanwhile, the second focused coding activities support immediate feedback (quickly understanding the errors and conducting debugging). It may not be easy to reveal students’ coding strategies by looking into qualitative gender data in the CT studies, as real-time feedback can easily fall into a trial-and-error loop if there is no proper intervention [18]. The second intriguing idea is to compare how different cross-institutions and cross-disciplines affect gender performance. The gender characteristics demonstrated in the same context (e.g., Asian culture) would differ from the context in Western areas [6]. For example, we plan to include a control group to analyse the outcomes of other courses (e.g., EFL) using a similar approach to see if there are any differences.

5.6. Implications

With growing interest on a global scale in reinforcing women’s participation in computer science, Goal 5.B of the United Nations (UN) Sustainable Development Goals (SDGs) prioritises the empowerment of all women and girls through information and communications technology (ICT) in order to achieve gender equality. Understanding this initiative is feasible and applicable in light of our findings offering robust support for the integration of ERs in the cross-disciplinary curriculum of young-learner CT education. In essence, SRs, one of the ER devices, can be used as a powerful technology to reform traditional stereotypes of boys being superior to girls at coding, robotics, and computing. Krommer [47] and Byrd [48] addressed a similar view that activity design should accommodate both genders, offering space and movement for boys when involving them in language production and arranging girls to work cooperatively and to express themselves while promoting their self-efficacy and confidence. Girls can be taught about CT beyond STEM to integrate target-language learning and CT skills. They are needed for reaching Goal 5.B of the UN Sustainable Development Goals for the empowerment of all women and girls to access technologies [49] and are in much demand for becoming 21st century digital citizens ([50,51]).

6. Conclusions

Only a few studies have specifically addressed gender-relevant SRs for interdisciplinary activities. Concerning the fact that a constructionism-based tangible device with three focused activities is one of the promising instructional strategies, SR-integrated gender-balanced learning activities for interdisciplinary activities which align CT and ML should not be ignored. This study examined the feasibility of SR-integrated focused activities and critically assessed their impact on gender-balanced interdisciplinary learning in an elementary context of education. It concludes that pedagogy-supported interdisciplinary activities with CT integration in a gender-based manner were helpful for promoting students’ learning in terms of their Mandarin linguistic knowledge, self-efficacies in programming, positive attitudes toward the use of SRs, and positive perceptions of the SR-integrated activities. The findings help expand the literature on the design of SRs in a gender-balanced way for interdisciplinary activities.

It is ascertained that SR-integrated cross-disciplinary programming education is incredibly beneficial and will be of great significance in the future. Drawing on the findings presented above, the low ratios of women in technologies can be reformed at the start of elementary education by integrating CT with pedagogy-informed gender-based interdisciplinary instruction. The use of SRs has actually balanced this gender gap. The number of women interested in coding and technology is expected to increase if a similar instructional design can be applied in other contexts, cross-subjects, or institutions.