Due to the continuous advancement of technologies and techniques, people are increasingly interested in emerging technologies. This term refers to new technologies that are characterized by radical novelty, relatively fast growth, coherence, prominent impact, uncertainty, and ambiguity (Rotolo et al., 2015). Currently, several emerging technologies affect people’s lives, such as artificial intelligence (AI; Haenlein and Kaplan, 2019), 5G/6G communication networks (Sheth et al., 2020), and quantum computers (Alvarez-Rodriguez et al., 2018). To provide students with an early understanding of the impact of these technologies on people’s future lives, popular science activities are slowly being introduced into the learning activities of elementary schools. Currently, one of the most commonly discussed issues is AI education (Kim and Park, 2017; Lin et al., 2021).

Among many issues involved in emerging technologies, the topic of AI has had a strong impact on life in recent years. The term “AI” first appeared at a 1956 Dartmouth conference (Bell, 2021). There are many definitions of AI, but the term is generally known to the public as a technology that simulates human-like intelligence through computer programs (Russell and Norvig, 2016). In addition to active research in AI and the expansion of relevant industries in recent years, many scholars have been exploring the application of AI in education (AIED). For example, Baker (2000) suggested that in the near future, AI applications in education will include models as scientific tools, models as components of educational artifacts, and models as bases for the design of educational artifacts. Colchester et al. (2017) also analyzed the current status of the use of learning systems through AI and noted that AI can assist in adapting learning styles. After analyzing learners’ learning models, AI can design appropriate methods of teaching and allow learners to study by themselves.

For elementary school students, the use of AI seems difficult. However, this perception has slowly changed. The content of AI education emphasizes AI popular science education rather than AI technology (Geraci, 2011; Zhang et al., 2021). Since AI is an important issue that affects current and future life, the goal of teaching it is not for students to take exams but to nurture students’ awareness of the influence of AI on life and improve their logical thinking and problem-solving (Lin et al., 2021). Therefore, the promotion of AI education from a young age is a matter of great urgency. To date, however, AI-supported emerging technologies cannot be integrated into the formal curriculum and can only be taught in informal curricula (i.e., popular science activities).

As human life continues to evolve and prosper at a rapid pace, scientific literacy has become a necessary capacity for society at large. According to the National Science Teachers Association (1982), a scientifically literate citizen must (1) understand the interaction between science, technology, and society and actively inquire and solve related problems; (2) have the ability to comprehend and apply scientific concepts, theories, and facts; (3) understand the essential spirit of science and make positive and meaningful judgments about social issues; (4) apply scientific methods, concepts and correct scientific attitudes to solve everyday problems; and (5) have a deeper understanding and more active involvement in the relationship between science, technology, and the environment. From the above description of scientific literacy, it is apparent not only that scientific literacy is required for professionals but also that every citizen needs information on this knowledge and related concepts and attitudes.

The goal of scientific development has gradually shifted from elite education to the cultivation of scientific literacy for all. Therefore, popular science activities act as an important channel to promote education for higher scientific literacy among all people. According to Dzan et al. (2015), popular science activities can increase students’ willingness to learn science by allowing them to engage in fun and interesting activities and broadening their knowledge.

Since AI affects our lives, it is important for AI education to enhance students’ awareness and knowledge of AI through interesting activities from a young age to promote universal scientific literacy (Hsu et al., 2022). However, because popular science activities are not part of the formal curriculum, they are often conducted as experiential events and as part of non-standard courses. Generally, experiential activities are scheduled for approximately 20–60 min, while informal curricula may be scheduled for several hours.

Informal curricula refer to educational activities that take place in off-campus settings (Jeffs and Smith, 2021). Informal curricula can be implemented in a variety of ways, including a series of lessons, seminars, short courses, and experiential activities. Popular science activities are often conducted in these common formats to disseminate scientific knowledge. In addition to other types of activities, popular science activities are often designed using theories, such as hands-on learning (Dewey, 1997), experiential learning (Kolb, 1981), or STEM learning frameworks (Kelley and Knowles, 2016).

The purpose of designing curricula based on relevant theories is to elicit motivation and learning outcomes in science subjects via popular science activities. Nugent et al. (2010) and Anand and Dogan (2021) note that the purpose of learning activities in informal curricula is to provide students with opportunities to engage more deeply in learning activities. Lin et al. (2015) arranged popular science activities through the Nanotechnology-based Popular Science Education Promotion and Teaching (NPSEPT) program. The results showed that the participants, elementary school students, achieved significant gains in nanotechnology learning performance and outcomes. The effectiveness of these popular science activities was attributed to the curriculum design and students’ willingness to learn. The National Research Council (2000) emphasized in the “schema for learning sciences” that when students identify themselves as science learners in an informal curriculum, this facilitates their learning of science.

Although some of the above studies have confirmed the effectiveness of popular science activities for enhancing science learning, some controversies remain. Currently, some popular science activities are promoted in the form of short courses. Because of time constraints, instructors are unable to fully impart science knowledge and guide inquiry and hands-on activities (Hight et al., 2021; Shu et al., 2021). Therefore, there is a paucity of information on whether popular science activities of short duration (a few hours) can help students achieve learning outcomes and motivate them for science studies.

On the other hand, previous studies have shown that popular science activities are effective in increasing students’ interest in and motivation for learning (Nugent et al., 2010; Lin et al., 2015; Anand and Dogan, 2021). However, popular AI science activities require high cognitive abilities, and students need to spend time writing programs (i.e., using computational thinking skills) in addition to assembling simple building block models (e.g., assembling machines). This study also investigated the effect of short hands-on AI activities on students’ interest in learning. Therefore, the first question of this study concerns the impact of short-term popular science activities in informal curricula on students’ AI achievement tests.

Teacher support includes both academic and emotional support (Patrick et al., 2007; Liu et al., 2018). Teacher academic support refers to students’ perception that teachers care about what and how much they learn, while teacher emotional support reflects students’ awareness that teachers care about students as individuals (Johnson et al., 1985). Many studies have confirmed the positive impact of teacher support on student learning. For example, Yu and Singh (2018) indicated that teacher support indirectly affects students’ mathematics achievement through their sense of mathematics self-efficacy and affects students’ interest in mathematics courses. Fredricks et al. (2018) examined how motivation and background influence boys’ and girls’ engagement in mathematics and science subjects. The results indicated that motivation and background factors were significantly associated with students’ engagement level, with girls being more likely to participate in subject activities because of teacher support. Li et al. (2021) also noted that teacher support of literacy development can improve students’ ability to build scientific models.

From the above literature, it is clear that teacher support is helpful to students (Marec et al., 2021). In the case of popular science activities, it is worthwhile to explore the ways in which teacher support is incorporated into students’ research practices. During such activities, because of the limited time, students often opt for teacher support when they encounter problems. However, in the case of AI hands-on activities, does the teacher’s guidance of students in programming limit students’ ability to try to develop computational thinking on their own? Is it the right choice to give direct answers to students when learning about AI? According to the above literature, the benefits of teacher support mostly appear in terms of motivation, self-efficacy, and interest in learning; in regard to cognitive learning outcomes, teacher support serves only to improve student learning overall. However, there is insufficient research on the impact of teacher support on students’ thinking patterns (e.g., creativity and logical thinking). Current popular science activities, as informal curricula, are intended not only to enhance students’ interest in learning science but also to help their thinking patterns evolve. Teacher support is expected to help students learn more effectively (Sadoughi and Hejazi, 2021), but there is a possibility that students may not be able to think and investigate independently in certain areas because of their reliance on teachers (e.g., Ma et al., 2021). Therefore, the second question pursued in this study concerns the relationship between the presence or absence of teacher support and students’ learning outcomes.

In summary, there are two main issues involved in conducting AI-related popular science activities in informal curricula. The first issue is whether short popular science activities only spur motivation and interest in learning or whether they also provide additional enhancement to learning outcomes. The second issue is that while teacher support in science education has been recognized by many scholars as contributing to students’ learning results, the role of teacher support in popular science activities needs to be further explored. The following are the research questions for this study:

The participants of this study were elementary school students in grades 5–6 who applied to participate voluntarily. All of the participants had basic knowledge about a visual programming language, for example, Scratch. A total of 22 students, including 16 boys and six girls, participated in the program; they were divided into 11 random pairs during the activity.

To investigate the two research questions, a 6-h AI education activity was conducted in 1 day. This activity was designed based on the STEM learning conceptual framework (Kelley and Knowles, 2016) and project-based learning (Triana et al., 2020) with both teaching sessions and hands-on activities. AI achievement pre- and posttests were arranged. In addition to lectures and hands-on exercises, group problem-solving activities were arranged (Figure 1). The schedule of activities included (a) Micro:bit and makecode teaching, (b) introduction to AI and the visualization of AI chips, (c) designing and solving problems in real-life scenarios (with teacher support), (d) designing and solving problems in real-life scenarios (without teacher support), and (e) result sharing.

In the activities on problem design and problem-solving in real-life scenarios, one activity was conducted with teacher support. In this activity, students could ask questions and the teacher provided prompts; if students did not understand or could not perform the task, the teacher provided answers. The second activity was conducted without teacher support, meaning that students were not allowed to ask any questions to the teacher or other groups during the activity (Figure 2).

Regarding the practices of designing and solving problems in real-life scenarios, the questions included face recognition (without teacher support) and color recognition (with teacher support), and the contexts were established as follows.

For the two research questions, the data analysis was divided into two parts. The first part focused on exploring the effectiveness of students’ participation in popular AI science activities. Thus, all the participants were asked to respond to the AI achievement pre- and posttest, and then the dependent t-test was conducted to analyze students’ performance growth.

The second part focused on comparing the activities with and without teacher support and examining the interrelationship among students’ learning outcomes in AI popular science activities (i.e., creativity, computational thinking, problem-solving, and the completeness of finished work). Thus, the AI artifacts presented by the group (including face recognition and color recognition) were analyzed with the developed scoring rubric. Two teachers in a related field in elementary school evaluated the work according to the rubric. The interrater reliability was 0.80. The average of the two teachers’ evaluations was used as the final grade.

The dependent t-test was conducted to understand student AI performance growth after joining the AI activities (Table 3). There is a significant difference in the overall learning effectiveness in terms of knowledge and skills in AI (t = −3.55, p = 0.002). Furthermore, a closer look at individual items reveals major differences among the dimensions: machine learning history (t = −3.81, p = 0.001) and AI concepts (t = −2.61, p = 0.016) both demonstrate improvement, while there are no significant gains for coding (t = 1.28, p = 0.213).

It can be concluded that students’ learning of emerging AI technologies can be enhanced through 6 h of popular AI science activities combined with lectures and hands-on activities in groups. However, coding does not show such major improvement, probably because coding requires longer periods of practice and development and the short hands-on activities cannot improve students’ learning effectiveness in coding skills.

With regard to teacher support, the literature review showed that students’ learning outcomes are better in learning activities with teacher support (Fredricks et al., 2018; Yu and Singh, 2018; Li et al., 2021). Table 4 shows that the scores of activities with guidance from the teacher are higher than those with no teacher support. Next, the role of teacher support in learning the emerging technologies of AI was explored.

The comparison of activities with and without teacher support shows that when students are provided with teacher assistance, work completion (r = 0.27), problem-solving ability (r = 0.35), and creativity (r = 0.00) are not significantly correlated with the algorithmic thinking skills (Table 5), whereas when students are not provided with teacher support (Table 6) and need to work alone in small groups, completion (r = 0.93), problem-solving (r = 0.97), and creativity (r = 0.87) are all highly correlated with algorithmic thinking skills.

The above analysis indicates that teacher support is related to better learning outcomes in AI popular science activities. However, an interesting finding is that students’ algorithmic thinking performance is not significantly correlated with creativity, problem-solving ability, or work completion in the presence of teacher support. The researchers theorize that when students encounter problems and teacher support is available, they become accustomed to seeking help from teachers rather than leveraging their individual algorithmic thinking ability.

On the other hand, when teacher support is not available, students’ algorithmic thinking performance is highly correlated with their creativity, problem-solving ability, and work completion on AI projects. This means that without teacher support, students must rely on their individual algorithmic thinking to solve problems. Furthermore, students’ creative performance and work completion are highly correlated with algorithmic thinking.

From these two studies, it is clear that hours long AI popular science activities are beneficial to students’ cognitive and skill acquisition. However, excessive support from teachers in the curriculum can limit students’ opportunities to practice their own algorithmic thinking. In other words, the design of future curricula needs to consider the level of teacher support and understand its impact on refining students’ high-level algorithmic thinking. Therefore, although popular science activities are conducted in a relatively short period of time, they can enhance students’ learning outcomes if they are designed appropriately. Second, learning activities with teacher support can enhance students’ learning outcomes, but they allow students to become accustomed to relying on their teachers and not think independently about the activities. In contrast, popular science activities without teacher support seem to be more effective in fostering students’ independent computational thinking and problem-solving abilities because they need to be completed independently.