Effects of voice assistant creation using different learning approaches on performance of computational thinking

Ting Chia Hsu; Ching Chang; Yi Wei Lin

doi:10.1016/j.compedu.2022.104657

Effects of voice assistant creation using different learning approaches on performance of computational thinking

Ting Chia Hsu, Ching Chang^*, Yi Wei Lin

^*Corresponding author for this work

Department of Technology Application and Human Resource Development

Research output: Contribution to journal › Article › peer-review

1 Citation (Scopus)

Abstract

Designing artificial intelligence (AI) artefact learning has gone beyond command-line-based instruction, to include a low-barrier threshold with block-based programming. Such instructional design must not solely emphasise AI workings. Rather, it must offer students computational thinking (CT) practice to support their AI-related artefact creation while reducing their AI anxiety about future job replacement or sociotechnical blindness. In this study, this research explored an experiential learning approach to improve CT along with AI application capabilities when engaging undergraduate students in creating a voice assistant application (VA app). A total of 56 students participated in the study. The control group (CG) of 26 students used a conventional subject-based learning method, while the experimental group (EG) of 30 students adopted an experiential learning method. This study aimed to examine the differences in the learning achievement of CT and AI concept, as well as the perspectives of AI anxiety, and CT; in the meanwhile, this study analysed the students' learning behaviours using sequential behavioural analysis to discuss the learning process. Results showed that the CT ability of the EG was better than that of the CG, although no significant difference was found between the two groups’ AI concepts and anxiety. The behaviour analysis also revealed that the EG students were willing to ask more questions, and conducted their VA evaluation, whereas the CG students were inclined to focus on the input and output of knowledge, and replicated what the teacher presented. Suggestions and implications are given for future research.

Original language	English
Article number	104657
Journal	Computers and Education
Volume	192
DOIs	https://doi.org/10.1016/j.compedu.2022.104657
Publication status	Published - 2023 Jan

Keywords

Artificial intelligence
Computational thinking
Experiential learning
Student learning behaviour
Voice assistant

ASJC Scopus subject areas

Computer Science(all)
Education

Access to Document

10.1016/j.compedu.2022.104657

Cite this

@article{ef368102ba3849069f1b3d483fd70197,

title = "Effects of voice assistant creation using different learning approaches on performance of computational thinking",

abstract = "Designing artificial intelligence (AI) artefact learning has gone beyond command-line-based instruction, to include a low-barrier threshold with block-based programming. Such instructional design must not solely emphasise AI workings. Rather, it must offer students computational thinking (CT) practice to support their AI-related artefact creation while reducing their AI anxiety about future job replacement or sociotechnical blindness. In this study, this research explored an experiential learning approach to improve CT along with AI application capabilities when engaging undergraduate students in creating a voice assistant application (VA app). A total of 56 students participated in the study. The control group (CG) of 26 students used a conventional subject-based learning method, while the experimental group (EG) of 30 students adopted an experiential learning method. This study aimed to examine the differences in the learning achievement of CT and AI concept, as well as the perspectives of AI anxiety, and CT; in the meanwhile, this study analysed the students' learning behaviours using sequential behavioural analysis to discuss the learning process. Results showed that the CT ability of the EG was better than that of the CG, although no significant difference was found between the two groups{\textquoteright} AI concepts and anxiety. The behaviour analysis also revealed that the EG students were willing to ask more questions, and conducted their VA evaluation, whereas the CG students were inclined to focus on the input and output of knowledge, and replicated what the teacher presented. Suggestions and implications are given for future research.",

keywords = "Artificial intelligence, Computational thinking, Experiential learning, Student learning behaviour, Voice assistant",

author = "Hsu, {Ting Chia} and Ching Chang and Lin, {Yi Wei}",

note = "Funding Information: The premise of conducting VA learning is a type of hands-on activity construction that affords students opportunities to gain knowledge of AI and CT after manipulating hands-on projects (Lin et al., 2020). This activity construction needs to support individual learning paces by providing scaffolding that helps them keep up with the teaching progress, and facilitates their VA problem-solving in many ways when problems arise (e.g., validating VA models). This complies with Van Brummelen et al.{\textquoteright}s (2021a) suggestion to be careful to support those falling behind in hands-on VA projects so they can keep up with the pace.A video tutorial is a feasible way to facilitate lessons for hands-on project information delivery, as suggested by Rodr{\'i}guez-Garc{\'i}a et al. (2019), as it can provide students with clear guidance and can support them to decompose materials into manageable units, while pointing out potential mistakes or possibly debugging when conducting troubleshooting (Rodr{\'i}guez-Garc{\'i}a et al., 2020). Specifically, the process of building VA apps involving training, testing, and making (Van Brummelen et al., 2019), supported by video tutorials not only facilitates CT development, such as analytical thinking, logical reasoning, and abstraction comprehension, but also reinforces students' reflection on their hands-on experiences of VA creation such as evaluating VA's prediction and generation (Rodr{\'i}guez-Garc{\'i}a et al., 2019). Participants not only perform the VA coding development using EL, but also flexibly examine the video tutorials during in-class instruction.As it is necessary to work through certain phases when doing VA app creation, it may not be easy to unpack the learning process without actually analysing students' learning behaviours. The VA learning content supported by video tutorials potentially plays a role in analysing students{\textquoteright} learning strategies, as simply playing videos forward or backwards can explicitly reveal how they locate specific solutions or process information delivery via their actions when they are involved in problem-solving and AI-creation tasks (e.g., Hsu et al., 2021). Particularly, as these two groups involved in different learning approaches might present different learning behaviours; how high-achieving students (divided by pretest) show a better strategy to search for intended information or reduce uncertain information using fast forward when compared to the low-achieving students. Briefly, investigating how two groups gain video review support while performing AI-artefact learning behaviours should be addressed.Apart from investigating learners' learning behaviours with video tutorial support, making students aware of the impact of AI on daily life is important. Thus, it is essential to report that whether such a lowering barrier of entry would stimulate their interest in acquiring AI knowledge and skills while lowering students{\textquoteright} AI anxiety toward future jobs replacement, sociotechnical blindness, and relevant AI configurations (e.g., relevant AI products or advanced AI skills) (Wang & Wang, 2022). As a high degree of AI anxiety would negatively inhibit performance with no intention of AI adoption. Concisely, we investigated the impacts of VA learning mediated by AI and CT on these two groups of CG and EG students, and examined their AI anxiety and their learning behaviours while being supported by video tutorials in the proposed activity. Results can thus contribute to understanding AI anxiety in AT-artefact VA app learning through CT construction, and offer potential teaching implications and suggestions to support instructors in their AI curriculum design.In this study, a series of phases in VA creation not only allows students to potentially accumulate CT skills; students could also acquire AI by imposing algorithms and datasets, validating and testing the results. In addition to articulating AI artefacts and skills mediated by CT development with a low-barrier design, it offers clear guidance to conduct practical VA supported by video tutorials. When students successfully unpack VA workings, students' AI anxiety, in term of fears about job replacement, sociotechnical blindness, and AI configurations (e.g., products or skills), can be reduced. Although the goal in educational settings is to support students to become AI innovators in different scenarios after creating AI-related artefacts, it is also important to be aware of students{\textquoteright} potential AI anxiety in the instructional design.While AI development is increasingly present in our daily lives, guiding students to recognise that what we see, hear, or learn is often mediated by simple algorithms has become important. Indeed, building AI-related artefacts is not such a complex project that requires command-line-based code. Rather, they can be created by applying simple CT concepts mediated by algorithms using block-based programming. As demonstrated by the current study, there is a low entry threshold to integrate CT into the AI-related VA artefacts tasks. Coupled with existing studies (e.g., Chen et al., 2021; Hsu et al., 2021; Zhou et al., 2020) that have successfully evidenced learning effects using low-barrier-to-entry AI practical hands-on tasks, critically designing activities with pedagogical-informed approaches and examining their results are under investigation. More studies can be done that consider possible alternative thresholds with low barriers with instructional support for students to engage in AI works and access CT skills, since simply putting components (e.g., AI, CT & VA) together cannot guarantee the expected outcomes.Such simple replication of what is shown by the teacher may have fostered an illusion of understanding among CG students as it was found from the analysis of the video viewing behaviours that they were highly engaged in rewinding and fast rewind behaviours. The low-achieving students in the CG frequently went fast backwards and reviewed the content, demonstrating their struggles in conducting VA validation, and their eagerness to figure out the solutions, whereas their counterparts in the EG did not. Meanwhile, it was common for the high-achieving students to navigate forward and to fast forward content, as they are assumed to be more likely to ignore previously learnt information (List & Ballenger, 2019). The retrieved behaviour patterns of the EG's high-achieving students were consistent with patterns of high groups demonstrated in List and Ballenger's (2019) study. They used scan forward and fast forward to work on AI evaluation by reducing the previously learnt information they had acquired, since they were situated in EL phases to support their cognitive knowledge (e.g., algorithm decoding and code construction) based on a framework of AI construction (Van Brummelen et al., 2019). However, their counterparts in the CG did not show similar patterns as they did not have fast-forward behaviours, showing that the high group in the CG still followed the linear structure of the video content to keep up with the instructional pace. This finding echoes Hsu et al.{\textquoteright}s (2021) discussion of the limitation of traditional hands-on tasks to actively construct knowledge. Overall, compared to the EG students, the CG students located the information by frequently going fast backwards to find their intended solutions (e.g., algorithm construction) in their AI practice (validating and testing). Although a lower barrier of entry to AI was tailored, a pedagogically supported approach (e.g., EL) in the instructional design to support learning should be taken into account.Although the behaviour analysis provided intriguing insights into the VA learning students engaged in during hands-on tasks, it was the CG's powerful reviewing behaviours supported by video tutorials that helped them to keep up with the progress. Hsin and Cigas's (2013) study embedded videos to scaffold students' learning during their participation in online course activities, and met the goal of aligning teaching activities with cognitive development, while Rodr{\'i}guez-Garc{\'i}a et al. (2019) provided students with video tutorials to successfully support primary-school students with little prerequisite knowledge to learn CT underlying modern AI-related problems. The embedded videos in the current study also triggered students' reflective thinking about their ideas of VA app practice, testing their ideas, and conceptualising creation, which may explain the first RQ regarding why both groups successfully completed the tasks without showing any significant difference in quantitative analysis. Publisher Copyright: {\textcopyright} 2022 Elsevier Ltd",

year = "2023",

month = jan,

doi = "10.1016/j.compedu.2022.104657",

language = "English",

volume = "192",

journal = "Computers and Education",

issn = "0360-1315",

publisher = "Elsevier Limited",

}

TY - JOUR

T1 - Effects of voice assistant creation using different learning approaches on performance of computational thinking

AU - Hsu, Ting Chia

AU - Chang, Ching

AU - Lin, Yi Wei

N1 - Funding Information: The premise of conducting VA learning is a type of hands-on activity construction that affords students opportunities to gain knowledge of AI and CT after manipulating hands-on projects (Lin et al., 2020). This activity construction needs to support individual learning paces by providing scaffolding that helps them keep up with the teaching progress, and facilitates their VA problem-solving in many ways when problems arise (e.g., validating VA models). This complies with Van Brummelen et al.’s (2021a) suggestion to be careful to support those falling behind in hands-on VA projects so they can keep up with the pace.A video tutorial is a feasible way to facilitate lessons for hands-on project information delivery, as suggested by Rodríguez-García et al. (2019), as it can provide students with clear guidance and can support them to decompose materials into manageable units, while pointing out potential mistakes or possibly debugging when conducting troubleshooting (Rodríguez-García et al., 2020). Specifically, the process of building VA apps involving training, testing, and making (Van Brummelen et al., 2019), supported by video tutorials not only facilitates CT development, such as analytical thinking, logical reasoning, and abstraction comprehension, but also reinforces students' reflection on their hands-on experiences of VA creation such as evaluating VA's prediction and generation (Rodríguez-García et al., 2019). Participants not only perform the VA coding development using EL, but also flexibly examine the video tutorials during in-class instruction.As it is necessary to work through certain phases when doing VA app creation, it may not be easy to unpack the learning process without actually analysing students' learning behaviours. The VA learning content supported by video tutorials potentially plays a role in analysing students’ learning strategies, as simply playing videos forward or backwards can explicitly reveal how they locate specific solutions or process information delivery via their actions when they are involved in problem-solving and AI-creation tasks (e.g., Hsu et al., 2021). Particularly, as these two groups involved in different learning approaches might present different learning behaviours; how high-achieving students (divided by pretest) show a better strategy to search for intended information or reduce uncertain information using fast forward when compared to the low-achieving students. Briefly, investigating how two groups gain video review support while performing AI-artefact learning behaviours should be addressed.Apart from investigating learners' learning behaviours with video tutorial support, making students aware of the impact of AI on daily life is important. Thus, it is essential to report that whether such a lowering barrier of entry would stimulate their interest in acquiring AI knowledge and skills while lowering students’ AI anxiety toward future jobs replacement, sociotechnical blindness, and relevant AI configurations (e.g., relevant AI products or advanced AI skills) (Wang & Wang, 2022). As a high degree of AI anxiety would negatively inhibit performance with no intention of AI adoption. Concisely, we investigated the impacts of VA learning mediated by AI and CT on these two groups of CG and EG students, and examined their AI anxiety and their learning behaviours while being supported by video tutorials in the proposed activity. Results can thus contribute to understanding AI anxiety in AT-artefact VA app learning through CT construction, and offer potential teaching implications and suggestions to support instructors in their AI curriculum design.In this study, a series of phases in VA creation not only allows students to potentially accumulate CT skills; students could also acquire AI by imposing algorithms and datasets, validating and testing the results. In addition to articulating AI artefacts and skills mediated by CT development with a low-barrier design, it offers clear guidance to conduct practical VA supported by video tutorials. When students successfully unpack VA workings, students' AI anxiety, in term of fears about job replacement, sociotechnical blindness, and AI configurations (e.g., products or skills), can be reduced. Although the goal in educational settings is to support students to become AI innovators in different scenarios after creating AI-related artefacts, it is also important to be aware of students’ potential AI anxiety in the instructional design.While AI development is increasingly present in our daily lives, guiding students to recognise that what we see, hear, or learn is often mediated by simple algorithms has become important. Indeed, building AI-related artefacts is not such a complex project that requires command-line-based code. Rather, they can be created by applying simple CT concepts mediated by algorithms using block-based programming. As demonstrated by the current study, there is a low entry threshold to integrate CT into the AI-related VA artefacts tasks. Coupled with existing studies (e.g., Chen et al., 2021; Hsu et al., 2021; Zhou et al., 2020) that have successfully evidenced learning effects using low-barrier-to-entry AI practical hands-on tasks, critically designing activities with pedagogical-informed approaches and examining their results are under investigation. More studies can be done that consider possible alternative thresholds with low barriers with instructional support for students to engage in AI works and access CT skills, since simply putting components (e.g., AI, CT & VA) together cannot guarantee the expected outcomes.Such simple replication of what is shown by the teacher may have fostered an illusion of understanding among CG students as it was found from the analysis of the video viewing behaviours that they were highly engaged in rewinding and fast rewind behaviours. The low-achieving students in the CG frequently went fast backwards and reviewed the content, demonstrating their struggles in conducting VA validation, and their eagerness to figure out the solutions, whereas their counterparts in the EG did not. Meanwhile, it was common for the high-achieving students to navigate forward and to fast forward content, as they are assumed to be more likely to ignore previously learnt information (List & Ballenger, 2019). The retrieved behaviour patterns of the EG's high-achieving students were consistent with patterns of high groups demonstrated in List and Ballenger's (2019) study. They used scan forward and fast forward to work on AI evaluation by reducing the previously learnt information they had acquired, since they were situated in EL phases to support their cognitive knowledge (e.g., algorithm decoding and code construction) based on a framework of AI construction (Van Brummelen et al., 2019). However, their counterparts in the CG did not show similar patterns as they did not have fast-forward behaviours, showing that the high group in the CG still followed the linear structure of the video content to keep up with the instructional pace. This finding echoes Hsu et al.’s (2021) discussion of the limitation of traditional hands-on tasks to actively construct knowledge. Overall, compared to the EG students, the CG students located the information by frequently going fast backwards to find their intended solutions (e.g., algorithm construction) in their AI practice (validating and testing). Although a lower barrier of entry to AI was tailored, a pedagogically supported approach (e.g., EL) in the instructional design to support learning should be taken into account.Although the behaviour analysis provided intriguing insights into the VA learning students engaged in during hands-on tasks, it was the CG's powerful reviewing behaviours supported by video tutorials that helped them to keep up with the progress. Hsin and Cigas's (2013) study embedded videos to scaffold students' learning during their participation in online course activities, and met the goal of aligning teaching activities with cognitive development, while Rodríguez-García et al. (2019) provided students with video tutorials to successfully support primary-school students with little prerequisite knowledge to learn CT underlying modern AI-related problems. The embedded videos in the current study also triggered students' reflective thinking about their ideas of VA app practice, testing their ideas, and conceptualising creation, which may explain the first RQ regarding why both groups successfully completed the tasks without showing any significant difference in quantitative analysis. Publisher Copyright: © 2022 Elsevier Ltd

PY - 2023/1

Y1 - 2023/1

N2 - Designing artificial intelligence (AI) artefact learning has gone beyond command-line-based instruction, to include a low-barrier threshold with block-based programming. Such instructional design must not solely emphasise AI workings. Rather, it must offer students computational thinking (CT) practice to support their AI-related artefact creation while reducing their AI anxiety about future job replacement or sociotechnical blindness. In this study, this research explored an experiential learning approach to improve CT along with AI application capabilities when engaging undergraduate students in creating a voice assistant application (VA app). A total of 56 students participated in the study. The control group (CG) of 26 students used a conventional subject-based learning method, while the experimental group (EG) of 30 students adopted an experiential learning method. This study aimed to examine the differences in the learning achievement of CT and AI concept, as well as the perspectives of AI anxiety, and CT; in the meanwhile, this study analysed the students' learning behaviours using sequential behavioural analysis to discuss the learning process. Results showed that the CT ability of the EG was better than that of the CG, although no significant difference was found between the two groups’ AI concepts and anxiety. The behaviour analysis also revealed that the EG students were willing to ask more questions, and conducted their VA evaluation, whereas the CG students were inclined to focus on the input and output of knowledge, and replicated what the teacher presented. Suggestions and implications are given for future research.

AB - Designing artificial intelligence (AI) artefact learning has gone beyond command-line-based instruction, to include a low-barrier threshold with block-based programming. Such instructional design must not solely emphasise AI workings. Rather, it must offer students computational thinking (CT) practice to support their AI-related artefact creation while reducing their AI anxiety about future job replacement or sociotechnical blindness. In this study, this research explored an experiential learning approach to improve CT along with AI application capabilities when engaging undergraduate students in creating a voice assistant application (VA app). A total of 56 students participated in the study. The control group (CG) of 26 students used a conventional subject-based learning method, while the experimental group (EG) of 30 students adopted an experiential learning method. This study aimed to examine the differences in the learning achievement of CT and AI concept, as well as the perspectives of AI anxiety, and CT; in the meanwhile, this study analysed the students' learning behaviours using sequential behavioural analysis to discuss the learning process. Results showed that the CT ability of the EG was better than that of the CG, although no significant difference was found between the two groups’ AI concepts and anxiety. The behaviour analysis also revealed that the EG students were willing to ask more questions, and conducted their VA evaluation, whereas the CG students were inclined to focus on the input and output of knowledge, and replicated what the teacher presented. Suggestions and implications are given for future research.

KW - Artificial intelligence

KW - Computational thinking

KW - Experiential learning

KW - Student learning behaviour

KW - Voice assistant

UR - http://www.scopus.com/inward/record.url?scp=85143773414&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85143773414&partnerID=8YFLogxK

U2 - 10.1016/j.compedu.2022.104657

DO - 10.1016/j.compedu.2022.104657

M3 - Article

AN - SCOPUS:85143773414

SN - 0360-1315

VL - 192

JO - Computers and Education

JF - Computers and Education

M1 - 104657

ER -

Effects of voice assistant creation using different learning approaches on performance of computational thinking

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this