Sage Journals: Discover world-class research

Abstract

Against the backdrop of the deep integration of chatbots into education, this study, based on Self-Determination Theory (SDT) and the UTAUT model, constructed a model of factors influencing college students’ self-directed learning ability in programming. Through a review of existing literature, six key determinants were identified: learning expectancy, effort expectancy, facilitation conditions, teaching interaction, learning motivation, and learning self-efficacy. The main goal of this study was to develop a model that outlines the interaction of these factors when students use chatbots to learn programming. To this end, a questionnaire survey was conducted among 296 college students who had used chatbots for programming self-study. The questionnaire consisted of a set of rigorously validated items. The collected data were carefully analysed, including the construction and validation of a structural equation model using AMOS software. The results show that learning motivation plays the most important role in the positive influence of chatbots on programming self-learning ability. Its effect was significantly greater than that of other factors, including teaching interaction, learning self-efficacy, learning expectancy, and effort expectancy. In addition, learning motivation was found to mediate the effect of learning self-efficacy on programming self-learning ability. Finally, based on the findings, the study provides a series of recommendations aimed at offering scientific and practical strategies to enhance college students’ ability to learn programming independently.

Plain language summary

As programming becomes an essential skill for college students, chatbots are increasingly used to support learning. Researchers surveyed 296 students who used chatbots for self-study. The analysis focused on six key factors: learning expectancy, effort expectancy, facilitation conditions, teaching interaction, learning motivation, and learning self-efficacy. The results show that learning motivation plays the most important role in the positive influence of chatbots on programming self-learning ability. Its effect was significantly greater than that of other factors, including teaching interaction, learning self-efficacy, learning expectancy, and effort expectancy. In addition, learning motivation was found to mediate the effect of learning self-efficacy on programming self-learning ability. This study provides useful suggestions for using chatbots to better support students in learning programming independently.

Keywords

chatbot programming self-learning ability self-determination theory UTAUT

Introduction

In recent years, with the development of information technology, artificial intelligence (AI) has been increasingly applied across various fields, such as economy, medicine, agriculture and education. Given the close connection between education and technology, the education has also undergone innovative transformations under the influence of AI development. Clarizia et al. (2018) consider chatbots as an innovative solution to bridge the gap between technology and education. Chatbots can easily detect learners’ gaps and respond accordingly, creating meaningful interactions that significantly benefit learners’ performance and memory (Deng & Yu, 2023). Meanwhile, programming-being the foundation of AI, has also been gradually emphasized. Today, learning programming is no longer limited to computer science majors, but has gradually become a general course in many science and engineering disciplines. Programming is essentially a practical activity that generates large amounts of behavioural data during the learning process. Such data can be collected implicitly and in real time. Given their strong data processing capabilities, chatbots hold great potential for assisting programming learning. Previous studies have identified three types of chatbot response modes: pattern-based, retrieval-based, and generative models (Manaswi, 2018). This study focuses on generative chatbots, which generate appropriate responses based on past and current user context.

Current research on the teaching capabilities of chatbots mainly focuses on three areas: design and technical capabilities (Kohnke, 2023; Lin & Mubarok, 2021), support for language education, particularly English learning (Koç & Savaş, 2025), and educational support for the biomedical field. While previous studies have noted tthe positive impact of chatbots on learning programming (Clarizia et al., 2018), few have examined how chatbots influence programming self-learning ability. In today’s digital society, learners need strong autonomous learning skills to master programming languages. Therefore, based on Self-Determination Theory (SDT), this study constructs a structural equation model (SEM) to investigate the factors through which chatbots affect college students’ programming self-learning ability, thereby expanding research on chatbot applications in programming education.

The present paper is structured as follows: Section 2 provides a comprehensive literature review. Section 3 describes the variable definitions and construction of the hypothesized model. Section 4 describes the research materials and methods. Section 5 is the analysis of the model and the display of results. Section 6 is a discussion of the results and limitations of the study. Section 7 provides suggestions based on the comprehensive research results.

Review of Related Studies in the Literature

In this study, the Unified Theory of Acceptance and Use of Technology (UTAUT) is selected as the research model framework, combined with the Self-Determination Theory to explore the influence factors of chatbots on college students’ programming self-learning ability, and the related theories involved are shown as follows:

Chatbots in Education

Chatbots are software tools that interacts with users through text or voice conversations on specific topic or within specific domain (Smutny & Schreiberova, 2020). Compared with traditional e-learning systems, they serve as an interactive mechanism with broad application potential in educational settings(Kiptonui, 2013). Chatbots are capable of interacting with students in multiple roles, including as tutors, advisors, classmates, or even gaming companions (Pérez-Marín, 2021). Current research on chatbots in education mainly fall into three categories. The first category mainly explores the support of chatbots to current education, for example, Essel et al. (2022) found through a controlled experiment that students interacting with chatbots performed better than those interacting with course instructors, supporting the value of using chatbots in higher education. Deng and Yu (2023) used meta-analysis to summarize the impact of chatbots on education. They found that chatbot-based learning was more effective than traditional learning in terms of explicit reasoning, academic performance, knowledge retention, and learning interest. 5 The second category mainly explores the effects of chatbots on language learning. For example, Chien et al. (2022) found that the contextual learning environment based on LINE ChatBot significantly improved learners’ English speaking and listening skills. Kohnke (2023) argued that chatbots can be used as supplementary tools for students’ language learning. The third category mainly investigates the effects of chatbots on students’ motivation (Yin et al., 2021), learning experience (Sandu & Gide, 2019) and motivation (Chiu et al., 2023). However, there is limited research on how chatbots influence college students’ autonomous learning ability in programming. Chatbots have shown significant potential and innovation in supporting challenging subjects like programming (Essel et al., 2022), therefore, this paper constructs a structural equation model based on UTAUT and SDT to explore the factors influencing college students’ autonomous learning ability in programming with the help of chatbots.

Research on UTAUT

The Unified Theory of Acceptance and Use of Technology (UTAUT) is one of the technology acceptance models developed based on TAM (Technology Acceptance Model), TRA (Theory of Reasoned Action), DOI (Diffusion of Innovation), and TPB (Theory of Planned Behaviour; Dečman, 2015). It reflects how individuals’ experience and knowledge influence their willingness to accept information technology. UTAUT is a theoretical model used to explain and predict factors influencing users’ acceptance and use of technology (Menon & Shilpa, 2023). Building on the integration of TAM, TRA, DOI, and TPB, UTAUT introduces four fundamental constructs: performance expectancy, effort expectancy, social influence, and facilitating conditions. The UTAUT model can be applied to understand users’ acceptance and use of AI tools (Venkatesh, 2022). This study adopts three constructs from UTAUT: performance expectancy, effort expectancy, and facilitating conditions. They were used to predict college students’ acceptance of using chatbots when learning programming, and to explore how chatbots influence their autonomous learning ability in programming. Performance expectation is defined as the degree of job performance gain that individuals believe they can obtain by using the system (Venkatesh et al., 2003). In this study, performance expectation was modified to learning expectation as an observational variable, indicating that students believed that using chatbots could improve their programming learning ability. Effort expectancy was defined as the ease of using the system (Venkatesh et al., 2003), that is, the degree of effort students need to put in when learning programming using the chatbot. Facilitating conditions refer to the extent to which individuals believed that the organizational and technical infrastructure needed to support is available (Venkatesh et al., 2003), meaning the level of infrastructure and technical support students receive when learning programming with chatbots.

Research on Self-Determination Theory (SDT)

Self-determination theory (SDT), proposed by American psychologists Deci and Ryan (1985), is a motivational theory of human self-determined behaviour. It highlights the critical role of three innate psychological needs - autonomy, competence, and relatedness, which are important determinants of individual motivation. According to Deci and Ryan, autonomy is the ability of an individual to master choices and work in the most fulfilling direction, which can effectively enhance motivation. Competence refers to having the knowledge and skills necessary to complete tasks and achieve success. Relatedness describes an individual’s sense of belonging and connection to significant others or groups, which helps to maintain motivation and engagement (Nikou & Economides, 2017; Ryan & Deci, 2000a). When instructional design adequately satisfies these psychological needs, students are more likely to engage actively in learning tasks. Fostering different types of motivation serves as a key source of energy that drives students’ participation in learning activities (Chiu, 2022). SDT defines individual motivation as the degree of autonomy an individual exhibits in learning activities,which is categorized into two major motivational orientations: (1) intrinsic motivation in the form of self-determination; and (2) extrinsic motivation in the form of being under control (Gan, 2020). Intrinsic motivation is associated with positive learning outcomes, such as academic performance and achievement (Augustyniak et al., 2016), and therefore, motivation has an impact on self-learning ability.

In summary, this study selects learning expectancy, effort expectancy and facilitating conditions from UTAUT, along with learning motivation from SDT, as key variables in the research model. This helps further explore the influencing factors and the mechanism of the chatbot’s effect on college students’ programming self-learning ability.

Research Model and Hypotheses

Variable Definition

In addition to extracting some of the variables from the above theories, we also introduced two variables: teaching interaction and learning self-efficacy. Regarding teaching interaction, Moore (1989) believes that learning interaction can be divided into three types: teacher-student interaction, student-student interaction, and student-content interaction. Scholars L. Chen (2004) also believes that online learning interaction includes exchanges between learners and various material resources. This study adopts a similar view, focussing primarily on the interaction between students and chatbots. Learning self-efficacy refers to the learner’s belief or confidence in their ability to guide their own learning behaviour and achieve positive outcomes. Individuals with high self-efficacy will choose more challenging tasks, set higher goals for themselves, and persist in achieving them (Bandura, 1977). Therefore, the research variables involved in this paper are defined as shown in Table 1.

Table 1.

Definition of Study Variables.

Variable	Definition in this paper
Learning expectancy (LE)	The extent to which learners learn programming effectively through chatbots
Effort expectancy (EE)	The level of effort learners need to put in to learn programming using chatbots
Facilitating conditions (FC)	The extent to which learners are empowered with infrastructural and technical support to learn programming using chatbots
Teaching interaction (TI)	Communication and interaction between learners and chatbots in the programming learning process
Learning motivation (LM)	The degree of autonomy shown by learners in programming learning
Learning self-efficacy (LSE)	Beliefs or confidence held by learners in using chatbots to learn programming that are used to guide their learning behaviours and produce good results
Programming self-learning ability (PSA)	Learners acquire, analyse, and apply knowledge independently as they learn to programme

Model Hypotheses

In UTAUT, performance expectancy and EE are shown to be positively related to behavioural intentions (Venkatesh et al., 2003). The direct relationship between FC and behavioural intentions was confirmed in a later revision of the model by Dwivedi et al. (2019). In this study, behavioural intentions were modified to LM, which is the intrinsic driving force or psychological motivation that causes an individual’s learning behaviour. Accordingly, the following hypotheses are proposed in this paper:

H1. LE has a significant positive effect on LM.

H2. EE has a significant positive effect on LM.

H3. FC has a significant positive effect on LM.

H4. LE has a significant positive effect on college students’ PSA.

H5. EE has a significant positive effect on college students’ PSA.

H6. FC has a significant positive effect on college students’ PSA.

SDT focuses on the effect of environment on the basic psychological needs of autonomy, competence and relatedness. When these three basic psychological needs are satisfied, intrinsic motivation can be generated. Deci and Ryan (2013) pointed out that the environment of self-directed learning facilitates the generation and cultivation of intrinsic motivation. This intrinsic motivation can maintain enthusiasm, creativity and effort in learning (Augustyniak et al., 2016). Compared with intrinsic motivation, extrinsic motivation focuses more on the consequences resulting from the task than on the task itself (Dysvik & Kuvaas, 2013). In other words, learners’ behaviours are directed towards the desired outcome to be achieved (Deci & Ryan, 2000). Accordingly, the following hypotheses are proposed in this paper:

H7. LM has a significant positive effect on PSA.

In traditional classroom teaching, the teacher is the embodiment of knowledge and authority, while students passively accept and memorize. This approach often ignores students’ individual differences and diverse needs. In contrast, the application of chatbots in education highlights student-centred learning and enhances students’ learning experience (Sandu & Gide, 2019). Chatbots are able to respond naturally through the dialogue interface and have the ability to interact with users easily (Smutny & Schreiberova, 2020). This helps solve the problem of insufficient interaction between students and teachers (Clarizia et al., 2018). Moreover, high-frequency interactions can help to improve students’ understanding and enhance their LM. Accordingly, this paper proposes the following hypotheses:

H8. TI has a significant positive effect on LM.

H9. TI has a significant positive effect on college students’ PSA.

Some studies have shown that LSE is closely related to self-learning ability (Cheng, 2023). For example, Kim et al. (2022) found that self-efficacy significantly predicted learners’ temptation feelings, affective factors, and academic achievement. They based this on a sample of 123 college students studying the influence of self-efficacy on self-directed learning and learning support in online learning. Meanwhile, some studies have proved that LSE is closely related to motivation, commitment to learning, and the use of metacognitive strategies (Cheng, 2023). For instance, Chi and Xin (2006) proved that the higher the subjects’ sense of efficacy, the higher the endogenous motivation for learning. Building on this, further research has found that self-efficacy can stimulate learning motivation by enhancing the sense of achievement and the pursuit of success (Firmansyah et al., 2018). Bassi et al. (2007) demonstrated that students with a high sense of self-efficacy placed a higher emphasis on academic achievement and a higher level of desire to be educated than those with a low self-efficacy. Additionally, Zhang et al. (2021) found a significant correlation between learning engagement and LSE through empirical analyses. Learning engagement was defined by three independent dimensions, namely, behavioural, emotional, and cognitive (Fredricks et al., 2004). Accordingly, the following hypotheses were made in this paper:

H10. LSE has a significant positive effect on LM.

H11.LSE has a significant positive effect on college students’ PSA.

H12. Learning expectations has a significant positive effect on LSE.

H13. EE has a significant positive effect on LSE.

H14. FC has a significant positive effect on LSE.

H15. TI has a significant positive effect on LSE.

Some past studies have noticed that there is a certain relationship between LM and LSE and self-learning ability. For example, Cheng (2023) has shown through empirical research that self-efficacy can indirectly affect English online self-learning ability via the mediator effect of LM. We expect that the higher a person’s LSE beliefs about motivation regulation, the more effective the use of motivation regulation strategies will be (Trautner & Schwinger, 2020). Accordingly, the following hypotheses are proposed in this paper:

H16. LSE can indirectly influence PSA through the mediator effect of LM.

Model Assumptions

Based on the above theoretical assumptions, combined with the characteristics of chatbots, a theoretical research model was constructed as shown in Figure 1.

Figure 1.

Proposed model.

Materials and Methods

Data Collection

The National Undergraduate Mathematical Modelling Competition is the largest foundational subject competition among Chinese universities. Programming implementation is a critical component of this event. Therefore, we selected university students who had participated in this competition and were primarily responsible for the programming tasks as our survey respondents. Participation in this competition is often incorporated into university coursework or organized by the schools, so many students were assigned to participate rather than joining solely by personal motivation. Therefore, this group effectively represents a large segment of college students with basic programming skills and a certain level of autonomous learning ability, which aligns well with the focus of this study on programming self-learning ability. A random sampling method was employed, and a total of 296 students were surveyed using a structured questionnaire.

During the survey, we provided participants with examples of commonly used generative AI chatbots, such as ChatGPT, Kimi, and Cursor. Before participation, all participants were informed of the study’s purpose, assured of the confidentiality of their responses, and informed of their right to withdraw at any time. They were then asked to indicate which platforms they used most frequently. The study design minimized risk by using an anonymous online questionnaire without collecting sensitive personal information, thus ensuring participants’ safety and privacy. Furthermore, before participating in the survey, respondents are required to read an informed consent form. Only respondents who confirm their consent to participate will be directed to the questionnaire page, ensuring compliance with ethical standards. Among the 296 returned questionnaires, 42 were excluded from the analysis due to unreasonably short response times or uniform responses across all items. The remaining 254 responses were deemed valid, resulting in a valid response rate of 85.8%.

Participants

The university students who participated in the study covered different majors and grades, as shown in Table 2. There were 113 male students and 191 female students, accounting for 44.5% and 55.5% respectively. There were 27 freshmen students, accounting for 10.6%. There were 53 sophomores, accounting for 20.9%. The number of junior students is 89, accounting for 35%. The number of senior students is 85, accounting for 33.5%. The majors involved are mainly science and engineering, and humanities, accounting for 87.1% in total.

Table 2.

Demographic Statistics of Participants.

Categories	Items	Frequency	Percentage
Gender	Male	113	44.50
Gender	Female	141	55.50
Grade	Freshman	27	10.60
	Sophomore	53	20.90
	Junior	89	35.00
	Senior	85	33.50
Major	STEM	92	36.20
	Business	70	27.60
	Humanities	59	23.20
	Others	33	12.90

Research Survey

This study is quantitative in nature and data were collected using a questionnaire. Through mutual validation and supplementation of the influencing factors proposed in the literature, the influencing factors of chatbots on college students’ ability to learn independently on programming were formulated as LE, EE, FC, TI, LM, and LSE. The questionnaire was adapted from validated scales used in previous studies. The measurements of LE, EE, and FC were adapted from Venkatesh’s et al. (2003) scales for measuring performance expectations, effort expectations, and facilitative conditions, with adjustments made to the wording to suit the specific context.. TI was based on Liu’s (2020) measurement items for teaching interaction in online learning. LM and LSE were adapted from Pintrich’s (1991) scale, with a total of 8 items. The measurement of PSA is adapted from Yang’s (2023) measurement scale for learners’ online self-directed learning ability. To ensure the accuracy of the questionnaire statements, back-translation tests were conducted on some of the referenced foreign-language scales during the questionnaire design process. By consulting experts in the field of online learning research, the rationality of the questionnaire items was evaluated. After evaluation, the content of 7 items was modified. Through a small-scale test of the questionnaire, project analysis, and factor analysis, the final questionnaire was formed as shown in Table 3. The questionnaire uses a 5-point Likert scale, with response options ranging from ‘1 Strongly Disagree’ to ‘5 Strongly Agree’.

Table 3.

Operationalized Definitions of Scale Dimensions and Variables.

Variables	Items	References
LE	LE1:I found using chatbots useful for my programming studies.	Venkatesh et al. (2003)
	LE2:Learning to programme using a chatbot will improve my programming skills.
	LE3:Using a chatbot will improve my programming learning.
	LE4:Learning to programme is easier with a chatbot.
EE	EE1:When learning to programme, the interaction between the chatbot and me is clear and easy to understand.	Venkatesh et al. (2003)
	EE2:It is easy for me to learn programming using chatbots.
	EE3:I can use chatbots to learn programming skilfully.
FC	FC1:I have a chatbot platform that is useful for learning programming.	Venkatesh et al. (2003)
	FC2:I can get enough guidance when learning programming using chatbots.
	FC3:Chatbots are good for learning programming.
TI	TI1:The chatbot responds to my questions in a timely manner during the learning process.	Liu (2020)
	TI2:The chatbot participates in collaborative and discussion activities in my learning process.
	TI3:The chatbot clearly communicates its expectations for learning activities and outcomes.
LM	LM1:I want to learn to programme better than the people around me.	Pintrich (1991)
	LM2:If I work hard enough, I can master programming.
	LM3:Learning programming skills is important to me.
	LM4:Being able to understand the programming language is a very satisfying part of the learning process
LSE	LSE1:I expect to master programming skills.	Pintrich (1991)
	LSE2:I am confident that I understand the basic concepts of the programming I am learning.
	LSE3:I am confident that I can understand the most complex concepts of the programming I am learning.
	LSE4:I am confident that I can master the programming skill.
PSA	PSA1:Learning to programme with a chatbot has made me a more effective learner and more competitive with my classmates.	Yang (2023)
	PSA2:When I use a chatbot to learn programming, I am able to create a specific learning plan based on my learning tasks.
	PSA3:Using a chatbot increases my theoretical and practical knowledge of programming.
	PSA4:When I use chatbot to learn programming, I often reflect on my learning process and learning methods and improve them in time.

Data Analysis

First, we used SPSS 26.0 to analyse the variables with descriptive statistics and reliability. Then we conducted a validation factor analysis (CFA) using AMOS26.0 and validated the theoretical model for goodness-of-fit. Finally, we tested the mechanism through which the variables influence PSA. In addition, we also empirically tested the mediator effect of LM on the relationship between college students’ PSA and LSE.

Results

Descriptive Analysis of Variables

The results of the normality test for the descriptive statistics are show in Table 4. The mean values of all dimensional variables are above 3. The absolute value of the skewness is within 3, and the kurtosis values are within 10. These values fall within the standard range(Prager et al., 2011), indicating that the data satisfy the requirements of an approximate normal distribution.

Table 4.

Results of Our Normality Test for the Descriptive Statistics and Measurement Question Items for Each Dimension.

Dimension	Items	Average	Standard	Skewness	Kurtosi
LE (learning expectation)	LE1–LE4	3.21–3.89	0.89–1.00	−0.55 to 0.06	−0.64 to −0.05
EE (effort expectancy)	EE1–EE3	3.25–3.52	1.02–1.10	−0.30 to −0.22	−0.66 to −0.46
FC (facilitating conditions)	FC1–FC3	3.40–3.91	0.89–1.08	−0.66 to −0.26	−0.58 to 0.07
TI (teaching intention)	TI1–TI3	3.34–3.43	0.84–1.06	−0.22 to −0.06	−0.59 to −0.24
LM (learning motivation)	LM1–LM4	3.11–3.53	1.00–1.10	−0.48 to 0.05	−0.53 to −0.38
LSE (learning self-efficacy)	LSE1–LSE4	3.44–3.71	0.94–1.05	−0.43 to −0.17	−0.46 to −0.26
PSA (programming self-learning ability)	PSA1–PSA4	3.15–3.38	0.97–1.08	−0.28 to −0.05	−0.57 to −0.27

CFA Analysis

In this study, SPSS26.0 and AMOS26.0 were used to measure the fit of the model by testing the reliability and construct validity. As shown in Table 5, the Cronbach’s alpha values of the seven latent variables ranged from .741 to .946, with an overall reliability of .914. All values exceeded the minimum criterion of 0.7, indicating a good fit of the measurement model and high internal consistency. Meanwhile, construct validity was assessed through convergent validity. According to Hair et al. (2021), a scale demonstrates good convergent validity when it meets two conditions: (1) the factor loading value of each measurement item is greater than 0.7; (2) the component reliability value of each potential variable is greater than 0.7. The CFA results are shown in Figure 2. Most items had factor loadings above 0.7 on their corresponding latent variables. Although a few loading are between 0.6 and 0.7, the CR values for all latent variables ranged from 0.740 to 0.947, all above the required threshold. These results indicate that the scale has good convergent validity.

Table 5.

Results of the Reliability Analysis.

Variables	N	Cronbach’s alpha	AVE	CR
LE	4	.816	0.53	0.817
EE	3	.741	0.488	0.74
FC	3	.783	0.554	0.788
TI	3	.854	0.672	0.86
LM	4	.915	0.73	0.915
LSE	4	.9	0.694	0.901
PSA	4	.946	0.817	0.947

Figure 2.

The results of CFA.

Goodness of Fit of the Model

As shown in Table 6, multiple fit indices were used to assess the model’s goodness of fit, including the chi-square to degrees of freedom ratio (X²/ df ), goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), normed fit index (NFI), incremental fit (IFI), and root mean square error of approximation (RMSEA). The results were as follows: GFI = 0.997 > 0.90, AGFI = 0.881 > 0.8, IFI = 0.983 > 0.90, NFI = 0.925 > 0.90, RMSEA = 0.033 < 0.05, and X2/df = 1.278, which is between 1 and 3, which indicates that the constructed model in this paper has good fit.

Table 6.

Goodness of Fit Metrics of the Model.

Indicators	GFI	AGFI	IFI	NFI	RMSEA	x ²/df
Results	0.997	0.881	0.983	0.925	0.033	1.278

Measured Results

The Bootstrapping algorithm in AMOS26.0 was used to repeatedly extract 2,000 subsamples for the operation. The confidence level of the Bias-corrected Confidence Intervals method was set to 95% (α = .05). Based on this, the final structural equation model and its path coefficients were obtained, as shown in Figure 3.

Figure 3.

Structure of the standardized coefficient model for college students’ PSA.

As shown in Table 7, which primarily presents the path coefficients of each variable affecting the dependent variable, except for the FC→PSA path, the critical ratio (C.R.) of the other 5 paths are all greater than 1.96 and p ≤ .005. For example, the path coefficient of LE→LSE is 0.359, with C.R. > 1.96 and p ≤ .001, which suggests that LSE can directly and significantly affect LM. The same rule applies to the other paths.

Table 7.

Path Coefficient.

Path relationship	Estimate	S.E.	C.R.	p
LSE→PSA	0.259	0.06	4.76	***
LM→PSA	0.364	0.074	4.988	***
LE→PSA	0.185	0.081	3.367	***
EE→PSA	0.176	0.059	3.368	***
FC→PSA	0.079	0.072	1.575	.115
TI→PSA	0.294	0.063	5.167	***

***

A significance level of .01.

Mediator Effect Analysis

With the help of AMOS26.0, the mediator effect model was constructed with LSE as the independent variable, PSA as the dependent variable, and LM as the mediator, as shown in Figure 4. The model showed good fit: =55.625, df = 51, RMSEA = 0.019, and NFI = 0.980.

Figure 4.

Mediator effect model.

In this model, LSE significantly predicted both LM (path coefficient = 0.633) and PSA (path coefficient = 0.342). LM also significantly predicted PSA (path coefficient = 0.610). Moreover, all factor loadings of the observed variables (LSE1–LSE4, LM1–LM4, PSA1–PSA4) are above 0.8, confirming good construct validity.

Based on the operating procedure, the sample size was set at 5,000 with a 95% confidence interval. The final analysis results are shown in Table 8. In the mediator effect model, the direct effect of LSE on PSA is 0.418, with a confidence interval of [0.364, 0.597], indicating significance. The indirect effect is 0.471, with a confidence interval of [0.279, 0.569], also significant. These results indicate that LM mediates the relationship between college students’ PSA and LSE, supporting Hypothesis 16: LSE has a positive effect on college students’ PSA through the mediation of LM.

Table 8.

Bootstrapping Test for the Mediator Effect of LM.

Path	Estimate	95% CI
Path	Estimate	Lower	Upper
Direct effect
LSE→PSA	0.418	0.364	0.597
Indirect effect
LSE→LM→PSA	0.471	0.279	0.569

Hypothesis Test Results

In summary, the results of the hypothesis testing are shown in Table 9. FC does not have a significant direct impact on college students’ PSA, suggesting that learners’ self-learning ability is driven by subjective factors, while objective conditions play a relatively minor role.

Table 9.

Hypothesis Test Results.

N	Hypothesis	Results
H1	LE has a significant positive effect on LM.	True
H2	EE has a significant positive effect on LM.	True
H3	FC has a significant positive effect on LM.	True
H4	LE has a significant positive effect on PSA.	True
H5	EE has a significant positive effect on PSA.	True
H6	FC has a significant positive effect on PSA.	False
H7	LM has a significant positive effect on PSA.	True
H8	TI has a significant positive effect on LM.	True
H9	TI has a significant positive effect on PSA.	True
H10	LSE has a significant positive effect on LM.	True
H11	LSE has a significant positive effect on PSA.	True
H12	LE has a significant positive effect on LSE.	True
H13	EE has a significant positive effect on LSE.	True
H14	FC has a significant positive effect on LSE.	True
H15	TI has a significant positive effect on LSE.	True
H16	LSE can indirectly influence PSA through the mediator effect of LM.	True

Discussion

Discussion of Model

In this study, variables were extracted by reviewing relevant literature and a model was constructed based on UTAUT and SDT. Model validation was carried out through questionnaire data analysis. The innovation of the model is reflected in the combination of SDT and UTAUT, which broadens the explanatory scope of PSA research. Additionally, the inclusion of TI, LM and LSE emphasizes both the teaching characteristics of chatbots and the learner-centred nature of autonomous learning. While existing studies on the application of chatbots in education is mainly related to language learning and medicine (Chien et al., 2022; Kohnke, 2023), this study explored their impact on programming-related autonomous learning through empirical analysis, offering a theoretical framework for future research on chatbot applications in technical education.

The Relationship Between LM and PSA

LM had the most significant effect in positively influencing learners’ ability to PSA using chatbots, aligning with the findings of Ait Baha et al. (2023) who found that chatbot-based learning enhances intrinsic motivation and fosters positive learning attitudes. This emphasizes the key role of LM in programming learning supported by chatbots. Through interaction and personalized experiences, chatbots can stimulate students’ interest in learning (Ait Baha et al., 2023), encourage active participation, and improve programming skills. In addition, students in exam-oriented and teacher-centred education systems, such as that in China, may rely less on external support when using emerging technologies and more on their own intrinsic motivation. The autonomous and self-directed nature of chatbot-assisted programming learning further amplifies the importance of learning motivation, especially when learners face complex, abstract tasks with limited external guidance. This may help explain why LM emerged as the most significant factor in this study.

The Relationship Between LE, EE, and PSA

The results of the study show that LE and EE have a positive effect on college students’ PSA, supporting the finding of Foroughi et al. (2024), who confirmed that performance expectancy and EE positively influence the use of ChatGPT in educational contexts. This suggests that when students have higher expectations for programming learning and are willing to invest effort, their PSA can be significantly enhanced.

The Relationship Between TI and PSA

TI also positively influences college students’ PSA, suggesting that chatbots can address the challenge of insufficient student-teacher interactions (Clarizia et al., 2018). Personalized interaction with chatbots cater to individual learning preferences and needs (Ait Baha et al., 2023), thus improving learners’ self-learning ability to a certain extent. Through real-time feedback and guidance, students can better understand the content and improve their PSA. This aligns with the findings of Essel et al. (2022), who found through quantitative analysis that interaction with assistant chatbots has a positive impact on academic performance.

The Relationship Between FC and PSA

FC has no significant positive effect on PSA. Previous studies also found that FC has no significant effect on students’ use of interactive and clear applications (Ameri et al., 2020; Arain et al., 2019). This suggests that when designing and implementing educational strategies, in addition to focussing on the convenience of learning resources, we should pay more attention to stimulating learning motivation and offering personalized support to enhance outcomes. Although convenience plays an important role in the learning environment, its impact in PSA is not critical. This may mean that students perceive the use of chatbots as simple to operate without additional resources or equipment (Strzelecki, 2024). Moreover, chatbot-assisted learning is often accessed voluntarily and informally, where personal motivation and perceived usefulness may play more dominant roles than environmental conditions. Prior studies also suggest that when systems are perceived as easy to use or when users already have sufficient support knowledge, the impact of facilitating conditions becomes negligible (Venkatesh et al., 2003). In this study, students who are familiar with programming tasks, in particular, already possess the necessary technical infrastructure and digital skills. This reduces their dependence on external support.

The Mediator Effect of LM

This paper established that LSE can indirectly affect PSA through LM, which also verifies Cheng’s (2023) findings. Relevant studies have shown that the higher self-efficacy leads to stronger intrinsic motivation (Chi & Xin, 2006). According to Self-Determination Theory (Ryan and Deci, 2000b), individuals need to experience a sense of competence and effectiveness in their tasks in order to sustain intrinsic motivation. Self-efficacy plays a central role in fostering this sense of competence, thereby enhancing learners’ motivation to engage in autonomous learning. Dong (2025) further argue that promoting students’ autonomy significantly improves their learning experiences and motivation. This motivational state, in turn, encourages more active and autonomous learning behaviours. So we need to pay full attention to the roles of LSE and LM in programming self-learning ability. By enhancing students’ LSE and stimulate their LM through effective teaching strategies, their PSA can be further improved.

Limitations and Future Research

Although this study confirms the influence factors of chatbots on college students’ PSA through empirical analysis, several limitations remain. First, the target population of this paper is only for college students. While relevant to the context of programming education, this group may not be representative of the broader student population. These participants likely possess stronger self-motivation and higher programming proficiency than average students, which may have introduced selection bias and limited internal validity. Academic level may also shape learning behaviour—prior studies (Sáiz-Manzanares et al., 2023) suggest that undergraduates and postgraduates differ in their use and perception of chatbot-assisted learning. Additionally, the study did not collect data on important demographic factors such as prior chatbot experience, frequency of use, or familiarity with similar technologies, which may have influenced participants’ responses and limited the depth of interpretation. Future research should consider expanding the sample to include students from more diverse academic levels and backgrounds, and incorporate a broader range of demographic variables to improve the representativeness of the findings.

Second, while this study examined TI and FC as chatbot- related variables, other environmental factors such as instant response of chatbots, personalized teaching, and learning scenarios are also important factors affecting college students’ ability to learn programming autonomously.

Finally, this paper selected UTAUT as the basis of the research model. In UTAUT, gender, age, experience, and voluntariness are used as moderating variables. Although existing data indicate that gender does not influence the use of chatbots, age, experience, and voluntariness do influence people’s use of technology (Menon & Shilpa, 2023; Mogaji et al., 2021). Given that the experimental sample primarily consists of college students participating in a mathematical modelling competition, moderator variables were not included in the actual research. In order to better explore the influence of chatbots on college students’ PSA, more influencing factors can be considered in future research, and experience can be used as a moderating variable.

Conclusions and Improvement Countermeasures

This study explores the influence factors of chatbots on college students’ PSA through the integration of SDT and UTAUT. By targeting college students with a certain level of programming proficiency, collecting data via online questionnaires, and validating the proposed model, this research elucidates both the key factors and the underlying mechanism through which chatbot use impacts PSA. The hypothesis testing results indicate that FC have no significant positive influence on college students’ PSA, whereas LSE can indirectly influence PSA through LM. Based tested path relationships in the model, this part offers targeted suggestions for enhancing chatbot-based programming self-learning outcomes, focussing on three aspects: the interactivity of chatbots, feedback and learners’ self-belief:

Considering that the effect of TI has the second most significant impact on college students’ PSA, following only LM, and that TI positively influences LM, it’s essential to strengthen the interactive functions of chatbots. Research by Hiremath et al. (2018)suggesting that students can benefit from using chatbots to ask questions, receive timely responses, and obtain personalized support. To enhance learner engagement, chatbots can be designed to include interactive learning activities such as quizzes, programming exercises, and online discussions. These features help stimulate students’ intrinsic motivation by encouraging active participation. Moreover, chatbots can support learners by providing individualized learning plans and goal-setting guidance, which helps translate motivation into practical actions and ultimately improves their programming skills.

From the positive influence of LE and EE on college students’ PSA, it can be seen that chatbots should not only provide answers directly, but also guide learners to think independently through skilful prompting statements. These prompts should encourage deep thinking about programming concepts, helping learners understand related knowledge more comprehensively. At the same time, chatbots should guide learners to reflect on and summarize their learning experience. This helps create a self-growth environment where learners gradually recognize their own process through continuous experimentation, thinking, and reflection, thereby enhancing their learning expectancy. Chatbots should also provide immediate positive feedback and encouragement. This not only helps to enhance learners’ LSE but also inspires them to have higher expectations of their efforts. As a result, learners develop stronger programming self-awareness and engage in the learning process more actively.

Since TI has a positive effect on LSE, and LSE also has a positive effect on PSA, learners should internalize the concept of independent learning and lifelong learning. They should cultivate an awareness of their own learning progress as a way to continuously strengthen LSE. The understanding and confidence gained from secondary school learning can be transformed into improved academic performance (Sandu & Gide, 2019). Through the chatbot platform, learners can not only control their own learning pace but also continuously identify and solve problems in practice. This process helps them gain a deeper understanding of programming knowledge.

Footnotes

We sincerely thank the editors and all reviewers for their comments and suggestion. We also appreciate the people who patiently participated in the study and the valuable data they provided.

ORCID iDs

Xue Ran

Zhigang Li

Yalin Yang

Ethical Considerations

This study did not require further ethics committee approval as it did not involve animal or human clinical trials and was not unethical. The study was designed as an anonymous online survey with no sensitive or personally identifiable questions,and participations were entirely voluntary.

Consent to Participate

We declare that all participants voluntarily gave informed consent after being informed of the study’s purpose,confidentiality,and their right to withdraw before submitting the questionnaire. Informed consent was obtained through an explicit consent statement presented at the beginning of the online survey.

Author Contributions

All authors contributed to the study conception and design. Conceptualization,validation,writing—original draft,writing—review and editing and investigation were performed by Xue Ran. Writing—review and editing and finding were performed by Zhigang Li. Investigation and writing—review and editing were performed by Yalin Yang.

Funding

The authors disclosed receipt of the following financial support for the research,authorship and/or publication of this article: This work was funded by the Major Project on Research into Reform of Graduate Education and Teaching at Chengdu University of Technology (Grant number: 2024YJG008).

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research,authorship,and/or publication of this article.

Data Availability Statements

The datasets analysed during the current study are available from the corresponding author on reasonable request.

Potential Benefits and Risk Assessment

The potential benefits include advancing understanding of chatbot-assisted learning in programming education,and offering participants an opportunity to reflect on their learning experiences. The potential risks were minimal,involving only possible minor discomfort from survey questions,and were mitigated through anonymity and voluntary participation. Overall,the benefits to society and participants clearly outweighed the limited risks.

References

Ait Baha

El Hajji

Es-Saady

Fadili

(2023). The impact of educational chatbot on student learning experience. Education and Information Technologies, 29(8), 10153–10176. https://doi.org/10.1007/s10639-023-12166-w

Ameri

Khajouei

Ameri

Jahani

(2020). Acceptance of a mobile-based educational application (LabSafety) by pharmacy students: An application of the UTAUT2 model. Education and Information Technologies, 25(1), 419–435. https://doi.org/10.1007/s10639-019-09965-5

Arain

A. A.

Hussain

Rizvi

W. H.

Vighio

M. S.

(2019). Extending UTAUT2 toward acceptance of mobile learning in the context of higher education. Universal Access in the Information Society, 18(3), 659–673. https://doi.org/10.1007/s10209-019-00685-8

Augustyniak

R. A.

Ables

A. Z.

Guilford

Lujan

H. L.

Cortright

R. N.

DiCarlo

S. E.

(2016). Intrinsic motivation: An overlooked component for student success. Advances in Physiology Education, 40(4), 465–466. https://doi.org/10.1152/advan.00072.2016

Bandura

(1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2), 191–215. https://doi.org/10.1037//0033-295x.84.2.191

Bassi

Steca

Fave

A. D.

Caprara

G. V.

(2007). Academic self-efficacy beliefs and quality of experience in learning. Journal of Youth and Adolescence, 36, 301–312. https://doi.org/10.1007/s10964-006-9069-y

Cheng

(2023). Online learning autonomy an empirical study on affecting factors of university students’ English. China Educational Technology, 9, 123–131.

Chen

(2004). An investigation into “interactivity” and the related concepts. Journal of Distance Education, 3, 12–16. https://doi.org/10.13541/j.cnki.chinade.2004.03.004

Chien

Y.-C.

T.-T.

Lai

C.-H.

Huang

Y.-M.

(2022). Investigation of the influence of artificial intelligence markup language-based LINE ChatBot in contextual English learning. Frontiers in Psychology, 13, 785752. https://doi.org/10.3389/fpsyg.2022.785752

10.

Chi

Xin

(2006). The measure of learning motivation and the relationship between it and self-efficacy of college students. Psychological Development and Education, 2, 64–70.

11.

Chiu

T. K. F.

(2022). Applying the self-determination theory (SDT) to explain student engagement in online learning during the COVID-19 pandemic. Journal of Research on Technology in Education, 54(sup1), S14–S30. https://doi.org/10.1080/15391523.2021.1891998

12.

Chiu

T. K. F.

Moorhouse

B. L.

Chai

C. S.

Ismailov

(2023). Teacher support and student motivation to learn with Artificial Intelligence (AI) based chatbot. Interactive Learning Environments, 32(7), 3240–3256. https://doi.org/10.1080/10494820.2023.2172044

13.

Clarizia

Colace

Lombardi

Pascale

Santaniello

(2018). Chatbot: An Education Support System for Student. Paper presented at Cyberspace Safety and Security: 10th International Symposium, Amalfi, Italy, 29–31 October 2018. https://doi.org/10.1007/978-3-030-01689-0_23

14.

Deci

E. L.

Ryan

R. M.

(1985). The general causality orientations scale: Self-determination in personality. Journal of Research in Personality, 19(2), 109–134. https://doi.org/10.1016/0092-6566(85)90023-6

15.

Deci

E. L.

Ryan

R. M.

(2000). The “What” and “Why” of goal pursuits: Human needs and the self-determination of behavior. Psychological Inquiry, 11(4), 227–268. https://doi.org/10.1207/s15327965pli1104_01

16.

Deci

E. L.

Ryan

R. M.

(2013). Intrinsic motivation and self-determination in human behavior. Springer Science & Business Media.

17.

Dečman

(2015). Modeling the acceptance of e-learning in mandatory environments of higher education: The influence of previous education and gender. Computers in Human Behavior, 49, 272–281. https://doi.org/10.1016/j.chb.2015.03.022

18.

Deng

(2023). A meta-analysis and systematic review of the effect of chatbot technology use in Sustainable Education. Sustainability, 15(4), 2940. https://doi.org/10.3390/su15042940

19.

Dong

(2025). Self-determination theory perspectives on the influence of digital learning engagement on motivation in extracurricular learning activities: Considering the mediating role of digital self-efficacy. Learning and Motivation, 90, 102135. https://doi.org/10.1016/j.lmot.2025.102135

20.

Dwivedi

Y. K.

Rana

N. P.

Jeyaraj

Clement

Williams

M. D.

(2019). Re-examining the Unified Theory of Acceptance and Use of Technology (UTAUT): Towards a Revised Theoretical Model. Information Systems Frontiers, 21(3), 719–734. https://doi.org/10.1007/s10796-017-9774-y

21.

Dysvik

Kuvaas

(2013). Intrinsic and extrinsic motivation as predictors of work effort: The moderating role of achievement goals. British Journal of Social Psychology, 52(3), 412–430. https://doi.org/10.1111/j.2044-8309.2011.02090.x

22.

Essel

H. B.

Vlachopoulos

Tachie-Menson

Johnson

E. E.

Baah

P. K.

(2022). The impact of a virtual teaching assistant (chatbot) on students’ learning in Ghanaian higher education. International Journal of Educational Technology in Higher Education, 19(1), 57. https://doi.org/10.1186/s41239-022-00362-6

23.

Firmansyah

Komala

Rusdi

(2018). Self-efficacy and motivation: Improving biology learning outcomes of senior high school students. JPBI (Jurnal Pendidikan Biologi Indonesia), 4(3), 203–208. https://doi.org/10.22219/jpbi.v4i3.6878

24.

Foroughi

Senali

M. G.

Iranmanesh

Khanfar

Ghobakhloo

Annamal-ai

Naghmeh-Abbaspour

(2024). Determinants of intention to use Chat-GPT for educational purposes: Findings from PLS-SEM and fsQCA. International Journal of Human–Computer Interaction, 40(17), 4501–4520. https://doi.org/10.1080/10447318.2023.2226495

25.

Fredricks

J. A.

Blumenfeld

P. C.

Paris

A. H.

(2004). School engagement: Potential of the concept, state of the evidence. Review of Educational Research, 74(1), 59–109. https://doi.org/10.3102/00346543074001059

26.

Gan

(2020). How learning motivation influences feedback experience and preference in Chinese University EFL Students. Frontiers in Psychology, 11, 296. https://doi.org/10.3389/fpsyg.2020.00496

27.

Hair

J. F.

Jr Hult

G. T. M.

Ringle

Sarstedt

Danks

Ray

(2021). Partial least squares structural equation modeling (PLS-SEM) using R: A workbook. Springer Nature.

28.

Hiremath

Hajare

Bhosale

Nanaware

Wagh

K. S.

(2018). Chatbot for education system. International Journal of Advance Research, Ideas and Innovations in Technology, 4(3), 37–43.

29.

Kim

Y. E.

S. L.

Shin

(2022). How temptation changes across time: Effects of self-efficacy for self-regulated learning and autonomy support. Educational Psychologist, 42(3), 278–295. https://doi.org/10.1080/01443410.2021.2009774

30.

Kiptonui

(2013). Chatbot technology: A possible means of unlocking student potential to learn how to learn. Educational Researcher, 4(2), 218–221.

31.

Koç

F. Ş.

Savaş

(2025). The use of artificially intelligent chatbots in English language learning: A systematic meta-synthesis study of articles published between 2010 and 2024. ReCALL, 37(1), 4–21. https://doi.org/10.1017/s0958344024000168

32.

Kohnke

(2023). A pedagogical chatbot: A supplemental language learning tool. RELC Journal, 54(3), 828–838. https://doi.org/10.1177/00336882211067054

33.

Lin

C. J.

Mubarok

(2021). Learning analytics for investigating the mind map-guided AI Chatbot approach in an EFL flipped speaking classroom. Educational Technology & Society, 24(4), 16–35.

34.

Liu

(2020). College students’ online self-regulated learning: The impact and intervention of teaching interaction. Dissertation. Shaanxi Normal University.

35.

Manaswi

N. K.

(Ed.). (2018). Developing chatbots. In Deep Learning with applications using Python (pp. 145–170). Apress.

36.

Menon

Shilpa

(2023). “Chatting with ChatGPT”: Analyzing the factors influencing users’ intention to use the open AI’s ChatGPT using the UTAUT model. Heliyon, 9(11), e20962. https://doi.org/10.1016/j.heliyon.2023.e20962

37.

Mogaji

Balakrishnan

Nwoba

A. C.

Nguyen

N. P.

(2021). Emerging-market consumers’ interactions with banking chatbots. Telematics and Informatics, 65, 101711. https://doi.org/10.1016/j.tele.2021.101711

38.

Moore

M. G.

(1989). Editorial: Three types of interaction. American Journal of Distance Education, 3(2), 1–7. https://doi.org/10.1080/08923648909526659

39.

Nikou

S. A.

Economides

A. A.

(2017). Mobile-based assessment: Integrating acceptance and motivational factors into a combined model of self-determination theory and technology acceptance. Computers in Human Behavior, 68, 83–95. https://doi.org/10.1016/j.chb.2016.11.020

40.

Pérez-Marín

(2021). A review of the practical applications of pedagogic conversational agents to Be used in school and University classrooms. Digital, 1(1), 18–33. https://doi.org/10.3390/digital1010002

41.

Pintrich

P. R.

(1991). A manual for the use of the motivated strategies for learning questionnaire (MSLQ). National Center for Research to Improve Postsecondary Teaching and Learning (NCRIPTAL), University of Michigan.

42.

Prager

Razzi

Blaj-Ward

Schmidt

J. J.

(2011). Becoming a behavioral science researcher. A guide to producing research that matters. International Journal of Social Research Methodology, 14(4), 327–335. https://doi.org/10.1080/13645579.2011.597123

43.

Ryan

R. M.

Deci

E. L.

(2000a). Intrinsic and extrinsic motivations: Classic definitions and new directions. Contemporary Educational Psychology, 25(1), 54–67. https://doi.org/10.1006/ceps.1999.1020

44.

Ryan

R. M.

Deci

E. L.

(2000b). Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55(1), 68–78. https://doi.org/10.1037/0003-066X.55.1.68

45.

Sáiz-Manzanares

M. C.

Marticorena-Sánchez

Martín-Antón

L. J.

González Díez

Almeida

(2023). Perceived satisfaction of university students with the use of chatbots as a tool for self-regulated learning. Heliyon, 9(1), e12843. https://doi.org/10.1016/j.heliyon.2023.e12843

46.

Sandu

Gide

(2019). Adoption of AI-Chatbots to Enhance Student Learning Experience in Higher Education in India. In 2019 18th International Conference on Information Technology Based Higher Education and Training (ITHET), 1–5. https://doi.org/10.1109/ITHET46829.2019.8937382

47.

Smutny

Schreiberova

(2020). Chatbots for learning: A review of educational chatbots for the Facebook Messenger. Computers & Education, 151, 103862. https://doi.org/10.1016/j.compedu.2020.103862

48.

Strzelecki

(2024). To use or not to use ChatGPT in higher education? A study of students’ acceptance and use of technology. Interactive Learning Environments, 32(9), 5142–5155. https://doi.org/10.1080/10494820.2023.2209881

49.

Trautner

Schwinger

(2020). Integrating the concepts self-efficacy and moti-vation regulation: How do self-efficacy beliefs for motivation regulation influence self-regulatory success? Learning and Individual Differences, 80, 101890. https://doi.org/10.1016/j.lindif.2020.101890

50.

Venkatesh

(2022). Adoption and use of AI tools: A research agenda grounded in UTAUT. Annals of Operations Research, 308(1-2), 641–652. https://doi.org/10.1007/s10479-020-03918-9

51.

Venkatesh

Morris

M. G.

Davis

G. B.

Davis

F. D.

(2003). User acceptance of information technology: Toward a Unified View. MIS Quarterly, 27(3), 425–478. https://doi.org/10.2307/30036540

52.

Yang

(2023). Construction and empirical research of online self-regulated learning model supported by visualization technology. [Dissertation]. Northeast Normal University.

53.

Yin

Goh

T.-T.

Yang

Xiaobin

(2021). Conversation technology with micro-learning: The impact of chatbot-based learning on students’ learning moti-vation and performance. Journal of Educational Computing Research, 59(1), 154–177. https://doi.org/10.1177/0735633120952067

54.

Zhang

Huang

(2021). The empirical study on the relationship between undergraduates engagement and self-efficacy in learning. Education Research Monthly, 11, 83–90. https://doi.org/10.16477/j.cnki.issn1674-2311.2021.11.012