behavioral

sciences
Article
The Impact of AI Usage on University Students’ Willingness for
Autonomous Learning
Ling Wang 1,2 and Wenye Li 1, *

1 Institute of Education, Nanjing University, Nanjing 210093, China; 602023120006@smail.nju.edu.cn

2 College of Education, Yili Normal University, Yining 835000, China
* Correspondence: wenyeli@smail.nju.edu.cn

Abstract: As artificial intelligence (AI) technology becomes increasingly integrated into education,
understanding the theoretical mechanisms that drive university students to adopt new learning
behaviors through these tools is essential. This study extends the Expectation-Confirmation Model
(ECM) by incorporating both cognitive and affective variables to examine students’ current AI usage
and their future expectations. The model includes intrinsic and extrinsic motivations, focusing on
three key factors: positive emotions, digital efficacy, and willingness for autonomous learning. A
survey of 721 valid responses revealed that positive emotions, digital efficacy, and satisfaction signifi-
cantly influence continued AI usage, with positive emotions being particularly critical. Digital efficacy
and perceived usefulness also impact satisfaction, but long-term usage intentions are more effectively
driven by positive emotions. Furthermore, digital efficacy strongly affects the willingness for au-
tonomous learning. Therefore, higher education institutions should promote AI technology, enhance
students’ expectation-confirmation levels, and emphasize positive emotional experiences during AI
use. Adopting a “human–machine symbiosis” model can foster active learning, personalized learning
pathways, and the development of students’ digital efficacy and innovation capabilities.

Keywords: artificial intelligence technology; expectation-confirmation model; willingness for au-

tonomous learning; positive emotions; digital efficacy

Citation: Wang, L.; Li, W. The Impact 1. Introduction

of AI Usage on University Students’
Prior to the advent of artificial intelligence (AI) technology, autonomous learning
Willingness for Autonomous
had already been extensively explored. Over the past decade, AI technology has experi-
Learning. Behav. Sci. 2024, 14, 956.
https://doi.org/10.3390/bs14100956
enced unprecedented exponential growth, posing significant challenges to the practices
and development of higher education [1,2]. Simultaneously, research and exploration
Academic Editors: Rebecca Upsher, into autonomous learning within both formal and informal learning environments have
Claire Heard and Sumeyra Yalcintas deepened. AI impacts university students’ learning through personalized learning experi-
Received: 31 August 2024 ences, adaptive guidance, intelligent tutoring systems, immersive learning technologies,
Revised: 5 October 2024 and automated content creation [3]. Notably, generative AI technology, by simulating
Accepted: 10 October 2024 human-like interactive communication, allows students to access enriched and person-
Published: 16 October 2024 alized learning resources, enabling them to control their learning pace and steps, thus
reshaping their perception of autonomous learning [4,5]. The effective use of generative
AI technology for autonomous learning is becoming increasingly critical for university
students to acquire skills and adapt to the rapidly changing job market, potentially leading
Copyright: © 2024 by the authors.
to significant differences in their learning outcomes and employment opportunities. Essel
Licensee MDPI, Basel, Switzerland.
and Vlachopoulos demonstrated through pre- and post-tests that students interacting with
This article is an open access article
chatbots performed better academically than those interacting with teachers [6]. Atalas’
distributed under the terms and
research further indicated that students perceived generative AI technology not only as a
conditions of the Creative Commons
source of personalized learning support, offering them round-the-clock tailored resources,
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
but also as an aid for literature searches, abstract reading, and even hypothesis generation
4.0/).
based on data analysis [7], thereby deepening their learning based on the latest research

Behav. Sci. 2024, 14, 956. https://doi.org/10.3390/bs14100956 https://www.mdpi.com/journal/behavsci

Behav. Sci. 2024, 14, 956 2 of 16

trends [8]. These studies collectively underscore the substantial value of AI technology in
higher education, particularly in its recognized potential to support university students’
learning [9]. However, existing research falls short of fully explaining the specific factors
influencing university students’ willingness to engage in autonomous learning with AI
technology.
This study focuses on whether university students’ willingness for autonomous learn-
ing has shifted with the advent of AI technology as a disruptive tool. Given that generative
AI can be applied in both formal and informal learning environments, this research expands
beyond classroom-based autonomous learning to comprehensively examine university
students’ autonomous learning in both formal and informal contexts. To investigate the
impact of generative AI technology on students’ willingness for autonomous learning, it is
essential first to analyze their willingness to use generative AI and the factors influencing
their continued use; second, to examine whether the intention for continued use affects
their willingness for autonomous learning; and finally, to clarify the specific mechanisms
underlying this influence. Addressing these questions will aid educators and policymakers
in understanding how best to integrate these technologies into higher education to support
personalized learning for university students.

2. Literature Review and Research Hypotheses

2.1. Expectation-Confirmation Model and Its Limitations
The theoretical framework for this study is based on the Expectation-Confirmation
Model (ECM), initially proposed by Bhattacherjee [10]. ECM has been extensively applied
in various domains, including e-commerce, online learning, and technology usage [11–13].
The model comprises four core constructs: expectation confirmation (the extent to which an
individual’s initial expectations are met after using an information system) [14], perceived
usefulness (the degree to which an individual believes that using a system will enhance their
performance) [15], satisfaction (an individual’s positive evaluation of their experience) [16],
and continuance intention (the individual’s willingness to continue using the current
system) [17]. Compared to the Technology-Acceptance Model (TAM), ECM focuses more
on the psychological motivations following the initial use of technology, making it more
reflective of the satisfaction experienced during the technology-usage process. Therefore,
this study employs ECM to investigate the factors influencing university students’ use of
generative AI technology for learning, providing insights that can help educators create
conditions conducive to learning and AI technology usage in higher education.
As illustrated in Figure 1, the original model posits that an individual’s satisfaction
and perceived usefulness of a technology determine their intention to continue using it.
Satisfaction, in turn, is influenced by expectation confirmation and perceived usefulness,
with expectation confirmation serving as a key determinant of perceived usefulness. Past
research has widely employed ECM in studies related to information technology, providing
a theoretical foundation for understanding university students’ satisfaction with AI tech-
nology. Generative AI, compared to other information technologies, is a more personalized,
intelligent, and instantly responsive learning tool that subtly alters users’ emotions and
behaviors, leading them to find value in the interaction with the technology, rather than
just its intrinsic technical value. Therefore, this study incorporates new external variables
involving interactions with generative AI technology into the original ECM model to verify
whether the model still holds. This further addresses the factors influencing college stu-
dents’ satisfaction with AI technology and the mechanisms affecting their continued use of
generative AI for autonomous learning.
This research focuses on three critical external variables: digital efficacy, positive
emotions, and the willingness for autonomous learning. First, digital efficacy reflects
whether university students possess the innovative and exploratory characteristics required
in a technologically advanced learning environment and whether they believe in their
ability to master AI usage strategies to support personalized and deep learning [18]. Second,
positive emotions are indicative of the appeal of generative AI technology to students
 Behav. Sci. 2024, 14, 956 3 of 16

and reveal the alignment between the resources provided by the technology and the
students’ learning goals [19], influencing their future continuance intention. Third, while
the traditional ECM and its extended models typically culminate in the formation of
continuance intention, this study introduces the construct of willingness for autonomous
Behav. Sci. 2024, 14, x FOR PEER REVIEW 3 of 1
learning to examine whether continued satisfaction and intention to use generative AI
technology lead to an increase in students’ autonomous learning willingness.

Figure1.1.The
Figure The original
original ECM.
ECM.

2.2. Digital Efficacy

This research focuses on three critical external variables: digital efficacy, positiv
Self-efficacy,
emotions, and as thea subjective
willingness factor
formediating
autonomous between motivation
learning. anddigital
First, behavior, refers reflect
efficacy
to an individual’s belief in their ability to accomplish a specific task or set of tasks—a
whether university students possess the innovative and exploratory characteristics re
self-perception of capability [20]. In this study, digital efficacy refers to university students’
quired in a technologically advanced learning environment and whether they believe in
perceived competence and confidence in utilizing AI technology. This perception reflects
their ability
whether to master
students AI usageleverage
can effectively strategies AIto support
tools personalized
to access and deep learning
valuable information and [18]
Second, positive
guidance, overcomeemotions are indicative
potential limitations of the appealrelated
or misconceptions of generative
to AI, andAI technology
achieve their to stu
desired learning outcomes. Conversely, students who doubt their technological abilities and th
dents and reveal the alignment between the resources provided by the technology
students’
may learning
adopt surface goals [19],
learning influencing
approaches, focusingtheir
on future continuance
rote memorization andintention. Third, whil
meeting mini-
mal
the academic
traditional requirements
ECM and its [21]. Bhattacherjee
extended models posited that expectation
typically culminate confirmation
in the formation not of con
only enhances individuals’ satisfaction with technology but also
tinuance intention, this study introduces the construct of willingness for autonomoustrengthens their digital
efficacy in using the technology [10]. When university students find that the functionalities
learning to examine whether continued satisfaction and intention to use generative A
and outcomes of generative AI technology meet or exceed their expectations, they are more
technology lead to an increase in students’ autonomous learning willingness.
likely to use the technology confidently and effectively, experience higher satisfaction in
their learning process, and regard AI tools as vital resources supporting their autonomous
2.2. Digital
learning [22].Efficacy
Similarly, in the context of online universities in South Korea, digital self-
efficacy and perceived
Self-efficacy, as usefulness
a subjective were significant
factor predictors
mediating of learners’
between satisfaction
motivation [23].
and behavior, re
Additionally, students’ perceptions of technological innovation and their experience
fers to an individual’s belief in their ability to accomplish a specific task or set of tasks— with it
can influence theirof
self-perception willingness
capability to [20].
use the Intechnology
this study, anddigital
the extent to which
efficacy theytointegrate
refers university stu
it into their learning processes [24]. Moreover, Zadorozhnyy and Lee demonstrated that
dents’ perceived competence and confidence in utilizing AI technology. This perception
individuals’ confidence in their digital technology skills significantly impacts their willing-
reflects whether students can effectively leverage AI tools to access valuable information
ness to engage in second language communication during informal digital learning [25].
andsuggests
This guidance, thatovercome potential
digital efficacy limitations
can directly enhance orstudents’
misconceptions related to AI,
learning autonomy, and achiev
thereby
their desired learning outcomes.
promoting autonomous learning [26,27]. Conversely, students who doubt their technological abil
itiesBased
may adopt
on these surface learning
findings, approaches,
this study hypothesizes focusing on rote memorization
that expectation confirmation posi-and meeting
tively
minimalinfluences
academic digital efficacy, and, in
requirements turn,
[21]. digital efficacy
Bhattacherjee positively
posited thatimpacts satisfaction,
expectation confirmation
continuance intention,
not only enhances and the willingness
individuals’ satisfaction for autonomous learning.
with technology but also strengthens their dig
ital efficacy in using the technology [10]. When university students find that the function
2.3. Positive Emotions
alities and outcomes of generative AI technology meet or exceed their expectations, they
Positive emotions refer to the degree of engagement students have with their learning
are more likely to use the technology confidently and effectively, experience higher satis
content and the positive experiences they derive from the learning process [28]. In this study,
faction emotions
positive in their learning process,
are defined and regard
as the emotional AI tools as
experiences vital resources
of university supporting
students when thei
autonomous learning [22]. Similarly, in the context of online universities
using generative AI technology for learning, including feelings of pleasure, satisfaction,in South Korea
digital self-efficacy
excitement, and a senseandof perceived
immersion.usefulness wereserve
These emotions significant predictors
as intrinsic of learners’
motivations that satis
faction
drive [23]. Additionally,
judgment and behavior, students’
and theyperceptions
are based onofthe
technological
fulfillment ofinnovation
expectations.andIt their ex
isperience
important to distinguish
with between
it can influence positive
their emotionstoand
willingness usesatisfaction, as bothand
the technology relate
thetoextent to
which they integrate it into their learning processes [24]. Moreover, Zadorozhnyy and Le
demonstrated that individuals’ confidence in their digital technology skills significantly
impacts their willingness to engage in second language communication during informa
digital learning [25]. This suggests that digital efficacy can directly enhance students
Behav. Sci. 2024, 14, 956 4 of 16

emotional experiences but differ conceptually. The key difference lies in the timing of
measurement: satisfaction is fundamentally a retrospective evaluation following use [29],
whereas positive emotions focus on the quality of emotions experienced during the process
of interacting with the technology, acting as a motivational force throughout the experience.
Therefore, positive emotions and satisfaction are theoretically distinct constructs.
Oliver’s Expectation-Confirmation Theory has established that when a consumer’s
actual experience aligns with or exceeds expectations, positive emotional experiences tend
to increase [30]. Previous research has concluded that students who have a positive attitude
towards using ChatGPT for language learning are more likely to engage emotionally with
the tool and perceive it as a beneficial resource for language acquisition [31]. Moreover, such
positive emotions are associated with higher behavioral intentions to use ChatGPT [32]
and strongly predict the actual use of ChatGPT for autonomous English learning outside
the classroom [33].
Based on these findings, this study hypothesizes that expectation confirmation pos-
itively influences positive emotions, and, in turn, positive emotions positively impact
continuance intention.

2.4. Willingness for Autonomous Learning

A high continuance intention to use generative AI technology among university stu-
dents is likely to lead to a positive outcome: an increased willingness for autonomous
learning. Willingness for autonomous learning refers to the learners’ tendency and incli-
nation to actively set learning goals, select appropriate learning methods, and engage in
self-monitoring during the learning process. This concept emphasizes students’ proactivity
and self-regulation in learning, which are crucial for lifelong learning and personal devel-
opment. In the age of AI, autonomous learning can fully harness the educational benefits of
generative AI, thereby enhancing learners’ independence and self-sufficiency [34]. In this
study, willingness for autonomous learning is defined as university students’ inclination to
actively utilize generative AI technology for learning activities and planning.
Previous studies have shown that AI interventions can increase students’ enjoyment
and willingness to communicate in foreign language learning while significantly reducing
anxiety levels, indicating that AI positively impacts spoken language learning [35]. Such
interventions can guide students towards autonomous learning and enhance their learning
motivation [36]. Moreover, students’ academic backgrounds may also play a role in under-
standing their willingness for autonomous learning, as prior research suggests that learning
experiences can influence students’ willingness to use technology for learning [37]. Based
on these findings, this study hypothesizes that university students’ continuance intention
to use generative AI positively influences their willingness for autonomous learning and
that willingness for autonomous learning varies across students with different academic
backgrounds.
In summary, the theoretical model of this study is illustrated in Figure 2. The revised
model integrates satisfaction and expectation confirmation regarding technology use, dig-
ital efficacy as a personal trait, positive emotions as intrinsic motivation, and perceived
usefulness as extrinsic motivation. Through this model, the study aims to address the
following four research questions:
Does the Expectation-Confirmation Model (ECM) apply to explaining university stu-
dents’ satisfaction and continuance intention in using generative AI for learning activities?
Can expectation confirmation, through digital efficacy and positive emotions, enhance
students’ satisfaction and continuance intention?
Does continuance intention improve students’ willingness for autonomous learning?
Does willingness for autonomous learning differ among university students from
different academic backgrounds?
 dents’ satisfaction and continuance intention in using generative AI for learning activities
Can expectation confirmation, through digital efficacy and positive emotions, en
hance students’ satisfaction and continuance intention?
Does continuance intention improve students’ willingness for autonomous learning
Behav. Sci. 2024, 14, 956 Does willingness for autonomous learning differ among university students 5 of 16 from
different academic backgrounds?

Figure2.2.Theoretical
Figure Theoretical model.
model.

3. Research Methodology
3. Research Methodology
3.1. Questionnaire Design
3.1. Questionnaire Design
The questionnaire is divided into three sections. The first section collects basic demo-
graphicThe questionnaire
information such is
as divided into three
gender, grade level,sections.
major, and The
thefirst
typessection collects basic demo
of AI technologies
graphic
used information
by the such
participants. Theassecond
gender, gradecomprises
section level, major, and across
27 items the types of AI technologie
8 variables, all
measured
used by the using a 5-point Likert
participants. The scale.
secondThese include
section four core
comprises 27variables: expectation
items across 8 variables, al
confirmation,
measured using perceived usefulness,
a 5-point Likert satisfaction,
scale. Theseand continuance
include intention;
four core one outcome
variables: expectation con
variable: willingness for autonomous learning; and two newly introduced variables: self-
firmation, perceived usefulness, satisfaction, and continuance intention; one outcome var
efficacy and positive emotions. The specific items and their sources are detailed in Table 1,
iable: willingness for autonomous learning; and two newly introduced variables: self-ef
with item wording adapted to fit the context of this study. The third section includes
ficacy
an and positive
open-ended emotions.
question aimed atThe specific
gathering items and
qualitative their sources
insights are detailed
into university in Table 1
students’
with item wording
experiences adapted
with generative AIto fit the context
technology of learning
in their this study. The third
activities, whichsection
servesincludes
to an
open-ended question aimed at gathering qualitative
supplement and contextualize the quantitative results. insights into university students’ ex
periences with generative AI technology in their learning activities, which serves to sup
Table 1. Scale
plement anditems and design basis.
contextualize the
quantitative results.
Latent Variable Measurement Items Source
Table 1. Scale items and design basis.
1. The learning outcomes from using generative AI exceeded my expectations.
Latent Expectation
Variable 2. Using generative Measurement
AI has enriched myItems
learning experience. Source
Confirmation 1. 3. Generative AI provides the learning resources I need.
The learning outcomes from using generative AI exceeded my expectations.
4. The use of generative AI tools meets my expectations and needs.
Expectation Con- 2. Using generative AI has enriched my learning experience.
1. I believe that using generative AI enhances my learning efficiency.
firmation 2. I3.believe
Generative AIgenerative
that using providesAIthe learning
improves resources
my learning I need.
outcomes.
3. I believe that using generative AI helps me achieve my learning goals.needs.
Adapted from Da-
Perceived Usefulness 4. The use of generative AI tools meets my expectations and Adapted from
vis etDavis
al. (1989) [15]
4. Using generative AI aids in my deep understanding of the learning content.
1. I believe 5. that
Using using generative
generative AI tools AI
is an enhances
effective usemy
of learning
my time. efficiency. et al. (1989) [15]
Perceived Useful-
2. I1.believe that with
I am satisfied usingthegenerative
benefits thatAI improves
generative my learning
AI brings outcomes.
to my learning.
ness
3. I believe that using
2. The information andgenerative
suggestionsAI helps by
provided megenerative
achieve AI myduring
learning
the goals.
Satisfaction
learning process are useful.
3. Generative AI technology meets my personalized learning needs.
4. I am satisfied with the way generative AI interacts with me (e.g., response
speed, response quality).
Behav. Sci. 2024, 14, 956 6 of 16

Table 1. Cont.

Latent Variable Measurement Items Source

1. I intend to continue using generative AI technology.
Adapted from Davis
Continuance Intention 2. I will consider using generative AI to assist me when I have learning needs.
et al. (1989) [15]
3. I would recommend generative AI technology to other students for learning.
1. I am willing to use generative AI for autonomous learning.
Willingness for 2. After using generative AI tools, I am better at planning my learning
Adapted from Davis
Autonomous Learning activities.
(1989) [15] and
3. After using generative AI tools, I actively seek and learn more related
Venkatesh et al. (2003)
knowledge.
[38]
4. After using generative AI tools, I find it easier to find motivation and
direction for learning.
1. I can use AI technology to obtain the information I need.
2. I am confident that I can learn AI-related skills even in unfamiliar areas. Adapted from
Self-Efficacy 3. I expect that I can keep up with the advancements in AI technology and Compeau and Higgins
complete diverse learning tasks. (1995) [39]
4. I can customize AI technology based on my learning needs.
1. Using generative AI tools makes me more curious and interested in learning,
prompting me to explore further based on specific dialogues.
2. Using generative AI tools effectively reduces my procrastination and Adapted from Ozkan
Positive Emotions
learning anxiety. and Koseler (2009) [40]
3. When using generative AI for learning, I often feel positive and optimistic,
even when facing challenges.

3.2. Reliability and Validity Testing

This study employed confirmatory factor analysis (CFA) to assess the reliability and
validity of the measurement scale. Data validity was examined using MPLUS 8.3, focusing
on item reliability, composite reliability (CR), and convergent validity (AVE). The results
indicated that all items had reliability coefficients greater than 0.770, with squared multiple
correlations ranging between 0.593 and 0.682. The lowest CR value exceeded 0.8, and the
lowest AVE value was above 0.628. These findings demonstrate that the reliability and
validity of the variables meet acceptable standards, thereby supporting the subsequent
structural equation modeling (SEM) analysis (see Table 2).

Table 2. Reliability and validity of measurement variables.

Internal Squared
Item Reliability Consistency Multiple Convergent
Variable
(STD) (Cronbach’s Correlation Validity (AVE)
Alpha) (SMC)
Expectation
0.781–0.807 0.871 0.611–0.652 0.628
Confirmation
Perceived
0.779–0.813 0.898 0.606–0.661 0.639
Usefulness
Satisfaction 0.790–0.816 0.88 0.625–0.666 0.648
Continuance
0.789–0.804 0.838 0.622–0.646 0.634
Intention
Willingness for
Autonomous 0.791–0.818 0.878 0.626–0.670 0.643
Learning
Digital Efficacy 0.770–0.816 0.875 0.593–0.665 0.637
Positive
0.794–0.826 0.88 0.630–0.682 0.647
Emotions
Behav. Sci. 2024, 14, 956 7 of 16

3.3. Research Procedure

After completing the initial design of the questionnaire, a pilot study was conducted
by inviting 35 students through a WeChat group to participate. Based on their feedback,
careful revisions were made to the questionnaire to ensure that respondents could accu-
rately complete the formal survey. The formal survey was administered online, and we
established clear inclusion and exclusion criteria. The inclusion criteria required partic-
ipants to be college students who had used generative artificial intelligence technology
in their classroom or after-class studies and who volunteered to participate in this study.
The exclusion criteria included students with a history of serious psychological illness and
those unable to complete the questionnaire. To ensure the representativeness of the sample,
we surveyed different types of higher education institutions and selected students from
various majors and grades. Teachers distributed the questionnaire through course QQ
groups or presented it via course PowerPoint presentations to invite students to participate
in the survey. After obtaining consent from both the schools and the students, we explained
the purpose and process of the study to them and assured them that all data would be
processed anonymously to protect the privacy of the participants. During the data collec-
tion process, we obtained the necessary information through distributing questionnaires.
The questionnaire included questions about demographic variables as well as other topics
related to the research theme.
The online survey was conducted between 6 May 2024 and 6 June 2024. After data
collection and cleaning, a total of 752 questionnaires were received. Following a screening
process to eliminate invalid responses, 721 valid questionnaires were retained, resulting
in an effective response rate of 95%. In terms of demographics, the sample consisted of
442 males (61.3%) and 279 females (38.7%). A total of 258 respondents (35.78%) were
from Double First-Class universities, while 463 respondents (64.22%) were from regular
universities. Regarding grade level, 119 were freshmen (16.5%), 131 were sophomores
(18.2%), 175 were juniors (24.2%), and 296 were seniors (41.1%). The disciplines represented
included 141 students (19.6%) from the humanities, 202 students (28%) from social sciences,
190 students (26.4%) from natural sciences, and 188 students (26%) from engineering. In
terms of academic performance, 43.8% of the students (316) were ranked in the top 25% of
their class, 28.9% (208) were ranked between 26 and50%, 17.9% (129) were ranked between
51 and 75%, and 9.4% (68) were in the bottom quartile (see Table 3).

Table 3. Demographic data of the participants (N = 721).

Demographic Variables N %
Female 279 38.7
Gender
Male 442 61.3
Double First-Class universities 258 35.78
Types of universities
Regular universities 463 64.22
Freshmen 119 16.5
Sophomores 131 18.2
Grade
Juniors 175 24.2
Seniors 296 41.1
Humanities 141 19.6
Social Sciences 202 28
Disciplinary types
Natural Sciences 190 26.4
Engineering 188 26
Top 25% 316 43.8
Academic 26–50% 208 28.9
Performance 51–75% 129 17.9
Bottom Quartile 68 9.4
Behav. Sci. 2024, 14, 956 8 of 16

4. Research Results
4.1. Descriptive Statistics of Variables
Descriptive statistics were calculated by summing and averaging the measurement
items, yielding insights into the factors influencing university students’ intentions to use
generative AI technology, as presented in Table 4. A one-sample t-test was conducted with
a mean value of 3, representing the midpoint of the scale. A significant result indicates
that the variable’s data significantly differ from the midpoint. The results show that
all variables significantly differed from the midpoint, suggesting that there is notable
variation in university students’ willingness to use generative AI technology for learning
and autonomous study, which warrants further analysis and attention.

Table 4. Descriptive statistics of variables.

Standard One-Sample
Variable Mean Min Max
Deviation t-Test
Expectation Confirmation 3.47 1.052 1 5 88.539 ***
Perceived Usefulness 3.462 1.057 1 5 87.964 ***
Satisfaction 3.500 1.064 1 5 88.071 ***
Continuance Intention 3.472 1.089 1 5 85.64 0***
Willingness for Autonomous
3.276 1.069 1 5 87.299 ***
Learning
Digital Efficacy 3.257 1.058 1 5 87.697 ***
Positive Emotions 3.483 1.063 1 5 90.372 ***
Note: *** p < 0.001.

The mean values for each dimension (as shown in Table 4) indicate that, overall, univer-
sity students had a positive experience using generative AI for learning, with perceptions
across variables generally above average. Among these, satisfaction with use (M = 3.50,
SD = 1.064) scored the highest, followed by positive emotions (M = 3.483, SD = 1.063), both
of which were significantly above the average level. This indicates that students generally
have a high acceptance of generative AI technology, maintaining positive emotions such as
excitement, curiosity, and focus during its use.
Next in line were continuance intention (M = 3.472, SD = 1.089) and expectation confir-
mation (M = 3.470, SD = 1.052), suggesting that while students felt that the performance
of generative AI in their learning largely met their expectations, it did not significantly
exceed them, resulting in a moderate level of continuance intention. The mean scores
for perceived usefulness (M = 3.462, SD = 1.057), willingness for autonomous learning
(M = 3.276, SD = 1.069), and digital efficacy (M = 3.257, SD = 1.058) were relatively lower,
indicating that some students lack sufficient confidence in using digital technology, which
hinders their ability to fully leverage generative AI as a learning tool. This suggests that
students’ learning habits and methods have not significantly changed, and their perceived
usefulness of the technology remains modest.
The context in which students use generative AI technology may explain these find-
ings. The survey results indicate that the primary scenarios for using generative AI are
completing assignments (63.11%) and searching for necessary information (53.08%). In
contrast, scenarios that involve reviewing for exams and proactively learning cutting-edge
knowledge in their field accounted for only 40.36% and 41.19%, respectively. This suggests
that, for many students, generative AI is a highly practical tool that aids them in efficiently
completing academic tasks and gathering required materials. However, in learning contexts
that require deeper understanding and autonomous exploration, students may still prefer
to rely on traditional learning methods.

4.2. Path Analysis of the Structural Equation Model

This study employed Mplus 8.3 to construct the structural model. The model demon-
strated good fit indices, as indicated by the following values: χ²/df = 3.28, CFI = 0.920,
 strated good fit indices, as indicated by the following values: χ²/df = 3.28, CFI = 0.920, TLI
= 0.900, RMSEA = 0.056, and SRMR = 0.047. These results suggest that the model’s fit indi-
ces met the acceptable criteria, indicating a good fit to the data [41]. Table 5 presents the
results of the path analysis, while Figure 3 illustrates the final SEM model. The analysis
Behav. Sci. 2024, 14, 956 9 of 16
revealed that four out of the five hypotheses proposed in the original Expectation-Confir-
mation Model (ECM) were supported, with the exception of the hypothesis that perceived
that usefulness influences continuance intention. This finding suggests that the ECM is
TLI
still=applicable
0.900, RMSEAin the= context
0.056, and SRMR = 0.047.
of university Thesecontinued
students’ results suggest
use of that the model’s
generative fit
AI tech-
indices met the acceptable criteria,
nology, addressing Research Question 1. indicating a good fit to the data [41]. Table 5 presents
the results of the path analysis, while Figure 3 illustrates the final SEM model. The anal-
ysis
Table revealed that four
5. Path analysis out of
results ofhypotheses.
the five hypotheses proposed in the original Expectation-
Confirmation Model (ECM) were supported, with the exception of the hypothesis that
Independent Variable Dependent
perceived that Variable Estimate continuance
usefulness influences Standard intention.
Error p-Value
This Result
finding suggests that the
Expectation Confirma- ECM is still applicable in the context of university students’ continued use of generative AI
Perceived Usefulness 0.52 0.035 0.000 Supported
tion technology, addressing Research Question 1.
Expectation Confirma-
Digital
Table Efficacy
5. Path analysis results of0.235
hypotheses. 0.037 0.001 Supported
tion
Expectation Confirma-
Independent Variable Satisfaction
Dependent Variable 0.164
Estimate 0.042 Error
Standard 0.000
p-Value Supported
Result
tion
Expectation Confirmation Perceived Usefulness 0.52 0.035 0.000 Supported
Expectation Confirma-
Expectation Confirmation Positive
DigitalEmotions
Efficacy 0.561
0.235 0.034
0.037 0.000
0.001 Supported
Supported
tion
Expectation Confirmation Satisfaction 0.164 0.042 0.000 Supported
Digital Efficacy
Expectation Confirmation Satisfaction
Positive Emotions 0.216
0.561 0.034
0.034 0.000
0.000 Supported
Supported
Digital Efficacy
Perceived Usefulness Satisfaction
Continuance Intention 0.216
−0.035 0.034
0.039 0.000
0.362 NotSupported
Supported
Perceived Usefulness
Perceived Usefulness Continuance Intention
Satisfaction − 0.035
0.309 0.039
0.039 0.362
0.000 Not Supported
Supported
Perceived Usefulness Satisfaction 0.309 0.039 0.000 Supported
Positive Emotions
Positive Emotions Continuance Intention
Continuance Intention 0.148
0.148 0.038
0.038 0.000
0.000 Supported
Supported
Satisfaction
Satisfaction Continuance
Continuance Intention
Intention 0.143
0.143 0.044
0.044 0.001
0.001 Supported
Supported
Digital
DigitalEfficacy
Efficacy Continuance
Continuance Intention
Intention 0.131
0.131 0.04
0.04 0.000
0.000 Supported
Supported
Willingness
Willingness for for
Autonomous
Continuance Intention 0.314 0.035 0.000 Supported
Continuance Intention Autonomous Learning 0.314 0.035 0.000 Supported
Learning
Willingness for
Digital Efficacy Willingness for Autonomous 0.234 0.035 0.000 Supported
Digital Efficacy Autonomous Learning 0.234 0.035 0.000 Supported
Learning

Figure3.3.Hypothesis
Figure Hypothesistesting
testingresults
resultsof
ofthe
thefinal
finalmodel.
model.Note:
Note: ***
*** pp<<0.001.
0.001.

Additionally, all seven hypotheses related to the newly introduced variables, positive
emotions and digital efficacy, were supported, thereby addressing Research Question 2.
Finally, the hypothesis that continuance intention influences willingness for autonomous
learning was also validated, answering Research Question 3.

4.3. Mediation Effects

To further address Research Question 2, this study employed the bootstrap method to
examine the multiple mediation pathways from expectation confirmation to satisfaction
Behav. Sci. 2024, 14, 956 10 of 16

and continuance intention. Specifically, five significant mediation pathways were analyzed.
As shown in Table 6, in the pathway from expectation confirmation to satisfaction, the
mediation effect of perceived usefulness was 0.161 (p < 0.001), with a confidence interval
of (0.160, 0.256), while the mediation effect of digital efficacy was 0.051 (p < 0.001), with a
confidence interval of (0.111, 0.206).

Table 6. Mediation effect test results.

Dependent
Pathway Effect BootSE BootLLCI BootULCI
Variable
Mediation Effect 0.211 0.024 0.16 0.256
Expectation Confirmation → Perceived
0.161 0.024 0.111 0.206
Satisfaction Usefulness → Satisfaction
Expectation Confirmation → Digital Efficacy →
0.051 0.012 0.027 0.073
Satisfaction
Mediation Effect 0.137 0.023 0.091 0.184
Expectation Confirmation → Positive Emotions
0.083 0.009 0.039 0.127
→ Continuance Intention
Continuance
Expectation Confirmation → Digital Efficacy →
Intention 0.031 0.011 0.009 0.052
Continuance Intention
Expectation Confirmation → Satisfaction →
0.023 0.022 0.005 0.042
Continuance Intention
Note: BootSE, BootLLCI, and BootULCL refer to the standard error, lower limit, and upper limit of 95% confidence
interval of indirect effect obtained by bootstrap method, respectively.

In the pathway from expectation confirmation to continuance intention, the mediation

effect of positive emotions was 0.083 (p < 0.001), with a confidence interval of (0.039, 0.127),
the mediation effect of digital efficacy was 0.031 (p < 0.001), with a confidence interval of
(0.009, 0.052), and the mediation effect of satisfaction was 0.023 (p < 0.01), with a confidence
interval of (0.005, 0.042). None of the confidence intervals for these five pathways included
zero, indicating that the mediation effects are significant.

4.4. Moderation Effect Analysis

The analysis revealed significant differences in the impact of the continuance intention
to use generative AI technology on the willingness for autonomous learning across different
academic disciplines, with students from social sciences, natural sciences, and engineering
showing notable variance when compared to those from the humanities. This finding
addresses Research Question 5 (see Table 7). The model’s path coefficients were analyzed
to further understand these differences, revealing that the moderation effect was strongest
among social sciences students (0.518), followed by natural sciences students (0.442), and
was weakest among humanities students (0.314).

Table 7. Moderation effect test results.

Academic Background
Pathway
Social Natural Engineering
Humanities
Sciences Sciences Sciences
Continuance Intention →
Willingness for 0.314 *** 0.518 *** 0.442 *** 0.382 ***
Autonomous Learning
Digital Efficacy →
Willingness for 0.257 *** 0.264 *** 0.362 *** 0.176 ***
Autonomous Learning
Note: *** indicates p < 0.001.

These results suggest that students from different academic backgrounds may possess
distinct cognitive frameworks. Humanities disciplines, such as literature and history, often
Behav. Sci. 2024, 14, 956 11 of 16

emphasize understanding and interpreting human culture, thought, and history. The
nature of these fields may lead to a more conservative approach to adopting technology
for autonomous learning. In contrast, social sciences, including fields like economics, law,
and management, typically stress critical thinking and the analysis of social phenomena.
Research in social sciences often requires extensive data collection and analysis, areas where
generative AI technology can assist students in processing large datasets and extracting
valuable insights. As a result, these students recognize the potential of technology to
enhance learning efficiency and quality and are more inclined to explore how it can support
their autonomous learning. Students in natural and engineering sciences are accustomed
to using various software tools for experiment simulation and data analysis, and while
generative AI technology may not be their only tool for autonomous learning, it plays a
significant role in their educational toolkit.
Similarly, the analysis of the impact of digital efficacy on the willingness for au-
tonomous learning showed significant differences across disciplines when compared to
humanities students. The moderation effect was strongest among natural sciences students
(0.362), followed by social sciences (0.264) and humanities students (0.257), with the lowest
moderation effect observed among engineering students (0.176). This indicates that im-
provements in digital efficacy significantly enhance the willingness of humanities students
to use generative AI technology for autonomous learning.

5. Discussion and Implication

5.1. Discussion and Conclusions
This study extends the Expectation-Confirmation Model (ECM) to the context of uni-
versity students using generative AI technology for learning, incorporating three additional
variables: positive emotions, digital efficacy, and willingness for autonomous learning. By
constructing a theoretical model, this research provides valuable insights into the factors
influencing university students’ learning experiences with generative AI, thereby enriching
the application scenarios of the ECM and offering practical implications for enhancing
students’ willingness to engage in autonomous learning using AI technology.
Consistent with other research conclusions [42,43], expectation confirmation was
found to significantly predict not only perceived usefulness, as proposed in the original
ECM, but also the newly added variables, positive emotions and digital efficacy. The higher
the degree of expectation confirmation, the higher the levels of digital efficacy, perceived
usefulness, and positive emotions associated with using generative AI for learning. Ex-
pectation confirmation significantly influenced satisfaction through perceived usefulness
and digital efficacy, highlighting the mediating role of these factors between expectation
confirmation and satisfaction. Among them, digital efficacy emerged as the most critical
factor influencing satisfaction, corroborating findings by Bai [44] and Zheng [45], who
identified technological self-efficacy as a key predictor of technology use behavior and
experiences.
As expected, students’ continuance intention to use generative AI technology in the
future was significantly influenced by satisfaction, positive emotions, and digital efficacy.
Interestingly, while perceived usefulness significantly affected satisfaction, it was insuffi-
cient to drive long-term continuance intention to use AI for learning, diverging from the
original ECM. The impact of perceived usefulness on continuance intention was mediated
by satisfaction, underscoring the mediating role of satisfaction between expectation confir-
mation and continuance intention. From a motivational theory perspective, this finding
aligns with Wang and Sun’s research [46], which suggests that perceived usefulness, as
an external motivator, stems primarily from students’ pursuit of efficiency in learning
tools and is relatively unstable. Although it significantly impacts satisfaction, it may not
consistently motivate students to continue using the technology for learning.
Among satisfaction, positive emotions, and digital efficacy, positive emotions were
identified as the most critical factor influencing continuance intention. This may be because
generative AI, as a virtual tutor, creates a community of inquiry through interactions with
Behav. Sci. 2024, 14, 956 12 of 16

students. The more positive emotional investment students make in this community, the
more they perceive the social presence of the virtual tutor, which fosters and sustains
their connection, enhances learning engagement and enjoyment, and leads to meaningful
learning experiences [47]. In this context, the reflective and constructive processes of
knowledge acquisition through human–computer interaction bring students learning joy
and reduce anxiety [48], which, in turn, stimulates their continued use of the technology.
Finally, the study confirmed that continuance intention positively influences the
willingness for autonomous learning, with digital efficacy also showing a significant impact.
Moreover, the research identified significant differences in the willingness to use generative
AI for autonomous learning across students from different academic disciplines.

5.2. Implications for Practice

While AI technology has garnered significant attention from higher education insti-
tutions and society at large, with the educational sector eager to integrate it into teaching
practices, it is evident that merely providing resources and technological support is insuffi-
cient. To ensure that AI technology effectively enhances student learning and development,
it is essential to focus on factors such as students’ digital efficacy, positive emotions, and
willingness for autonomous learning and to develop corresponding strategies and mea-
sures.

5.2.1. Enhance AI Technology Promotion to Increase Expectation Confirmation

Given the importance of expectation confirmation for student satisfaction and contin-
uance intention, universities should strengthen the promotion of AI technology to improve
students’ expectation confirmation. Policymakers and university leaders need to implement
clear initiatives or strategies to address the growing importance of AI in the 21st century
and raise awareness among graduates of the key demands related to AI. As the “Innovation
Action Plan for Artificial Intelligence in Higher Education” emphasizes, a multi-stakeholder
collaborative education mechanism in the AI field should be established [49]. Based on this,
universities should create public service platforms for AI science popularization aimed at
youth and the general public, actively participating in outreach activities. They can also
invite industry leaders as guest speakers to discuss the impact of AI on future workplaces.
Additionally, universities can use various channels such as social media, campus seminars,
and presentations to showcase case studies and success stories of AI applications, helping
students form reasonable and positive expectations about AI technology. Teachers can
enhance students’ curiosity and expectations by demonstrating the relevance of generative
AI to real-world applications, such as using AI to create student profiles and academic
planning, making AI technology more relevant to their real needs. Furthermore, univer-
sities should adopt diverse incentive strategies to encourage students to engage deeply
with AI technology, as high levels of engagement can stimulate greater expectations and
enthusiasm for future intelligent technologies.

5.2.2. Foster Positive Emotions in AI Use to Achieve a “Human–Machine Symbiosis” Deep

Learning Model
As AI is increasingly integrated into education, the growing autonomy of technology
poses a threat to the development of human agency. To promote more autonomous and
deep learning among students in the age of AI, it is crucial to enhance their intrinsic mo-
tivation rather than relying solely on extrinsic motivators like perceived usefulness [50].
There is also a need to be cautious of the phenomenon of reverse alienation caused by
excessive reliance on intelligent technology. Ouyang et al. argued that the output of gener-
ative AI may not always align with individual intentions, and a forward-looking learning
approach involves questioning and adjusting the content through human feedback [51].
Focusing on students’ positive emotions when using generative AI for learning can stim-
ulate curiosity and the desire to explore, leading to higher-order thinking processes and
interactions with disciplinary knowledge that are internalized as a learning interest and
Behav. Sci. 2024, 14, 956 13 of 16

eventually transformed into a higher level of intellectual pursuit [52]. As Sun pointed out,
individuals leverage their subjectivity to positively adapt AI technology, using its functions
to meet their learning needs, thereby enabling deep learning to occur [53]. Universities
should create a supportive digital learning environment where students feel confident
and joyful when using generative AI, enhancing their engagement with learning under
the guidance of teachers, peers, and AI agents. For instance, building AI-based learn-
ing communities can foster interaction and the sharing of experiences among teachers,
peers, and students, thus strengthening students’ motivation to learn. Teachers should
adhere to a student-centered educational philosophy, focusing on developing students’
higher-order thinking and enhancing their autonomy in learning. Moreover, universities
should organize reflective activities on technology use, helping students to understand
the potential negative impacts of over-reliance on technology, such as information silos,
reverse domestication of technology, and risks of losing control over technology, thereby
cultivating a proper understanding of technology and ethical behavior in its use. This
approach will encourage students to maintain self-awareness, retain more autonomy, and
develop a “human–machine symbiosis” learning model that leads to more efficient and
deeper learning experiences.

5.2.3. Cultivate Digital Efficacy to Empower More Effective Personalized Learning

Personalized learning centers on the learner, focusing on their individual needs and
future development. With the advent of AI technology, all students have access to and
can use it, but not all students can adapt to this technology. That is, not every student
can transition from passively receiving information to actively controlling technology and
resources. Personalized learning can only be achieved when learners actively control
technology and resources, flexibly adjusting their learning methods to meet their indepen-
dent learning needs. Therefore, in the era of intelligent technologies, enhancing digital
self-efficacy provides a safeguard for the development of personalized learning among
university students. As Trimmel and Bachmann mentioned, students proficient in multiple
technologies tend to have higher levels of motivation in learning [54,55], and university
students with higher digital self-efficacy are more likely to use technology to devise person-
alized learning strategies [56]. This may be because students with higher digital self-efficacy
are better at recognizing and leveraging the possibilities offered by technology and can
choose technological tools based on their learning contexts, values, and intentions [57].
Universities should provide ample technical resources and support to enhance students’
digital self-efficacy by focusing on the construction of smart campuses, platform operations,
and creating a conducive cultural environment. They should establish clear AI usage
policies and user guides to reduce students’ anxiety about using AI technologies, help
students self-regulate and manage their own learning processes in informal learning spaces,
and enhance their adaptability to personalized learning. In formal learning environments,
teachers should pay greater attention to students’ emotional experiences, motivational
beliefs, learning styles, and other non-cognitive factors. They can leverage AI technology to
create online resource libraries and intelligent recommendation systems [58], recommend-
ing the most appropriate resources based on students’ learning behaviors, interests, and
needs. Furthermore, students should be encouraged to participate in the creation and use
of resources, fostering an active learning community that enables personalized learning.
Given the differences in how students from various academic disciplines use AI tech-
nology, universities should pay particular attention to the digital self-efficacy of humanities
and social sciences students. These students need to understand the core functions and
operational logic of AI technology, clarify the role of technology in their learning, and
be encouraged to experiment with generative AI technologies to support their learning—
without completely replacing traditional learning methods. As Miyagawa stated, in a world
increasingly assisted by computers, traditional humanities skills such as idea creation and
expression become even more valuable [59].
Behav. Sci. 2024, 14, 956 14 of 16

5.3. Limitations and Directions for Future Research

When interpreting the results of this study, several limitations should be taken into
account. First, the data used in this research were collected through self-reported surveys,
which may not fully capture the entire breadth and complexity of the constructs being
measured. We encourage future research to collect qualitative data to cross-validate the
findings. Second, our data are cross-sectional, which limits our ability to establish causal
and reciprocal relationships between variables. Future studies may need to collect lon-
gitudinal data to provide a more dynamic picture of how positive emotions and digital
efficacy interact to promote students’ willingness to engage in autonomous learning using
AI technology. Third, the sample in this study was drawn from universities in mainland
China, which may restrict the cross-cultural generalizability of the findings.

Author Contributions: Conceptualization, L.W. and W.L.; methodology, W.L.; validation, L.W.; inves-
tigation, L.W.; writing—original draft preparation, L.W.; writing—review and editing, W.L.; Funding
acquisition, W.L. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the National Education Planning Topics General Project
“Research on deep learning for undergraduates in the “Top Notch Program” from the perspective of
artificial intelligence empowerment: connotation reconstruction, influence mechanism and promotion
strategy” (ID:BIA240122) and Postgraduate Research & Practice Innovation Program of Jiangsu
Province (ID:KYCX24_0098). The APC was funded by these projects.
Institutional Review Board Statement: This study collected data from human participants in line
with Article 32 in “Measures for the Ethical Review of Life Sciences and Medical Research Involving
Humans”; it is exempt from needing formal ethical approval.
Informed Consent Statement: The necessary informed consent form was set at the beginning of the
questionnaire, and informed consent was obtained from all subjects involved in the study.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Moscardini, A.O.; Strachan, R.; Vlasova, T. The role of universities in modern society. Stud. High. Educ. 2020, 47, 812–830.
[CrossRef]
2. Cotton, D.R.E.; Cotton, P.A.; Shipway, J.R. Chatting and Cheating: Ensuring Academic Integrity in the era of ChatGPT. Innov.
Educ. Teach. Int. 2023, 61, 228–239. [CrossRef]
3. du Boulay, B.; Mitrović, A.; Yacef, K. Handbook of Artificial Intelligence in Education; Edward Elgar Publishing: Cheltenham, UK,
2023.
4. Iku-Silan, A.; Hwang, G.-J.; Chen, C.-H. Decision-Guided Chatbots and Cognitive Styles in Interdisciplinary Learning. Comput.
Educ. 2023, 201, 104812. [CrossRef]
5. Ryan, T.; French, S.; Kennedy, G. Beyond the Iron Triangle: Improving the Quality of Teaching and Learning at Scale. Stud. High.
Educ. 2019, 46, 1383–1394. [CrossRef]
6. Essel, H.B.; Vlachopoulos, D.; Tachie-Menson, A.; Johnson, E.E.; Baah, P.K. The Impact of a Virtual Teaching Assistant (Chatbot)
on Students’ Learning in Ghanaian Higher Education. Int. J. Educ. Technol. High. Educ. 2022, 19, 57. [CrossRef]
7. Atlas, S. ChatGPT for Higher Education and Professional Development: A Guide to Conversational AI. 2023. Available online:
https://digitalcommons.uri.edu/cba_facpubs/548 (accessed on 6 August 2024).
8. Berg, C. The Case for Generative AI in Scholarly Practice. 2023. Available online: https://papers.ssrn.com/sol3/papers.cfm?
abstract_id=4407587 (accessed on 6 August 2024).
9. Popenici, S.A.D.; Kerr, S. Exploring the Impact of Artificial Intelligence on Teaching and Learning in Higher Education. Res. Pract.
Technol. Enhanc. Learn. 2017, 12, 22. [CrossRef] [PubMed]
10. Bhattacherjee, A. Understanding Information Systems Continuance: An Expectation Confirmation Model. MIS Q. 2001, 25,
351–370. [CrossRef]
11. Kim, S.H.; Bae, J.H.; Jeon, H.M. Continuous Intention on Accommodation Apps: Integrated Value-based Adoption and Expecta-
tion Confirmation Model Analysis. Sustainability 2019, 11, 1578. [CrossRef]
12. Dai, H.M.; Teo, T.; Rappa, N.A.; Huang, F. Explaining Chinese University Students’ Continuance Learning Intention in the MOOC
Setting: A Modified Expectation Confirmation Model Perspective. Comput. Educ. 2020, 150, 103850. [CrossRef]
Behav. Sci. 2024, 14, 956 15 of 16

13. Gupta, A.; Dhiman, N.; Yousaf, A.; Thurasamy, R. Social Comparison and Continuance Intention of Smart Fitness Wearables: An
Extended Expectation Confirmation Theory Perspective. Behav. Inf. Technol. 2021, 40, 1341–1354. [CrossRef]
14. Bhattacherjee, A.; Premkumar, G. Understanding Changes in Belief and Attitude toward Information Technology Usage: A
Theoretical Model and Longitudinal Test. MIS Q. 2004, 28, 229–254. [CrossRef]
15. Davis, F.D. Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology. MIS Q. 1989, 13,
319–340. [CrossRef]
16. Kuo, Y.-C.; Walker, A.E.; Schroder, K.E.E.; Belland, B.R. Interaction, Internet Self-efficacy, and Self-regulated Learning as Predictors
of Student Satisfaction in Online Education Courses. Internet High. Educ. 2014, 20, 35–50. [CrossRef]
17. Hong, J.; Lee, O.-K.D.; Suh, W. A Study of the Continuous Usage Intention of Social Software in the Context of Instant Messaging.
Online Inf. Rev. 2013, 37, 692–710. [CrossRef]
18. Blondeel, E.; Everaert, P.C.; Opdecam, E. Does Practice Make Perfect? The Effect of Online Formative Assessments on Students’ Self-
Efficacy and Test Anxiety; SSRN: Rochester, NY, USA, 2023; Volume 12, Available online: https://papers.ssrn.com/abstract=4357056
(accessed on 17 August 2024).
19. Paetsch, J.; Drechsel, B. Factors influencing pre-service teachers’ intention to use digital learning materials: A study conducted
during the COVID-19 pandemic in Germany. Front. Psychol. 2021, 12, 733830. [CrossRef]
20. Bandura, A. Self-efficacy: Toward a Unifying Theory of Behavioral Change. Psychol. Rev. 1977, 84, 191–215. [CrossRef]
21. Biggs, J.B. Teaching for Quality Learning at University: What the Student Does; McGraw-Hill Education: London, UK, 2011.
22. Lee, Y.H.; Hsieh, Y.C.; Hsu, C.N. Adding Innovation Diffusion Theory to the Technology Acceptance Model: Supporting
Employees’ Intentions to Use E-Learning Systems. Educ. Technol. Soc. 2011, 14, 124–137.
23. Joo, Y.J.; Lim, K.Y.; Kim, J. Locus of Control, Self-efficacy, and Task Value as Predictors of Learning Outcome in an Online
University Context. Comput. Educ. 2013, 62, 149–158. [CrossRef]
24. Altalhi, M. Toward a Model for Acceptance of MOOCs in Higher Education: The Modified UTAUT Model for Saudi Arabia. Educ.
Inf. Technol. 2021, 26, 1589–1605. [CrossRef]
25. Zadorozhnyy, A.; Ju, S.; Lee, J. Informal Digital Learning of English and Willingness to Communicate in a Second Language:
Self-efficacy Beliefs as a Mediator. Comput. Assist. Lang. Learn. 2023, 5, 1–21. [CrossRef]
26. Chik, A. Digital Gaming and Language Learning: Autonomy and Community. Lang. Learn. Technol. 2014, 18, 85–100.
27. Jacobs, G.E.; Castek, J. Digital Problem Solving: The Literacies of Navigating Life in the Digital Age. J. Adolesc. Adult Lit. 2018, 61,
681–685. [CrossRef]
28. Chan, C.K.Y.; Hu, W. Students’ Voices on Generative AI: Perceptions, Benefits, and Challenges in Higher Education. Int. J. Educ.
Technol. High. Educ. 2023, 20, 43. [CrossRef]
29. Ding, Y. Looking Forward: The Role of Hope in Information System Continuance. Comput. Hum. Behav. 2018, 91, 127–137.
[CrossRef]
30. Oliver, R.L. A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions. J. Mark. Res. 1980, 17, 460–469.
[CrossRef]
31. Bearman, M.; Ajjawi, R. Learning to Work with the Black Box: Pedagogy for a World with Artificial Intelligence. Br. J. Educ.
Technol. 2023, 54, 1160–1173. [CrossRef]
32. Getenet, S.; Cantle, R.; Redmond, P.; Albion, P. Students’ Digital Technology Attitude, Literacy and Self-efficacy and Their Effect
on Online Learning Engagement. Int. J. Educ. Technol. High. Educ. 2024, 21, 3. [CrossRef]
33. Liu, G.; Ma, C. Measuring EFL Learners’ Use of ChatGPT in Informal Digital Learning of English Based on the Technology
Acceptance Model. Innov. Lang. Learn. Teach. 2023, 18, 125–138. [CrossRef]
34. Darvishi, A.; Khosravi, H.; Sadiq, S.; Gašević, D.; Siemens, G. Impact of AI Assistance on Student Agency. Comput. Educ. 2024,
210, 104967. [CrossRef]
35. Zhang, C.; Meng, Y.; Ma, X. Artificial Intelligence in EFL Speaking: Impact on Enjoyment, Anxiety, and Willingness to Communi-
cate. System 2024, 121, 103259. [CrossRef]
36. Kamita, T.; Ito, T.; Matsumoto, A.; Munakata, T.; Inoue, T. A Chatbot System for Mental Healthcare Based on SAT Counseling
Method. Mob. Inf. Syst. 2019, 2019, 9517321. [CrossRef]
37. Owens, J.; Lilly, F. The Influence of Academic Discipline, Race, and Gender on Web-use Skills among Graduate-level Students. J.
Comput. High. Educ. 2017, 29, 286–308. [CrossRef]
38. Venkatesh, V.; Morris, M.G.; Davis, G.B.; Davis, F.D. User Acceptance of Information Technology: Toward a Unified View. MIS Q.
2003, 27, 425–478. [CrossRef]
39. Compeau, D.R.; Higgins, C.A. Computer Self-Efficacy: Development of a Measure and Initial Test. MIS Q. 1995, 19, 189–211.
[CrossRef]
40. Özkan, S.; Koseler, R. Multi-dimensional students’ evaluation of e-learning systems in the higher education context: An empirical
investigation. Comput. Educ. 2009, 53, 1285–1296. [CrossRef]
41. Hu, L.; Bentler, P.M. Cutoff Criteria for Fit Indexes in Covariance Structure Analysis: Conventional Criteria Versus New
Alternatives. Struct. Equ. Model. Multidiscip. J. 1999, 6, 1–55. [CrossRef]
42. Ni, A.; Cheung, A. Understanding Secondary Students’ Continuance Intention to Adopt AI-powered Intelligent Tutoring System
for English Learning. Educ. Inf. Technol. 2023, 28, 3191–3216. [CrossRef]
Behav. Sci. 2024, 14, 956 16 of 16

43. Chai, C.S.; Lin, P.Y.; Jong, M.S.Y.; Dai, Y.; Chiu, T.K.; Qin, J. Perceptions of and Behavioral Intentions towards Learning Artificial
Intelligence in Primary School Students. Educ. Technol. Soc. 2021, 24, 89–101.
44. Bai, B.; Wang, J.; Chai, C.S. Understanding Hong Kong Primary School English Teachers’ Continuance Intention to Teach with
ICT. Comput. Assist. Lang. Learn. 2021, 34, 528–551. [CrossRef]
45. Zheng, J.; Li, S. What Drives Students’ Intention to Use Tablet Computers? An Extended Technology Acceptance Model. Int. J.
Educ. Res. 2020, 12, 101612. [CrossRef]
46. Wang, Q.; Sun, J. Research on the Influencing Factors of Users’ Continuous Contribution Behavior in Network Group Collabora-
tion. Libr. J. 2018, 37, 12–20. (In Chinese)
47. Liu, Q.; Pang, H.; Wang, X.H.; Zhu, J. Artificial Intelligence Support for Language Study: How Anthropomorphic Technology
Affects Study. J. Electrochem. Educ. Res. 2022, 10, 102–109. (In Chinese) [CrossRef]
48. Tai, T.Y.; Chen, H.J. The Impact of Google Assistant on Adolescent EFL Learners’ Willingness to Communicate. Interact. Learn.
Environ. 2023, 31, 1485–1502. [CrossRef]
49. The Ministry of Education. Notice on Issuing the Action Plan on Artificial Intelligence Innovation in Universities [EB/OL].
(2018-12-31). Available online: https://www.gov.cn/zhengce/zhengceku/2018-12/31/content_5443346.htm (accessed on 24
August 2024).
50. Dunn, T.J.; Kennedy, M. Technology Enhanced Learning in Higher Education: Motivations, Engagement and Academic Achieve-
ment. Comput. Educ. 2019, 137, 104–113. [CrossRef]
51. Ouyang, L.; Wu, J.; Jiang, X.; Almeida, D.; Wainwright, C.; Mishkin, P.; Zhang, C.; Agarwal, S.; Slama, K.; Ray, A.; et al. Training
Language Models to Follow Instructions with Human Feedback. Adv. Neural Inf. Process. Syst. 2022, 35, 27730–27744.
52. Lv, L.; Gong, F. Knowledge Purport: The Deep Driving Force Affecting the Quality of Undergraduate Learning Experience in
Top Universities: Based on the SERU Survey Data of Eight Universities in China and the United States (2017–2018). High. Educ.
Jiangsu 2019, 9, 57–65. (In Chinese) [CrossRef]
53. Sun, T.; Linzi, T. Game and Construction of Human-Technology Relationship in Artificial Intelligence Education: From Reverse
Domestication to Technical Mediation. Open Educ. Res. 2021, 27, 37–43. (In Chinese) [CrossRef]
54. Trimmel, M.; Bachmann, J. Cognitive, Social, Motivational, and Health Aspects of Students in Laptop Classrooms. J. Comput.
Assist. Learn. 2004, 20, 151–158. [CrossRef]
55. Li, B.; Bonk, C.J.; Wang, C.; Kou, X. Reconceptualizing Self-Directed Learning in the Era of Generative AI: An Exploratory
Analysis of Language Learning. IEEE Trans. Learn. Technol. 2024, 17, 15–29. [CrossRef]
56. Glynn, S.M.; Brickman, P.; Armstrong, N.; Taasoobshirazi, G. Science Motivation Questionnaire II: Validation with Science Majors
and Nonscience Majors. J. Res. Sci. Teach. 2011, 48, 1159–1176. [CrossRef]
57. Dings, R. Meaningful Affordances. Synthese 2020, 199, 1855–1875. [CrossRef]
58. Huang, A.Y.Q.; Lu, O.H.T.; Yang, S.J.H. Effects of Artificial Intelligence–Enabled Personalized Recommendations on Learners’
Learning Engagement, Motivation, and Outcomes in a Flipped Classroom. Comput. Educ. 2023, 194, 104684. [CrossRef]
59. Miyagawa, S. Ethics Education Must Accompany AI Development. Times Highigher Education. 2019. Available online: https:
//www.timeshighereducation.com/cn/news/ethics-education-must-accompany-ai-development-says-mit-expert (accessed on
18 September 2024).

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.