Emerging Educational
Technologies
Emerging
Educational Technologies, 2026, 1, 260102 https://doi.org/10.65553/EET.260102
Foreign
Language Anxiety, Enjoyment, and Classroom Psychological Resilience as
Predictors of English Majors’ Behavioral Intention to Use
Generative AI for English Learning
Puyang
Suna
aUniversity of Wales Trinity Saint
David, Wales, United Kingdom
Submission date: May 11, 2026; Revision date: October 20, 2026; Online: November 20, 2026
Abstract
The rapid development of
generative AI (GenAI) has significantly changed language learning. Existing
studies on learners’ acceptance of GenAI-assisted technologies mostly rely on
cognitive-based models, focusing on perceived usefulness and ease of use, while
neglecting learners’ emotional experiences and psychological resources. Drawing
on Control-Value Theory, this study examines how Foreign Language Enjoyment
(FLE), Foreign Language Anxiety (FLA), and classroom Psychological Resilience
(RES) predict English majors’ Behavioral Intention (BI) to use GenAI.
Questionnaire data from 433 Chinese undergraduate English majors were analyzed
using SPSS 27 and AMOS 26 through correlation, CFA, and SEM. Results showed
that FLE, FLA, and RES significantly predicted BI, with language-related
emotions stronger than RES. These findings highlight the need to incorporate
affective factors into GenAI-supported language learning design.
Keywords: Behavioral Intention;
English Majors; Foreign Language Anxiety; Foreign Language Enjoyment;
Generative Artificial Intelligence; Psychological Resilience
Introduction
GenAI
has emerged as one of the most transformative technological developments. In
comparison to previous educational technology applications, GenAI systems have
one advantage in that they can generate coherent texts, carrying on a
conversation, and responding intelligently to user feedback. These affordances
have created a lot of curiosity in the way they can be used in learning
languages especially in assisting in the development of academic writing (Khanh-Quynh &
Cong-Lem, 2026; Watson Todd, 2026).
Recent research indicates that GenAI has potential to transform the nature of
interaction of learners with language input and respondents by dissolving the
distinction between implementation devices and interactive learning associates
(Yang &
Zhao, 2024; Zhang & Derakhshan, 2025).
Alongside these technological innovations, scholars have been motivated to
learners accepting and using GenAI-assisted tools in more studies. A great part
of this has been based on the comprehensible technology approval models
including the Technology Acceptance Model (TAM) and the Unified Theory of
Acceptance and Use of Technology (UTAUT) that highlights mental assessment of
technology by the learners, like the feeling of usefulness, perceived ease of
use and social influence (Davis, 1989; Venkatesh et al.,
2003). Such models have found extensive
application in the educational scenarios, and they have proven to explain
Behavior Intention (BI) with high predictability rates. Nevertheless, there is
an emerging level of evidence that purely cognitive explanations can be
inadequate in explaining technology adoption in other realms of learning that
involve high emotionality. The process of language learning is generally
acknowledged to be emotionally defective, barring anxiety, enjoyment,
uncertainty, and self-evaluation (MacIntyre, 2017).
Consequently, scholars have started to doubt the capacity of the models that
pay minimal attention to the affective aspects to capture the interest of
learners towards technologies like GenAI that directly relates with language
production and identity of learners as language users.
Research
studies on SLA have continued to establish that learning of language is not
just a cognitive process but a process that is intensively influenced by the
emotional experience of the learners. Emotions have a role in the motivation of
the learner, their engagement, self-regulation and decisions related to
learning, influencing not only the results of learning but also the desire of
the learners to engage in the learning process (Dewaele, 2019; MacIntyre, 2017). In
this sense, affect does not play a peripheral role in language learning but a
key part of the classroom experience of learners. Control Value Theory (CVT) of
Achievement Emotions offers an in-depth insight into the role of the appraisal
of control and value by learners in developing a certain emotion and how
certain emotions impact the learning behavior (Pekrun, 2006; Pekrun et al.,
2011). CVT further asserts that the emotions of
enjoyment and anxiety will be influenced by how learners perceive that they can
advocate the outcome of the learning process and the value they attach to the
learning task. Such feelings are then the key determinants of the behavioral
options, persistence and engagement of the learners. In this context, the
Foreign Language Anxiety (FLA) and Foreign Language Enjoyment (FLE) have been
named as two especially vulnerable variables in the classroom. FLA is linked to
the fear of negative judgment, fear of communication, and pressure to perform (Horwitz et al.,
1986), whereas FLE is connected to positive
involvement, interest, and accomplishment among the learners (Dewaele and
MacIntyre, 2014). Notably, the recent studies point
out that these feelings are not mutually exclusive but partially independent as
learners can feel fun and worried about language learning situations at the
same time (Dewaele et al., 2018; Teimouri et al.,
2019). This rich emotional impact has significant
consequences to explain how learners react to GenAI mediated learning
conditions.
In
addition to emotional scenes, psychological resources of learners can also be
significant in influencing their reaction toward difficult learning conditions.
Psychological Resilience (RES) that is often defined as the capacity to bounce
back and survive stress and adapt to adversity has been acknowledged as one of
the main causes that underlies long-lasting engagement and academic persistence
(Luthar et
al., 2000; Masten, 2014).
Research using the educational setting has attributed the phenomenon of RES to
adaptive coping patterns, emotional control, and endurance amidst adversity. In
classes building language skills, RES can be especially significant as the very
process of learning a language requires a lot of mistakes, communicative
challenges, and critical assessment. Students who have more advanced degrees of
RES might be closer to overcoming these difficulties and be able to stay
engaged in the long term. Of more salient position in the context of emerging
technologies added as GenAI may be, RES can be. The problem with GenAI tools is
that they involve doubt, trial and error, and trial and error to generate an
accurate output, which means that the learner has to endure failure and
continue with experimentation. Even though its relevance is high, the research
on GenAI-assisted language learning is relatively little in terms of its
research on RES. The role of broader psychological resources is not yet
well-researched whilst most of the existing studies add to the affective or
cognitive variables.
Literature review
Theoretical
Foundation of Emotions and Learning-Related Behaviors
The
present study is grounded in CVT of achievement emotions (Pekrun, 2006;
Pekrun et al., 2011),
which provides a comprehensive framework for understanding how learners’
emotional experiences influence their learning-related behaviors. According to
CVT, achievement emotions arise from learners’ appraisals of control over
learning activities and the value they attach to those activities. These
emotions, in turn, shape motivation, engagement, persistence, and behavioral
intentions (Pekrun et al., 2011; Tze et al., 2016).
Within this theoretical framework, positive emotions such as enjoyment are
expected to facilitate approach-oriented behaviors, whereas negative emotions
such as anxiety may constrain participation or alter learners’ behavioral
choices. Importantly, CVT conceptualizes emotions as dynamic and
context-dependent rather than fixed traits, meaning that they fluctuate
according to situational characteristics and task demands (Pekrun et al., 2011).
Building
on CVT, the present study focuses on FLE and FLA as two central achievement
emotions in EFL contexts. FLE reflects learners’ positive engagement, interest,
and sense of accomplishment in language learning, while FLA captures
apprehension, fear of negative evaluation, and performance-related tension.
From a CVT perspective, these emotions are not merely outcomes of learning but
key antecedents of learners’ BI. When learners perceive greater control and
value in language learning activities, they are more likely to experience
enjoyment and demonstrate stronger willingness to engage in related behaviors.
Conversely, anxiety may reduce learners perceived control and influence their
willingness to participate, thereby shaping their technology-related decisions.
In
addition to achievement emotions, psychological resources such as classroom RES
may further influence learning-related behaviors. Although RES is not
classified as an achievement emotion within CVT, it represents learners’
adaptive capacity to cope with stress, recover from setbacks, and remain
engaged in challenging learning situations. In emotionally demanding contexts,
resilience may support sustained engagement and help learners regulate negative
emotions, thereby indirectly contributing to behavioral intentions. In this
sense, integrating RES into a CVT-informed framework enables a more
comprehensive understanding of how both emotional experiences and psychological
resources jointly shape learners’ behavioral choices.
This
theoretical framework is particularly relevant in technology-mediated language
learning environments such as those supported by GenAI. In GenAI-facilitated
contexts, learners are required to operate in uncertain environments, make
independent decisions, and engage in emotionally intensive communication with
intelligent systems (Zong & Yang, 2025). Such situational
characteristics may rapidly influence learners’ perceptions of control and
value, thereby shaping their emotional experiences and subsequent behavioral
intentions. Although CVT-based studies have examined the impact of emotions on
academic success and engagement, relatively few have extended this framework to
technology-related behavioral intentions in language learning contexts (Botes et al., 2020;
Putwain et al.,
2018). Given the rapid expansion of GenAI in
education, applying CVT to examine how FLE, FLA, and RES predict learners’ BI
represents a timely and theoretically grounded research direction.
Foreign
Language Anxiety in EFL Classrooms
FLA has also been thoroughly investigated as
an eminent negative effect in EFL classrooms and understood generally as a
contingent figuration of anxiety related to language learning situations (Horwitz et al., 1986; MacIntyre
and Gardner, 1991). Contrary to trait
anxiety, FLA is a resultant effect of learners’ perceptions of communicative
demands, their performance expectations and the evaluative pressure that is
specific to foreign language classrooms. Controlled-value-wise, it is possible
to see FLA as a state that can be caused by the perceived shortage of control
in learning activities among learners, which can deter approach-oriented
behaviors and decrease their intention to use GenAI-assisted learning tools.
An extensive body of evidence has written
about negative implications that FLA has on language learning experiences of
learners. Increased anxiety rates are always associated with a lack of
participation in the classroom, a decrease in the desire to communicate, and a
loss of self-confidence (Gregersen & Horwitz, 2002; MacIntyre et al., 1998; Teimouri et al., 2019). The presence of anxiety can also disrupt cognitive
processing because it pulls attentional resources, which have a detrimental
impact on the learners in terms of processing the input and accessing
linguistic information effectively (Eysenck et al., 2007; MacIntyre, 2017).
Recent studies have also emphasized the situational sensitivity of FLA showing
that levels of anxiety can change with task modality and interactional
requirements and mediation by technology (Teimouri et al., 2019). In digitally mediated spaces, anxiety can be
re-structured and not removed, with learners haggling new modes of performance
demands and technological ability. The insight into the functioning of FLA in
GenAI-assisted situations then demand a transition in thinking beyond the
deficit-focused conceptualizations and the investigation of the role of anxiety
in driving compensatory learning responses.
In addition to the conventional learning
outcomes, the recent studies have started to examine the way in which FLA
affects the response of the learners to the educational technologies. Learners
who are anxious can find unknown or complicated technologies threatening, which
results in the attempt to avoid, or resist (Howard
et al., 2016). Simultaneously, certain
online tools might offer less face-threatening conditions by lessening
immediate social assessment, which might potentially be an attractive offer to
learners who are exposed to high classroom anxiety levels (Lee,
2020; Lin and Warschauer, 2015).
Although the role of affective factors in
technology-driven language acquisition is receiving growing interest, there is
limited empirical evidence to study how directly FLA is linked to the use of BI
by learners to use GenAI. Since GenAI tools can change the dynamics of
interaction and decrease evaluative pressure, the study of the ability of FLA
to predict the willingness of learners to use it is theoretically and
methodologically significant.
Foreign Language Enjoyment
and Positive Emotional Experiences
As
a reaction to the previous concern on anxiety, new studies in the field of SLA
have begun to pay more attention to the importance of positive emotions,
especially FLE to influence the classroom experiences of learners (Dewaele and
MacIntyre, 2014; MacIntyre et al. 2019).
This trend is based on the impact of positive psychology (PP) that emphasizes
not only the disadvantages of learners but also their strengths and well-being.
FLE
is usually linked to the existence of positive classroom experiences,
significant interaction, and a feeling of progress and competence in the
learners. The empirical research indicated the positive correlation of FLE with
motivation, engagement, and persistence of learners to language learning tasks
among various elements (Dewaele et al., 2018; Ogas-Baldwin et al., 2019;
Saito et al., 2018).
Fun usually crops up when learners consider learning tasks challenging but not
insurmountable and when they are provided enough instructional and social help.
In addition, FLE has been also associated with the perception of autonomy and
agency of the learners in the learning process (Ogas-Baldwin et al., 2019).
Such technology-enhanced environments in which learners can dictate the pace of
learning, learning material, and interaction can thus be especially supportive
of enjoyment (Lin et al., 2025; Wu & Yang, 2025).
The presence of GenAI tools, which imply flexible and learner-focused
interaction, can enhance pleasure as learners are able to try out the language
without social ramifications. This underscores the possibility of FLE being one
of the crucial affective elements underlying voluntary technology adoption.
Notably,
it has been evidenced that FLE and FLA are not the extreme ends of the same
emotional range. Students can feel both pleasure and stress in the same
classroom and such experience can be based on the kinds of tasks demanded and
situational forces (Boudreau et al., 2018; Dewaele & MacIntyre, 2016).
This difference is in line with CVT and the significance of analyzing positive
and negative emotions as independent predictors of learning related behaviors.
In
reference to technology-supported language learning, the positive, enjoyable
emotions have been determined to aid the openness of learners to innovation and
experimentation (Scherer & Teo, 2019). Among learners
with a greater degree of enjoyment, there are more chances of perceiving the
technological challenges as growth avenue not as a threat. Nonetheless, it is
important to note that most attention to GenAI applications is paid, yet the BI
alteration, especially by FLE, to make learners apply GenAI in language
learning is poorly investigated.
Classroom Psychological Resilience as Learning
Resource
Psychological
resources of learners are also essential in determining the ways of responding
to challenges in foreign language classrooms in addition to their emotional
experiences. Of these resources, the growing popularity of RES in educational
psychology is as an important factor that contributes to favor of adaptive
learning behaviors in the conditions of stress and uncertainty (Masten, 2014;
Martin & Marsh,
2006).
The ability of the
individuals to adjust to adversity, recuperate in the aftermath of an
unfortunate turn of events, and be in efficient operation or reemerge back into
efficient functions in a difficult circumstance characterizes the concept of
RES (Luthar
et al., 2000; Masten, 2014).
The conceptualization of RES in the educational setting is becoming more
context-dependent and dynamic. It shifts into a conceptualized and variable
resource of the learner instead of a given personality attribute, determined by
interactions of the learners with the instructional setting and social forms of
support (Martin,
2013). This view can be applied especially in
the classroom-based research where resilience of learners can vary depending on
tasks, courses, along with learning needs.
In the foreign
language classrooms, the phenomenon of RES is particularly acute because of the
peculiarities of the language learning process: numerous mistakes, risks of
communication and evaluations. Empirical research has indicated that the
learners with increased academic or classroom RES are more apt to continue in
learning activities, managing negative feelings, and being engaged despite
challenges (Kim
& Kim, 2015; Martin & Marsh, 2006).
It has also been discovered that RES is responsive to emotional experiences and
may buffer the harmful quality of anxiety, as well as inhibits the harmful
impact of anxiety by facilitating the maintenance of engagement and adaptive
responses towards anxiety.
Notably, it has
become more assumed that RES is conceptualized as flexible and vulnerable to
instructional practices and not fixed as an individual difference (Martin, 2013).
Emotionally supportive classroom environments with constructive feedback and a
chance to experiment safely can be useful because they can help build
resilience in the long-term. GenAI, in GenAI-assisted learning scenarios, can
be used as a learning device or used as a resource to facilitate adaptive
coping in the aspect of GenAI in terms of perceived failure costs and iteration
of improvement.
Although
accumulated evidence on the concept of RES in a learning context has been
documented in learning environments, comparatively few studies have included
classroom RES in the models that address technology-related aspects of learning
behaviors. Within the framework of fast-changing educational technologies,
including GenAI, RES can potentially change the propensity of students to use
the new tools, accept the challenges in the beginning, and continue learning
via trial and error (Bond et al., 2018; Howard & Gigliotti, 2016).
This implies that classroom RES can provide a significant predictor of learners
BI to make use of GenAI to learn the language.
Behavioral Intention to use Generative AI to
learn English
BI is generally
viewed as the closest predecessor of real technology utilization and the
cornerstone of technology acceptance studies in educational settings (Ajzen, 1991;
Venkatesh et al.,
2003). BI is the motivational willingness of
learners to interact with the specific technology that depends on cognitive
assessment, emotion, and context.
The traditional TAM
has laid stress on perceived usefulness and perceived ease of use as vital
predictors of BI. Nevertheless, the fact that emotional and psychological
nature has no less significance on influencing the decisions on technologies
has been building up (Scherer & Teo, 2019; Teo, 2011).
Emotions determine the way learners evaluate the issue of technology and
psychological resources provide the assurance that they can handle possible
hardships.
Affectively-motivationally,
BI indicates the willingness of learners to use technology under conditions
that have implications on them emotionally. The history of emotional
experiences could influence the perception of learners regarding GenAI
empowering or overwhelming them. The incorporation of emotional and
psychological variables into the models of BI can therefore enable a more
detailed assay of the technology adoption compared to the mere existence of
frozen evaluative positions on one side.
GenAI has some
unique affordances in the context of language learning, such as a flexible
interaction, high-learner-control, and less instantaneous social evaluation (Yang & Zhao,
2024; Zong & Yang, 2025).
These characteristics can be especially applicable to the learners whose
experiences at the classroom are influenced by FLA, FLE, and RES.
GenAI-assisted learning and intelligent tutoring system studies have suggested
that the affective state of the learners can determine how they will perceive
GenAI-aided tools, as well as how they will incorporate adapted tools in the
learning process.
Although GenAI
applications are rapid growing in education, there are little studies that
empirically research the relationship between classroom feelings and RES, and
the predictive value of these factors aimed at predicting the use of GenAI by
learners in English learning. The current literature has been largely
concentrating on either cognitive perceptions or also moral issues and less
emphasis has been given to affective and motivational aspects. This is one of
the gaps that lead to a conclusion that studies should be conducted to
incorporate emotion variables and psychological resources in GenAI-related
learning behavior models.
Indeed, existing
literature has found the significance of FLA, FLE and RES in the development of
experience and behavior of learners in EFL classrooms. Empirical studies that
are based on CVT showed that classroom emotions are the main factors, which affect
motivation, engagement, and decisions related to learning (Dewaele &
MacIntyre, 2014; Pekrun et al., 2011).
In the meantime, research on RES has been able to point to the adaptive
abilities of learners to deal with any academic obstacles and emotional
pressures (Martin, 2013).
Nevertheless, one
can single out several weaknesses in the existing literature. To begin with,
emotional variables and psychological resources have been frequently
considered, and they have not been discussed and viewed in one integrative
theory. Second, with the increasing popularity of technology-enhanced language
learning, the affective outlooks on GenAI-assisted language learning are
underdeveloped. Third, English majors, as a group facing intensive language
learning demands and frequent evaluative situations, have received limited
attention in this emerging line of inquiry.
To address these
gaps, the present study integrates FLA, FLE, and classroom RES within a unified
framework to examine their predictive roles in English majors’ BI to use GenAI
for English learning.
Rationale and research questions
Al though recent
studies have been more interested in GenAI-assisted language learning, there
still exist several gaps in the literature. To begin with, emotional factors
and psychological resources are typically analyzed individually but not in a
single model that represents the complicated experiences of learners. Second,
the emotional aspect in the consideration of GenAI is not well developed,
especially where English as a Foreign Language (EFL) is involved, where
emotional engagement is the most prominent. Third, little attention has been
given to English majors in prior studies on GenAI acceptance in terms of the
high language learning requirements and the evaluation requirement often
imposed upon them. To fill these gaps, this research will incorporate FLA, FLE,
and classroom RES to analyze the predictive values of these theories on the BI
of the English majors to apply the GenAI in learning English. The quantitative
method and the structural equation modeling (SEM) of choosing a quantitative
approach of the study will help shed light on the relative contribution that
the emotions and the psychological resources have towards the process of GenAI
adoption.
Based on the
theoretical framework and research gaps identified above, the present study
addresses the following research questions:
RQ1.
What are the relationships among FLE, FLA, RES, and BI of English majors to use
GenAI to learn English?
RQ2. Do
FLE, FLA, and RES significantly predict English majors’ BI to use GenAI for
English learning?
RQ3.
Do the relative predictive strengths of FLE, FLA and RES have any differences
on BI in English majors to utilize GenAI?
Methodology
Participants and Sampling
A total of 455
undergraduate English majors from universities in China initially participated
in the study. After data screening, 433 valid questionnaires were retained for
subsequent analyses. Out of the final population, 232 were female (53.6%) and
201 were male (46.4%). In terms of academic year, 106 of the participants were
first-year students (24.5%), 125 were second-year students (28.9%), 98 were
third-year students (22.6%), and 104 were fourth-year students (24%),
indicating a relatively balanced distribution across undergraduate levels. The
reason English majors were chosen is due to the high language learning demands
that are usually involved in studying English and high chances of interacting
with digital learning tools, such as GenAI during the study of English. The
method used to recruit the participants in the study was a non-probability
sampling technique which was that of convenience sampling. This sampling
strategy might reduce the sample representativeness, but it is widely used in
both exploratory research and predictive research in the educational research.
All participants reported prior experience using GenAI tools for English
learning. Participation was voluntary, informed consent was obtained prior to
data collection, and participants were informed of their right to withdraw from
the study at any time without penalty.
Instruments
Data were collected
using a self-report questionnaire consisting of 42 items, including informed
consent, demographic information, background questions on GenAI use, and four
main measurement scales assessing FLE, FLA, RES, and BI to use GenAI.
Specifically, the
questionnaire included: informed consent, demographic information (5 items),
background information on GenAI use (6 items), FLE (10 items), FLA (10 items),
RES (6 items), and BI to use GenAI (4 items). Unless otherwise stated, all
scale items were rated on a five-point Likert scale ranging from 1 (strongly
disagree) to 5 (strongly agree). Reliability was examined using Cronbach’s α,
while construct validity was supported by reference to prior empirical studies
that have established the validity of the original instruments.
Foreign Language Anxiety
FLA was measured
using items adapted from the Foreign Language Classroom Anxiety Scale (FLCAS)
developed by Horwitz et al. (1986), which conceptualizes foreign language
anxiety as a situation-specific construct related to communication apprehension
and fear of negative evaluation. In the present study, 10 items were adapted to
assess learners’ anxiety during English learning activities involving GenAI,
and the scale was treated as a unidimensional construct, consistent with common
empirical practice. A sample item is: “I feel nervous when using GenAI tools to
complete English learning tasks.” The construct validity of the FLCAS has been
well established in previous research (Horwitz et al., 1986; MacIntyre &
Gardner, 1991; Teimouri et al., 2019).
Internal consistency reliability was examined using Cronbach’s α, with values
of 0.70 or above considered acceptable (Nunnally & Bernstein, 1994).
Foreign Language Enjoyment
FLE was measured
using items adapted from the Foreign Language Enjoyment Scale (FLES) proposed
by Dewaele & MacIntyre (2014), which assesses learners’ positive
emotional experiences in language learning. The present study employed 10 items
to measure overall FLE in GenAI-assisted English learning contexts. Although
the original scale includes multiple dimensions, enjoyment was treated as a
single global construct in this study. A sample item is: “Using GenAI makes
English learning more enjoyable for me.” Previous studies have provided strong
evidence for the construct validity of the FLES (Dewaele & MacIntyre, 2014; Dewaele et al.,
2018). Internal consistency reliability was
examined using Cronbach’s α, with values of 0.70 or above considered acceptable
(Nunnally &
Bernstein, 1994).
Psychological Resilience
RES was measured
using items adapted from the Brief Resilience Scale (BRS) developed by Smith et
al. (2008),
which conceptualizes RES as individuals’ ability to recover from stress and
adapt to adversity. In this study, 6 items were adapted to assess learners’ RES
in GenAI-assisted English learning contexts, and the scale was treated as a
unidimensional construct, consistent with the original scale design. A sample
item is: “I tend to recover quickly after difficulties in English learning with
GenAI.” Previous research has established the construct validity of the BRS (Smith et al., 2008;
Windle et al., 2011).
In the present study, internal consistency reliability was assessed using
Cronbach’s α.
Behavioral Intention to Use
GenAI
BI to use GenAI was
measured using 4 items adapted in accordance with the behavioral intention
construct in the TAM (Davis, 1989). Items were worded following standard
TAM-based conventions to reflect learners’ planned future use of GenAI for
English learning. A sample item is: “I intend to continue using GenAI for
English learning in the future.” The predictive validity of TAM-based BI
measures has been established in prior research (Teo, 2011;
Venkatesh et al.,
2003). In the present study, internal
consistency reliability was assessed using Cronbach’s α.
Data Collection Procedure
Data were collected
over a two-week period using an online questionnaire administered via the
Wenjuanxing platform. The survey link was distributed through course-related
online platforms and social media groups targeting undergraduate English
majors. Participants were required to provide electronic informed consent
before completing the questionnaire.
Following data
collection, several data-cleaning procedures were conducted to ensure response
quality. First, questionnaires with incomplete responses were removed. Second,
responses with extremely short completion times (less than 120 seconds) were
excluded, as such responses were considered unlikely to reflect careful reading
and thoughtful responding. Third, one questionnaire exhibiting straight-line
responding across Likert-scale items (standard deviation = 0) was removed.
After these procedures, the final dataset consisted of 433 valid
questionnaires.
Data Analysis
All statistical
analyses were conducted using IBM SPSS Statistics 27 and AMOS 26. Data
screening and preliminary analyses, including descriptive statistics,
reliability analysis, and Pearson correlation analysis, were performed using
SPSS.
Pearson correlation
coefficients were calculated to examine the bivariate relationships among FLE,
FLA, RES, and BI to use GenAI. To examine the measurement properties of the
instruments, confirmatory factor analysis (CFA) was conducted using AMOS to assess
the construct validity of the four latent variables. Subsequently, SEM was
employed to test the hypothesized predictive relationships among FLE, FLA, RES,
and BI to use GenAI. Standardized path coefficients (β) and corresponding
significance levels were used to evaluate the strength and significance of the
hypothesized paths.
Results
Descriptive statistics
Table 1 provides
descriptive statistics of all the variables of the study and also gives their
reliability estimate. As shown in the table, the mean scores of all four
constructs were above the midpoint of the five-point Likert scale, indicating
that participants generally reported moderate to relatively high levels of FLE,
FLA, RES and BI to use GenAI for English learning.
Regarding
reliability, the Cronbach’s α coefficients ranged from 0.879 to 0.95, exceeding
the commonly accepted threshold of 0.70, which indicates satisfactory internal
consistency (DeVellis, 2017; Nunnally & Bernstein, 1994).
Specifically, the BI scale demonstrated good reliability (α = 0.879), which is
considered adequate for research purposes, particularly for short scales with a
limited number of items (Tavakol & Dennick, 2011). Overall, these
results provide initial support for the adequacy of the measurement instruments
used in the study.
Table 1 Descriptive statistics and reliability of
the main variables
|
Variables
|
N
|
Mean
|
SD
|
Cronbach’s α
|
|
FLE
|
433
|
3.464
|
1.142
|
0.949
|
|
FLA
|
433
|
3.435
|
1.158
|
0.950
|
|
RES
|
433
|
3.512
|
1.149
|
0.917
|
|
BI
|
433
|
3.540
|
1.298
|
0.879
|
Correlation analysis
Pearson correlation
coefficients among the main variables are presented in Table 2. As shown in the
table, all variables were significantly correlated with each other at the 0.01
level, indicating systematic associations among FLE, FLA, RES and BI.
According to Cohen’s
(1988)
conventional benchmarks, correlation coefficients of approximately 0.10, 0.30,
and 0.50 represent small, medium, and large effect sizes, respectively. In the
present study, the observed correlation coefficients ranged from 0.81 to 0.92,
far exceeding the threshold for large effects. The strongest association was
observed between FLE and FLA (r = 0.924, p < 0.01), followed by the
relationship between FLE and RES (r = 0.916, p < 0.01). BI was also strongly
correlated with FLE (r = 0.840, p < 0.01), FLA (r = 0.853, p < 0.01), and
RES (r = 0.814, p < 0.01).
These results
indicate that learners’ emotional experiences and RES are closely intertwined
with their intention to use GenAI for English learning. The consistently high
correlation coefficients further justify the subsequent use of multivariate
modeling techniques to examine the relative predictive effects of the
independent variables.
Table 2 Pearson correlations coefficients among the
main variables
|
Variables
|
FLE
|
FLA
|
RES
|
BI
|
|
FLE
|
—
|
|
|
|
|
FLA
|
0.924**
|
—
|
|
|
|
RES
|
0.916**
|
0.887**
|
—
|
|
|
BI
|
0.840**
|
0.853**
|
0.814**
|
—
|
Note. ** p
< 0.01.
Measurement model evaluation
CFA was conducted
to evaluate the measurement model comprising four latent constructs: FLE, FLA,
RES, and BI to use GenAI. Model fit was assessed using multiple goodness-of-fit
indices, including the chi-square to degrees of freedom ratio (χ²/df), the
Comparative Fit Index (CFI), the Tucker–Lewis Index (TLI), and the Root Mean
Square Error of Approximation (RMSEA). Following widely accepted criteria, χ²/df
values below 3.0 indicate acceptable model fit (Kline, 2016),
CFI and TLI values above 0.90—and preferably above 0.95—suggest good to
excellent fit (Hu & Bentler, 1999), and RMSEA values
below 0.05 indicate excellent approximation of the data (Browne &
Cudeck, 1992).
The results
indicated that the measurement model demonstrated an excellent fit to the data,
with χ²/df = 1.33, CFI = 0.99, TLI = 0.99, and RMSEA = 0.028 (see Table 3 and
Figure 1). All fit indices met or exceeded the recommended thresholds,
providing support for the adequacy of the measurement model.
Prior to testing
the overall measurement model, separate single-factor CFAs were conducted for
each construct to examine unidimensionality. The results showed that the
single-factor models for FLE, FLA, RES, and BI each demonstrated acceptable to
excellent fit based on the same criteria (see Table 3). These findings support
the treatment of each construct as a unidimensional latent factor and provide
empirical justification for their inclusion in the subsequent structural model
analysis.
Table 3 Goodness-of-fit indices for the structural
model
|
|
χ²/df
|
CFI
|
TLI
|
RMSEA
|
|
FLE
FLA
RES
BI
Overall measurement model
|
1.258
1.045
1.677
2.065
1.336
|
0.997
1.000
0.996
0.998
0.989
|
0.996
0.999
0.994
0.993
0.988
|
0.024
0.010
0.040
0.050
0.028
|
Figure 1 Measurement model based on confirmatory
factor analysis
Structural model
Structural equation
modeling (SEM) was conducted to examine the predictive effects of FLE, FLA, and
RES on BI to use GenAI. The structural model demonstrated an excellent fit to
the data, with χ²/df = 1.33, CFI = 0.99, TLI = 0.99, and RMSEA = 0.028 (see
Table 3).
The standardized
path coefficients of the structural model are presented in Table 4. The results
showed that FLE significantly predicted BI (β = 0.539, p < 0.001). FLA also
exhibited a strong and significant predictive effect on BI (β = 0.812, p <
0.001). In addition, RES had a positive and significant effect on BI (β =
0.226, p < 0.001). Overall, the findings indicate that all three affective
variables exert significant predictive effects on EFL learners’ intention to
use GenAI.
With all these
combined, the SEM results demonstrate that all three predictors exert
statistically significant effects on English majors’ intention to use GenAI. In
addition to that, the standardized coefficients indicate substantial
differences in the relative strength of these effects, laying the groundwork
for further interpretation in the discussion section.
Table 4 Standardized path coefficients of the
structural model
|
Path
|
β
|
p
|
|
FLE → BI
|
0.539
|
***
|
|
FLA → BI
|
0.812
|
***
|
|
RES → BI
|
0.226
|
***
|
Note. *** p <
0.001
Figure 2 Structural equation model of the predictive
effects on BI
Discussion
General Presentation of the Results
The current
research attempted to determine the relationship and predictive value of FLE,
FLA, and RES on English majors to commit to English learning through GenAI. It
is possible to note that the combination of correlational analysis and SEM
allow the study to offer an in-depth insight into the emotional and
psychological processes behind the agreement of the learners to the emergent
technologies of GenAI in the language learning settings. As opposed to
interpreting the acceptance of GenAI by learners as a cognitive or instrumental
choice, this study places the adoption of GenAI in the emotional and
psychological context of language learning in learners.
From a broader
theoretical perspective, the findings resonate with recent calls to move beyond
traditional technology acceptance models that prioritize cognitive beliefs
(e.g., perceived usefulness) while under-theorizing affective dimensions (Davis, 1989;
Venkatesh et al.,
2003). The current study proves a more
emotion-sensitive explanation of technology adoption in language learning
settings because it shows that affective variables have significant predictive
value, even when other crucial elements of the technology adoption concept are
considered.
In general, the
prediction strength of the three variables is very significant in predicting
BI, but the strength varies widely. Emotions connected with the language,
especially FLA and FLE became more significant predictors compared to RES. This
pattern suggests that learners’ immediate affective experiences in language
learning contexts play a more central role in shaping GenAI adoption than more
general psychological traits. The findings indicate the need to contextualize
the technology acceptance within emotional ecology of language acquisition.
From the perspective of CVT (Pekrun, 2006), learners’ emotional experiences are
shaped by their perceived control over learning activities and the value they
attach to these activities, which subsequently guide behavioral decisions. For
English majors, the academic identity and learning objectives are directly
connected to the English proficiency and performance meaning so that GenAI
tools might be viewed as an instrumentally valuable learning tool. That is why
the presence of language-related emotions and their specific role in
influencing their intention to use GenAI might explain this very situation.
Interrelationships among FLE, FLA, RES and BI
On the bivariate
relationships level, the findings showed uniform high levels of association
between BI and all the three predictors. The size of these associations, which
was higher than traditional significance levels of large effect size,
demonstrates that the BI of learners to use GenAI is interwoven with both
emotional and psychological learning experience in English. This high degree of
correlations supports the idea that BI does not appear as a vacuum but is
embedded into a web of affective and psychological events.
FLE represents
learners’ positive emotional engagement with language learning, encompassing
interest, satisfaction, and a sense of accomplishment (Dewaele & MacIntyre, 2014).
The high level of association between FLE and BI suggests that learners who
find the process of learning English to be an enjoyable and meaningful
experience are more likely to make GenAI one of the learning strategies. This
finding is consistent with previous research highlighting the facilitating role
of enjoyment in promoting learner engagement and openness toward innovative
learning tools in language learning contexts (Dewaele & MacIntyre, 2014;
Pekrun, 2006).
This may reflect a tendency for positively engaged learners to explore and
integrate new learning tools that align with their existing motivation and
curiosity (Fredrickson, 2001; Pekrun, 2006).
At the same time,
the strong association between FLA and BI highlights the complex role of
negative emotions in shaping learners’ technology-related decisions. FLA is
commonly associated with concerns about performance, evaluation, and linguistic
inadequacy (Horwitz et al., 1986). The intention had a
strong correlation with the anxiety implying that those learners who showed
more stress and tension in their English language learning have a higher
tendency to seek external support. GenAI, that provides privacy in interaction,
instant feedback, and the minimization of social assessment, can be viewed as a
helpful tool in dealing with these emotional issues. This pattern aligns with
earlier studies suggesting that anxiety may motivate learners to seek
alternative or supportive learning resources when facing performance pressure
and evaluative concerns (Horwitz et al., 1986; MacIntyre, 2017).
RES was also
strongly associated with BI, indicating that learners who perceive themselves
as capable of recovering from difficulties tend to express stronger intentions
to use GenAI. RES reflects a general capacity to adapt to stress and persist in
the face of challenges (Smith et al., 2008). Resilient learners might
be more comfortable with such tools when interacting with the context of GenAI
where the process mostly includes experimentation, trial, and error and
uncertainty. This result is corroborated by the results of earlier studies that
highlight the importance of resilience as a facilitating psychological variable
that motivates learners to survive uncertainty and continue the learning
process even in difficult circumstances (Smith et al., 2008;
Windle et al., 2011).
Predictive Effects in the Structural Model
Although the
correlational results indicate high levels of relationships between the
variables, the SEM offers a better idea on which variable is more important by
explaining the shared variance. The outcome of SEM revealed that FLE, FLA, and
RES had statistically significant predictive measures on BI with highly
different effect measures.
FLA proved to be
the most powerful forecast in the model. The study finding implies that the
desire by learners to use GenAI is closely correlated to the necessity to cope
with the stress and uncertainty in language. Functionally, anxiety may heighten
learners’ awareness of their linguistic limitations and increase their
motivation to seek tools that offer support and reduce perceived risk (MacIntyre, 2017).
GenAI will be able to fulfill this role by offering prompt feedback, a way to
rehearse privately and an anxiety-free space of interaction, thus covering the
anxiety-related issues of learners.
FLE also
demonstrated a substantial predictive effect, though weaker than anxiety. This
indicates that enjoyment is also a key factor in maintaining the interest of
learners to continue to use GenAI especially in increasing openness, curiosity
and intrinsic motivation. It is possible that enjoyment will make learners
perceive GenAI not only as a remedial activity but as a more interesting and an
exciting part of their learning process.
RES, although
significant, exhibited a comparatively smaller effect. This suggests that RES
contributes to BI indirectly by enabling learners to persist in the face of
challenges rather than directly motivating technology adoption. Resilient
learners may be better equipped to tolerate initial frustration or uncertainty
associated with GenAI use, which supports continued engagement but does not, by
itself, strongly drive intention.
According to the
findings, the results are that language emotions, especially FLA and FLE, have
stronger predictive powers on their BI to use GenAI than RES. It indicates that
the situational emotional experiences of the language learning situations have
a more direct influence on contributing to the technology-related decisions,
but subsequently, the RES can be a more distant psychological resource that
facilitates the engagement indirectly. For English majors, whose learning
experiences are closely linked to language performance and self-evaluation,
emotions experienced in GenAI-mediated learning contexts are likely to be
especially salient.
Interpretation of the Predictive Patterns
Combined, the
findings indicate that BI by learners towards using GenAI should be seen as an
affectively motivated choice and not necessarily a rational or cognitive one.
The prevalence of emotional predictors reveals that learners resort to GenAI in
order to improve the performance as well as to control the emotional experience
of language learning (Yang & Zhao, 2024).
The strong role of
anxiety implies that GenAI may function as an emotional coping tool, helping
learners manage stress, fear of negative evaluation, and perceived linguistic
inadequacy. Simultaneously, the high importance of enjoyment underlines the
significance of positive emotional involvement in maintaining voluntary and
further GenAI use. Meanwhile, the psychological basis offered by RES gives
engagement a chance to work but is not the motivating factor.
Conclusion
To conclude, the
given research has shown that the willingness of English majors to use GenAI to
learn the English language can be described as the product of a complex
interaction between emotional and psychological variables. Emotions of learners
associated with language, FLA and FLE turned out to be particularly a good
predictor, and RES was supported in a minor, but relatively weak way. This
pattern underscores the centrality of affect in technology-related decisions
within language learning contexts, a perspective increasingly emphasized in
contemporary SLA research (Dewaele, 2019; MacIntyre, 2017).
Theoretically, the
research adds to the evidence of the work on the GenAI-assisted language
learning as foregrounding affective variables are considered the key elements
of the GenAI adoption. Although the classical models of technology acceptance
have focused mainly on cognition, i.e., on perception of usefulness and
perceived ease of use (Davis, 1989; Venkatesh et al., 2003),
the current results indicate that even in the context of the expanded
psychological resources, emotional experiences can still explain quite a lot of
variation. This research by combining FLE, FLA and RES in one predictive model
gives deeper insights into the interactions of the learner with GenAI within a
language learning environment (Fredrickson, 2001;
Pekrun, 2006).
The findings also
carry important pedagogical implications. The prominent role of FLA suggests
that GenAI may function as an affective support tool for learners who
experience stress, fear of negative evaluation, or linguistic insecurity (Horwitz et al.,
1986). Teachers may therefore consider
leveraging GenAI to provide low-stakes practice opportunities, individualized
feedback, and private rehearsal spaces that reduce emotional pressure. In
addition, the significant influence of FLE highlights the importance of
designing GenAI-mediated learning activities that foster interest, autonomy,
and positive engagement, which are known to support sustained motivation and
persistence (Dewaele & MacIntyre, 2014; Pekrun et al., 2017).
Although RES played a more modest role, its significance indicates that
learners’ general coping capacity may indirectly facilitate engagement with
GenAI-assisted learning, particularly in challenging or uncertain learning
environments (Smith et al., 2008; Windle et al., 2011).
Despite the
pedagogical implications discussed above, the interpretation of the present
findings should be situated within the methodological and contextual boundaries
of the study. Accordingly, several limitations need to be acknowledged. First,
the cross-sectional research design limits causal interpretations of the
observed relationships, as emotional experiences, RES, and BI may influence one
another dynamically over time (Pekrun, 2006). Longitudinal or experimental studies
are therefore needed to examine how learners’ emotions and GenAI use intentions
co-develop. Second, the reliance on self-report questionnaire data may
introduce common method bias and subjective response tendencies (Podsakoff et al.,
2003). Future research could incorporate
behavioral indicators, usage logs, or learning analytics to complement
self-reported intentions. Finally, as the sample consisted exclusively of
Chinese English majors, caution is warranted in generalizing the findings to
learners from other disciplinary, institutional, or cultural backgrounds.
Further research involving more diverse populations would enhance the external
validity of the conclusions.
Overall, by
highlighting the emotional and psychological foundations of GenAI use in
language learning, this study advances current understanding of why learners
adopt GenAI technologies and offers empirically grounded directions for both
theory development and pedagogical practice.
Acknowledgement
We gratefully acknowledge the anonymous reviewers for their diligent
efforts in reviewing this study.
Declaration
of Interest Statement
There is no potential conflict
of interest.
References
Cohen, J. (1988). Statistical power analysis
for the behavioral sciences (2nd ed.). Lawrence Erlbaum.
DeVellis, R. F. (2017). Scale Development: Theory
and Applications (4th ed.). Sage.
Kline, R. B. (2016). Principles and practice of
structural equation modeling (4th ed.). Guilford Press.
Masten, A. S. (2014). Ordinary magic: Resilience
in development. Guilford Press.
Saito, K., Dewaele, J.‐M., Abe, M., &
In'nami, Y. (2018). Motivation, emotion, learning experience, and second
language comprehensibility development in classroom settings: A cross-sectional
and longitudinal study. Language Learning, 68(3),
709-743. https://doi.org/10.1111/lang.12297
Author Bios
Puyang Sun is a PhD candidate
in English Language Education and English Literature at the
University of Wales Trinity Saint David, UK. Her research interests include
second language acquisition, foreign language emotions, positive psychology in
language education, and technology-enhanced language learning, with a
particular focus on generative artificial intelligence in EFL contexts.