The Role of GenAI in EFL: Impact on Learning Motivation and Outcome

doi:10.21203/rs.3.rs-5144171/v1

The purpose of this paper is to explore the role of GenAI (generative artificial intelligence) in EFL. Specifically, this paper focuses on its impact on learning motivation and outcome. With a total of five hundred and sixty-six students in four Chinese university, the questionnaire was shown to be reliable and valid. Research has revealed that GenAI has a strong effect on EFL learning motivation and outcomes in China; that GenAI plays an important role in learners’ personalized learning, learning interest and motivation through timely feedback and strong interactivity; and GenAI can enhance learner’s English listening and speaking skills, academic writing skills, vocabulary expansion and understanding of English-speaking countries' culture. The paper also revealed that personal factors such as grade, English proficiency and language environment in which students grew up have a strong effect on English learning outcomes. The research findings widens our understanding of Chinese student’s use of artificial intelligence in second language acquisition, and also reflect Chinese student’s AI anxiety in the era of information technology.

Biological sciences/Psychology

Health sciences/Diseases

GenAI

EFL

learning motivation

learning outcome

GenAI has caught educator’s attention globally[1]. Research has shown that GenAI tools have been widely used in intelligent tutoring systems, personalized curriculum design, and web-based chatbots[2–4]. With highly personalized learning experience, these tools have facilitated learner's learning efficiency [5, 6]. On early study found that ChatGPT, an GenAI language model built in November 2022, has clear advantages for idea generation and data identification when doing academic research[7]. Many countries and regions have published a series of favorable policies on the application of GenAI to education[8].

For language education, GenAI is not a substitute for teachers; instead, it can be regarded as a powerful teaching and learning mode[9, 10]. In many countries, such as America, China and India, the application of GenAI in education have significantly enhanced the efficiency and effectiveness of language education[11].

For English learners, GenAI reshapes the learning mode with a personalized learning method based on every learner's specific needs and abilities with higher learning efficiency[12]. Meanwhile, GenAI instantly assesses learner’s oral pronunciation, grammar, and vocabulary use and provides timely feedback and corrections[13]. In addition, by using AI-driven interactive learning applications to practice conversations, English learners improve their speech ability and increases the enjoyment of learning[14].

Despite the recognition of GenAI’s educational potential, systematic research on its impact on college students in terms of EFL’s learning effectiveness, motivation enhancement, and academic performance needs to be conducted. This research gap has attracted many educational scholars and practitioners. Foreign language educators, language teachers, and curriculum developers are particularly interested in the influence of GenAI on students’ academic writing skills and how to better integrate GenAI into English teaching practices[15]. However, with the global demand for GenAI, the exploration of effective teaching strategies and tools has become particularly important[16, 17]. As an innovative teaching aid, the application and effectiveness of GenAI tools in English teaching need to be further explored.

This study focuses on the effect of GenAI on EFL’s learning motivation, effectiveness and achievement. Questionnaires will be used, and a wide range of data to analyze how GenAI affects the learning effectiveness of college English learners in China will be collected. We hope that this study can provide new perspectives for second language acquisition, and support for the application of educational technology in language learning. As a result, language learners can improve the efficiency of their English learning, facilitate communication among people with different cultural backgrounds, and ultimately facilitate better integration into international communication and cooperation in the era of globalization.

GenAI, beyond traditional teaching methods, as an innovative learning aid, has shown unique value in language learning. One Study demonstrated that GenAI enhances the efficiency and quality of English learning and brings innovation and transformation to language education[18]. Understanding the theoretical foundations can help us better utilize GenAI as a tool for EFL.

2.1 Cognitive psychology

Cognitive psychology provides a framework for understanding the impact of GenAI on English learning. It offers models on how humans receive, process, and store information[19]. Empirical investigations within cognitive psychology have demonstrated the utility of memory theories, such as working memory and long-term memory constructs, in informing the AI-assisted review and memorization strategies to augment language learning effectiveness[20]. GenAI adeptly curate tailored learning resources, including reading materials, videos, and audio, calibrated to individual proficiency levels[21, 22] Moreover, GenAI has been harnessed in the oral English pronunciation correction systems and automated scoring mechanisms, which can effectively identify and rectify pronunciation errors. GenAI Apps accommodate flexible learning schedules and diverse learning environments with convenience. This personalized and convenient mode of language instruction represents a paradigm shift from traditional methods, underscoring its unparalleled efficiency and efficacy.

2.2 Self-determination theory

Self-determination theory, formulated by psychologists Deci and Ryan in the 1980s, posits that all organisms, including humans, possess inherent capabilities to develop intricate systems of intrinsic motivation and to pursue three fundamental psychological needs: autonomy, competence, and relatedness[23]. GenAI tools have been found their capacity to bolster learners' engagement, efficiency, and particularly human-computer interaction, which enables learners to perceive progress and accomplishment in their educational endeavors. This, in turn, cultivated confidence and sustained motivation in learning[24]. Within the realm of education, GenAI, functioning as a facilitator of learning, autonomously curates and explores diverse educational resources and nurtures student's autonomy, competence, and sense of relevance. Moreover, GenAI engenders and sustains positive motivational and behavioral patterns[25, 26].

2.3 Technology acceptance model

Technology Acceptance Model (TAM) posits that user's attitudes toward a given technology, which influence their behavioral intentions and ultimately determine their choice[27, 28]. Introduced by Fred Davis in 1989, TAM explicates user's acceptance behavior toward information technology, drawing upon principles of rational behavior and notably expectancy theory and attitude theory. This model underscores two pivotal determinants: perceived ease of use and perceived usefulness. Perceived ease of use pertains to users' perceptions of the ease or difficulty associated with employing certain technology. Consequently, users are inclined to embrace and utilize a technology perceived as user friendly. On the other hand, perceived usefulness denotes the extent to which users believe that technology will enhance their job performance. Users are more likely to utilize technology if they perceive it as beneficial to their work or study endeavors[29]. Widely employed across information systems, human-computer interaction, and educational technology domains, TAM serves as a predictive and explanatory framework for user acceptance of diverse technological products[30].

2.4 Krashen's input hypothesis

Krashen's input hypothesis postulates that language acquisition transpires when learners are exposed to linguistic input slightly beyond their current proficiency level, encapsulated by the "i+1" principle[31]. GenAI technology adeptly tailor input to match learner's abilities and progress, ensuring exposure to linguistically challenging yet comprehensible material. Leveraging an adaptive learning system, GenAI dynamically modulates input difficulty to ensure that learners consistently operate within the optimal learning zone and facilitate natural language proficiency. Additionally, GenAI mitigates learner anxiety and bolsters motivation through personalized positive reinforcement, encouragement, and assistance. Furthermore, GenAI enriches learning experiences through gamified strategies and diminish affective filters and enhancing language acquisition efficiency. Krashen's theoretical framework elucidates GenAI's potential to enhance language acquisition[32]. As GenAI technology advances and has broader applications, its capacity to expedite language acquisition should be further explored and realized.

Building upon the aforementioned theoretical underpinnings, three research questions are proposed as follows.

1.Is there any descriptive factors related to learning motivation and learning outcomes?

2.The use and understanding of GenAI have a significant positive impact on EFL’s learning motivation?

3.The use and understanding of GenAI have a significant positive impact on EFL’s learning outcomes?

1. Statistics Description

This study conducted a detailed survey on higher education students throughout China. In May to June 2024, we distributed 800 questionnaires in four Chinese university, located in Guangzhou and Changsha. 711 questionnaires were returned, with 88.88% recovery rate. Among them, 566 questionnaires were valid, for a recovery effectiveness rate of 79.61% as presented in table 1.

Table 1 Statistics description

Variant	options	frequency	Percentage (%)
Gender	male	312	55.124
	female	239	42.226
	others	15	2.65
Age	18-19	99	17.491
	20-21	219	38.693
	22-23	188	33.216
	Over 24-year-old	60	10.601
Grade	freshman	55	9.717
	Sophomore	106	18.728
	Junior	208	36.749
	Senior	197	34.806
English level	None	61	10.777
	CET4	100	17.668
	CET6	213	37.633
	TEM8	192	33.922
Language environment	Non-Chinese speaking area	58	10.247
	bilingual area	95	16.784
	dialect	218	38.516
	Mandarin	195	34.452
English learning duration (per week)	＜ 1 hour	179	31.625
	1 - 3 hours	320	56.537
	4 - 6 hours	46	8.127
	＞ 6 hours	21	3.71
Study or travel abroad experience	never	89	15.724
	Short-term (＜ 1 month)	263	46.466
	Medium-term (1 to 6 months)	162	28.622
	Long-term (＞ 6 months)	52	9.187

2. Methodology

In this study, all methods were carried out in accordance with relevant guidelines and regulations. The experimental protocols were approved by the Institutional Review Board of the first author’s working institution. Informed consent was obtained from all subjects or their legal guardian prior to their inclusion in the study.

The design of the questionnaire draws on recent research in the fields of artificial intelligence, second language acquisition, and educational psychology. It specifically references theories: Fred Davis's Technology Acceptance Model, Krashen's Input Hypothesis Theory, and Edward L. Deci and Richard M. Ryan's Self-Determination Theory. The questionnaire was structured into four dimensions: participant’s demographics, use and understanding of GenAI tools, EFL learning motivation, and learning effectiveness.

The first dimension is to collect participant's demographic information, including their gender, age, grade, English level, language environment in which they grew up, average English learning time (per week), and overseas study or travel experience. The second dimension is to assess undergraduate’s use frequency and understanding of GenAI tools. The third dimension focuses on the undergraduate’s motivational role of GenAI tools in English learning, such as increase learning interests, improve listening and speaking skills, enlarge their vocabulary and improve intercultural communication. The fourth dimension explores how well GenAI tools improve EFL learning outcomes: English speaking, listening, reading, writing, vocabulary, and intercultural communication.

All questions were evaluated on a five-point Likert scale (for GenAI1-GAi11: SD=Strongly Disagree=1, D=Disagree=2, NS=Not Sure=3, A=Agree=4, SA=Strongly Agree=5; for MTV1 to MTV13 and LO1 to LO9, Not Sure= NS=3, SA=Strongly Agree=1, A=Agree=2, NS=Not Sure=3,D=Disagree=4,SD=Strongly Disagree=5) to provide a picture of the impact of GenAI tools on higher education students' English learning. Questions of the specific measurement are shown in Table 2.

Table 2 Variable definitions and assignments

Dimension	variant	encoding	variable assignment
demography	gender	GND	M=1, F=2, Other=3
	age	AGE	18-19 = 1, 20-21 = 2, 22-23 = 3, 24+ = 4
	grade	GRD	Grade 1 = 1, Grade 2 = 2, Grade 3 = 3, Grade 4 = 4
	English Level	EXM	None=1,CET4=2,CET6=3,TEM8=4
	linguistic environment	ENV	Non-Chinese-speaking Area=1, Bilingual =2, Dialects=3, Mandarin=4
	English learning Duration	TIM	Less than 1 hour = 1, 1 to 3 hours = 2, 4 to 6 hours = 3, more than 6 hours = 4
	Study or travel abroad	ABD	None = 1, short-term = 2, medium-term = 3, long-term = 4
Use and understanding of GenAI	Has GenAI been used	GAI1	Never= 1, Occasional use = 2, Frequent use= 3, Everyday=4, Highly Dependent = 5
	Use frequency	GAI2	Never = 1, Several times a month = 2, Several times a day = 3, Several times a week = 4, Every day = 5
	Instant feedback	GAI3	SD=1, D=2, NS=3, A=4, SA=5
	Learning efficiency	GAI4	SD=1, D=2, NS=3, A=4, SA=5
	Match between generated content and needs	GAI5	SD=1, D=2, NS=3, A=4, SA=5
	user-friendly interface	GAI6	SD=1, D=2, NS=3, A=4, SA=5
	Generating quality	GAI7	SD=1, D=2, NS=3, A=4, SA=5
	interaction	GAI8	SD=1, D=2, NS=3, A=4, SA=5
	Alternatives to Traditional Learning Methods	GAI9	SD=1, D=2, NS=3, A=4, SA=5
	Effectiveness of English communication skills	GAI10	SD=1, D=2, NS=3, A=4, SA=5
	Personalized Learning	GAI11	SD=1, D=2, NS=3, A=4, SA=5
Learning motivation	Effective learning method	MTV1	SA=1, A=2, NS=3, D=4, SD=5
	GenAI has a future	MTV2	SA=1, A=2, NS=3, D=4, SD=5
	Help to understand cultural differences	MTV3	SA=1, A=2, NS=3, D=4, SD=5
	help English learning in the long run	MTV4	SA=1, A=2, NS=3, D=4, SD=5
	Share with English learning partners	MTV5	SA=1, A=2, NS=3, D=4, SD=5
	Make English learning more Fun	MTV6	SA=1, A=2, NS=3, D=4, SD=5
	Adapt to the new paradigm of English learning	MTV7	SA=1, A=2, NS=3, D=4, SD=5
	Enhance motivation for English learning	MTV8	SA=1, A=2, NS=3, D=4, SD=5
	Improve self-confidence in learning English	MTV9	SA=1, A=2, NS=3, D=4, SD=5
	improve self-directed learning	MTV10	SA=1, A=2, NS=3, D=4, SD=5
	Improve the depth of English Learning	MTV11	SA=1, A=2, NS=3, D=4, SD=5
	Clearer goals for English language learning	MTV12	SA=1, A=2, NS=3, D=4, SD=5
	Enhance the emotional experience of English learning	MTV13	SA=1, A=2, NS=3, D=4, SD=5
Learning outcome	spoken language	LO1	SA=1, A=2, NS=3, D=4, SD=5
	listening	LO2	SA=1, A=2, NS=3, D=4, SD=5
	writing	LO3	SA=1, A=2, NS=3, D=4, SD=5
	reading	LO4	SA=1, A=2, NS=3, D=4, SD=5
	Grammar	LO5	SA=1, A=2, NS=3, D=4, SD=5
	vocabulary	LO6	SA=1, A=2, NS=3, D=4, SD=5
	Cross-culture cultures	LO7	SA=1, A=2, NS=3, D=4, SD=5
	critical thinking	LO8	SA=1, A=2, NS=3, D=4, SD=5
	innovation capacity	LO9	SA=1, A=2, NS=3, D=4, SD=5

4.1 Validity testing

SPSS 20.0 was used to validate the reliability of the data for all the variables covered in the paper.

4.1.1 Normal distribution test

The basic premise of the statistical analysis was that the questionnaire data followed a normal distribution, so an analysis of the mean, standard deviation, skewness, and kurtosis of the questionnaire is required. The questionnaire data conform to a normal distribution(The absolute value of the skewness coefficient＜3; the absolute value of the kurtosis coefficient＜10). The results are shown in Table 3.

Table 3 Descriptive statistics

Variant	mean	standard deviation	skewness	kurtosis
GAI1	3.532	1.314	0.481	1.007
GAI2	3.571	1.246	-0.495	-0.883
GAI3	3.64	1.243	-0.507	-0.905
GAI4	3.592	1.26	-0.523	-0.907
GAI5	3.601	1.221	-0.519	-0.776
GAI6	3.594	1.282	-0.508	-0.976
GAI7	3.617	1.245	-0.524	-0.857
GAI8	3.549	1.287	-0.495	-0.956
GAI9	3.565	1.268	-0.494	-0.961
GAI10	3.558	1.307	-0.502	-0.975
GAI11	3.595	1.324	0.512	-1.02
MTV1	3.624	1.268	-0.561	-0.907
MTV2	3.657	1.26	-0.616	-0.765
MTV3	3.666	1.287	-0.704	-0.657
MTV4	3.618	1.246	-0.613	-0.719
MTV5	3.629	1.222	-0.558	-0.757
MTV6	3.67	1.223	-0.618	-0.707
MTV7	3.668	1.22	-0.622	-0.676
MTV8	3.67	1.214	-0.602	-0.679
MTV9	3.645	1.203	-0.472	-0.925
MTV10	3.668	1.284	-0.605	-0.868
MTV11	3.634	1.214	-0.55	-0.82
MTV12	3.641	1.232	-0.584	-0.755
MTV13	3.673	1.247	-0.632	-0.755
LO1	3.592	1.264	-0.48	-0.952
LO2	3.534	1.315	-0.481	-1.024
LO3	3.565	1.332	-0.524	-1
LO4	3.541	1.317	-0.513	-0.963
LO5	3.523	1.302	-0.49	-0.912
LO6	3.519	1.3	-0.491	-0.971
LO7	3.553	1.27	-0.499	-0.868
LO8	3.567	1.295	-0.507	-0.945
LO9	3.574	1.297	-0.491	-0.964

Table 3 shows that the absolute value of the skewness coefficient is less than 1, and the absolute value of the kurtosis coefficient is less than 3. These indicates that the data follow a normal distribution.

Cronbach's α test was performed. The results are shown in Table 4. The Cronbach's α‌ coefficients for each scale indicate a high level of reliability (＞0.8）.

Table 4 Cronbach's α‌ coefficients

Dimension	quantities	Cronbach's α‌ coefficient	confidence level
GAI	11	0.969	high reliability
MTV	13	0.972	high reliability
LO	9	0.962	high reliability

2. Validity analysis

Validity analysis is to measure the structural validity of the questionnaire. The evaluation indices include the KMO test, Bartlett's test of sphericity, cumulative contribution rate, and factor loading. The KMO test and Bartlett's test of sphericity are used to determine whether the data are suitable for factor analysis. The cumulative contribution rate represents the degree of cumulative validity of the common factors on the scale. The factor loading indicates the degree of correlation between the original variables and a common factor. The KMO value of the scale was 0.984. The approximate chi-square value of Bartlett's test of sphericity was 19063.113 (p<0.0001), which indicates that the scale is suitable for factor analysis. The specific test results are shown in Table 5. An eigenvalue greater than 1 was used as the criterion for factor extraction, resulting in the extraction of four main components. As shown in Table 5, the proportion of cumulative explained variance is 76.905%. The factors attributed to the 34 items correspond to the same dimensions as the initial item set. There is no cross-factor loading phenomenon, which shows that the structural validity is relatively high.

Table 5 Exploratory factor analysis

Table of factor loading coefficients after rotation						Commonality (common factor variance)
	Postrotation factor loading coefficients
	Factor 1	Factor 2	Factor 3		Factor 4
GAI1	0.219	0.817	0.201		0.134	0.774
GAI2	0.245	0.808	0.185		0.088	0.755
GAI3	0.251	0.815	0.187		0.056	0.766
GAI4	0.269	0.802	0.215		0.04	0.764
GAI5	0.183	0.835	0.151		-0.024	0.754
GAI6	0.255	0.824	0.179		0.031	0.777
GAI7	0.239	0.82	0.169		0.042	0.76
GAI8	0.284	0.795	0.213		0.059	0.761
GAI9	0.268	0.816	0.181		0.049	0.773
GAI10	0.285	0.808	0.179		0.019	0.767
GAI11	0.262	0.827	0.163		0.052	0.782
MTV1	0.817	0.235	0.217		0.064	0.774
MTV2	0.799	0.245	0.225		0.051	0.752
MTV3	0.818	0.227	0.218		0.074	0.773
MTV4	0.791	0.265	0.219		0.049	0.747
MTV5	0.766	0.266	0.252		0.061	0.724
MTV6	0.795	0.273	0.208		0.051	0.753
MTV7	0.786	0.244	0.249		0.022	0.74
MTV8	0.765	0.296	0.216		0.047	0.722
MTV9	0.791	0.211	0.216		0.026	0.717
MTV10	0.821	0.211	0.243		0.061	0.781
MTV11	0.811	0.262	0.175		0.034	0.758
MTV12	0.788	0.243	0.25		0.051	0.745
MTV13	0.811	0.252	0.215		0.086	0.776
LO1	0.258	0.216	0.806		0.063	0.767
LO2	0.287	0.206	0.812		-0.051	0.787
LO3	0.224	0.201	0.84		0.024	0.796
LO4	0.245	0.2	0.823		0.07	0.782
LO5	0.253	0.18	0.801		0.072	0.742
LO16	0.209	0.199	0.831		0.093	0.782
LO7	0.275	0.225	0.792		0.001	0.754
LO8	0.255	0.183	0.828		0.038	0.785
LO9	0.263	0.238	0.164		0.914	0.988
	cumulative interpretation variance ratio	28.903%		54.676%	74.042%	76.905%
	KMO value	0.984
	Bartlett's test of sphericity	approximate chi-square		19063.113
		P		0.000***
		df		528

Note: ***, **, and * represent 1%, 5%, and 10% significance levels.

4.2 Related Analysis

Related analysis are used to examine the correlation between higher education student’s use and understanding of GenAI tools and learning motivation, learning outcomes. As shown in Table 6, there was a highly significant positive correlation between higher education student’s perception and use frequency of GenAI tools and their learning motivation, learning outcomes. Specifically, the correlation coefficient between higher education student’s perception and use frequency of GenAI tools and learning motivation is 0.558, with a significance level of 0.000. It indicates that higher education student’s use and understanding of GenAI tools are associated with stronger learning motivation. Similarly, the correlation coefficient between higher education student’s use and understanding of GenAI tools and learning outcomes is 0.489, with a significance level of 0.000. This suggests that greater use and understanding of GenAI tools lead to better learning outcomes.

Table 6 Correlation analysis of use and understanding of GenAI with learning motivation, learning effectiveness

	MTV	LO
GenAI	0.558 (0.000***)	0.489 (0.000***)

Note: ***, **, and * represent 1%, 5%, and 10% significance.

Correlation analysis can verify the close relationship between higher education student’s GenAI acceptance and their learning motivation, learning outcomes. It does not mean that they have a significant relationship, therefore, their relationship needs to be further verified.

4.3 Regression Analysis

According to the proposed hypotheses 1 to 3, this study employed regression analysis to test these effects. Learning effectiveness and learning outcomes were treated as dependent variables, while gender, age, grade, English examination, language environment, English learning duration, overseas study and travel experience, and higher education student’s use and understanding of GenAI tools were regarded as independent variables. Regression analyses were conducted using two models. The results are presented in Table 7.

Table 7 The effects of undergraduate’s use and understanding of GenAI tools on learning motivation and learning outcome

variant	MTV		LO
variant	β	t	β	t
GenAI	0.078	2.173	0.115	2.662
GND	-0.035	-1.283	-0.062	-1.919
AGE	-0.063	-2.332	-0.054	-1.677
GRD	0.317	8.736	0.18	4.133
EXM	0.318	8.627	0.26	5.896
ENV	0.205	5.576	0.214	4.851
TIM	0.059	2.155	0.094	2.874
ABD	0.058	2.167	0.108	3.355
F	106.152***		52.722***
Adjustment of R²	0.598		0.423

Note: ***, **, and * represent 1%, 5%, and 10% significance.

Model 1 examined GenAI tools’ effect on higher education student’s learning motivation. The adjusted R²value was 0.598, and the F was 106.152, It reaches statistical significance (p < 0.001), which indicates that the predictor variables explained 59.8% of the variance in learning motivation. Specifically, the regression coefficient for higher education student’s use and understanding of GenAI tools was 0.078, with t value of 2.173 and p value of 0.030**. It verifies Hypothesis H1 in this study: students who frequently use GenAI are significantly more motivated than those who are not. This finding supports the Technology Acceptance Model, particularly the perceived usefulness factor, which influences user's attitudes and behavioral intentions toward technology[33]. If students find GenAI tools easy and convenient to use, they tend to use them for their learning. Ease of use reduces resistance to technology and lowers the threshold for usage, which enhance their learning motivation. When learners perceive that GenAI tools improve learning efficiency and effectiveness through personalized recommendations and interactive learning, they find these tools to be more useful[34].

Model 2 focused on the use and understanding of GenAI tools on student’s learning outcomes. It shows an adjusted R²value of 0.423 and F of 52.722 (p < 0.001), which indicates that the predictor variables explained 47.1% of the variance. The regression coefficient for student’s use and understanding of GenAI tools was 0.115, with t value of 2.662 and p value of 0.008***. These verifies Hypothesis H2: students who frequently uses GenAI tools achieve better English learning outcomes than those who uses GenAI tools with lower frequency. According to self-determination theory, GenAI tools provide personalized learning plans and self-paced learning paths. Hence, students are more motivated to learn when they feel in control of their learning. Active learning improves their learning outcomes. GenAI tools offer instant feedback and adaptive learning contents, which helps learners identify and strengthen their weaknesses and build on their strengths. As learners achieve their goals with GenAI tool’s assistance, their sense of self-confidence can be improved. Then their learning effectiveness will be reinforced because they take the initiative to learn. Additionally, GenAI tools help students to get involved in the social interactions. For example, it helps students to cross the language barriers through language translation and promote international communication and cooperation. At the meantime, dialogue robots provide a safe environment for people to better express their thoughts and emotions both in the oral and written form[35]. These social interactions enhance student’s sense of relatedness and community and boost their learning motivation and effectiveness. By fulfilling the psychological needs of autonomy, competence, and relatedness, GenAI tools promote student’s intrinsic learning motivation and improve learning effectiveness[36].

Research Question 1: Is there any descriptive factors related to learning motivation and learning outcomes?

As presented in Table 7, the grade level, English language proficiency, and language environment of upbringing significantly affected EFL learning motivation and learning outcomes. Higher grades were correlated with stronger positive effects of GenAI on motivation, learning outcomes. And the moderating effect of the language environment was particularly significant; learners from non-Chinese-speaking areas achieved the best learning outcomes. English proficiency also moderated learning achievement significantly, with higher English levels correlating with better learning effectiveness. Gender significantly affected learning achievement and outcomes, but had a non-significant effect on motivation. Additionally, age, frequency of weekly English use, and overseas study and travel experience positively influenced motivation, learning outcomes. These results highlight the important moderating roles of personal factors such as age, English proficiency, and the language environment in which one grows up.

Research Question 2: The use and understanding of GenAI have a significant positive impact on EFL’s learning motivation?

The study revealed that GenAI tools significantly enhance both intrinsic and extrinsic motivation by providing personalized learning plans, instant feedback, and interactive learning environments. This aligns with the core concepts of the TAM, where perceived usefulness is a key factor influencing learner's acceptance and use of new technologies. The perceived usefulness of GenAI tools enhances learner's motivation by increasing learning efficiency and effectiveness. Additionally, self-determination theory supports this study, indicating that GenAI tools meet learner's needs for autonomy, competence, and relevance, thereby stimulating intrinsic motivation [37].

According to Stephen Krashen's input hypothesis, learners benefit from linguistic input that is slightly above their current level, known as comprehensible input. GenAI tools provide students with personalized learning experience and tailor language input to the student's own ability and progress. This personalized input aligns with the learner’s “i + 1” level, ensuring that they consistently receive materials that challenge them just beyond their current proficiency. To improve student’s ability, GenAI tools assist them quickly to understand and correct their mistakes by offering immediate feedback. GenAI tools dynamically adjusts their difficulty of learning materials based on learner’s interactions and performance to ensure continuous exposure to appropriately challenging input. Additionally, GenAI tools maintain learner engagement through various ways, such as online courses, apps, or virtual assistants to provide ongoing comprehensible input. GenAI tools are designed to be interactive and enjoyable, increasing learner motivation and engagement. According to the input hypothesis, language acquisition is more effective when learners are interested in and actively engaged in language input. GenAI tools, like ChatGPT, Claude and Gemini, offer rich learning resources, including video, audio, and reading materials and provide a diverse range of language inputs to help learners grasp languages from multiple perspectives and contexts.

While generative artificial intelligence brings a lot of convenience to learning, it also inevitably has some side effects,for instance, the formation of dependence psychology. Over-reliance on GenAI may weaken student’s autonomous learning ability and intrinsic learning motivation. when students are used to getting answers quickly with the help of artificial intelligence tools, they may reduce the in-depth thinking and active exploration of problems, resulting in less initiative to learn. Taking EFL academic writing as an example, if students frequently use GenAI to generate contents, they may gradually lose the ability to think and express themselves independently. In the long run, when facing learning tasks, students may first think of using external tools, instead of relying on their own knowledge and ability to solve problems, which will have a negative impact on learning motivation.

Research Question 3: The use and understanding of GenAI have a significant positive impact on EFL’s learning outcomes?

The results showed that GenAI tools positively impact learning outcomes, especially listening comprehension, speaking, reading comprehension, and writing skills, particularly in their vocabulary[24, 38, 39]. The personalized learning paths and instant feedback mechanisms of GenAI tools allow learners to receive customized content and guidance according to their progress and abilities, enhancing learning efficiency. The interactivity and enjoyment of GenAI tools further increase learner engagement, making the learning process more vivid and engaging, thus improving learning effectiveness[40]. Furthermore, GenAI tools help learners apply their knowledge in real-life communication scenarios, enhancing their learning performances[41]. While GenAI in education shows great potential for optimizing the learning experience and improving educational effectiveness, it is crucial to guide students to use GenAI critically and consider the ethical implications of GenAI technology in education [42].

Likewise, it may limit creativity and cause AI anxiety. When students over-rely on AI-generated ideas and solutions, their own innovative thinking and creativity may be inhibited. For example, in the process of artistic creation or writing, if students always refer to the works generated by generative artificial intelligence, they may fall into the trap of imitation. And it is difficult to form a unique style and creativity, which will affect the learning effect and personal development. At the same time, GenAI can generate information with high pace, but this can also lead to information overload, causing anxiety and stress. Too much information makes students feel at a loss, and they do not know how to screen and deal with it. In addition, students may worry that they cannot master all the knowledge in face of technology explosion and result in AI anxiety and further reduces the learning effect.

This study explores the role of GenAI tools in L2 acquisition, with a focus on their impact on learning motivation and learning outcomes. It also examined the moderating effects of individual factors such as gender, age, and English learning level. The findings suggest that the use of GenAI tools significantly enhances learners' motivation and outcomes. These results align with the literature on the positive effects of GenAI applications in education.

The results of this study have important implications for educational practices. Educational institutions and teachers can use GenAI tools to enhance the language learning experience and improve learning efficiency. GenAI tools can provide learners with personalized learning paths and stimulate their interest and motivation through interactive and engaging learning environments. Additionally, educational policymakers should consider integrating GenAI tools into language teaching to promote educational innovation and improve learning effectiveness.

The application of GenAI tools in language learning also has broader socioeconomic implications[43]. By providing a personalized and efficient learning experience, GenAI tools help to narrow the gap in the distribution of educational resources and improve educational equity [44]. Furthermore, the widespread adoption of GenAI tools can help individuals develop globally competitive language skills and promote international exchange and cooperation [45].

While this study strongly supports the positive role of GenAI tools in English language learning, several limitations must be noted. At first, the questionnaire data rely on respondent’s self-reports, so that they may be subject to inaccuracies or memory bias. Secondly, although variables such as gender, age, and grade were controlled for. Other variables, such as individual professional background, native language, and learning styles, may have affected the results. The study may have been conducted in specific settings or with specific learner types, which limited the generalizability of the results. Finally, the use of GenAI tools may not have been consistently controlled. With different learners possibly using GenAI in various ways, the results may be affected. Future research should consider these factors for a more comprehensive understanding.

Data Availability

All methods were carried out in accordance with relevant guidelines and regulations. The data underlying the results presented in the study are avail-able within the manuscript.

Conflicts of Interest

There is no potential conflict of interest in our paper, and allauthors have seen the manuscript and approved to submit to your journal. We confirm that the content of the manuscript has not be-en published or submitted for publication elsewhere.

Jordan, M.I. and T.M. Mitchell, Machine learning: Trends, perspectives, and prospects. Science, 2015. 349(6245): p. 255-260.
Lameras, P. and S. Arnab, Power to the Teachers: An Exploratory Review on Artificial Intelligence in Education. Information, 2022. 13(1): p. 14.
Roll, I. and R. Wylie, Evolution and Revolution in Artificial Intelligence in Education. International Journal of Artificial Intelligence in Education, 2016. 26(2): p. 582-599.
Srinivasan, V., AI & learning: A preferred future. Computers and Education: Artificial Intelligence, 2022. 3: p. 100062.
Cope, B., M. Kalantzis and D. Searsmith, Artificial intelligence for education: Knowledge and its assessment in AI-enabled learning ecologies. Educational Philosophy and Theory, 2021. 53(12): p. 1229-1245.
Radanliev, P., Artificial intelligence: reflecting on the past and looking towards the next paradigm shift. Journal of Experimental & Theoretical Artificial Intelligence, 2024: p. 1-18.
Dowling, M. and B. Lucey, ChatGPT for (Finance) research: The Bananarama Conjecture. Finance Research Letters, 2023. 53: p. 103662.
Hsu, T., et al., Effects of a Pair Programming Educational Robot-Based Approach on Students’ Interdisciplinary Learning of Computational Thinking and Language Learning. Frontiers in Psychology, 2022. 13.
Bin, Y. and D. Mandal, English teaching practice based on artificial intelligence technology. Journal of Intelligent & Fuzzy Systems, 2019. 37(3): p. 3381-3391.
Lai, K.K. and H.H. Chen, A comparative study on the effects of a VR and PC visual novel game on vocabulary learning. Computer Assisted Language Learning, 2023. 36(3): p. 312-345.
Bhutoria, A., Personalized education and Artificial Intelligence in the United States, China, and India: A systematic review using a Human-In-The-Loop model. Computers and Education: Artificial Intelligence, 2022. 3: p. 100068.
Hsu, T., et al., Effects of a Pair Programming Educational Robot-Based Approach on Students’ Interdisciplinary Learning of Computational Thinking and Language Learning. Frontiers in Psychology, 2022. 13.
Wu, R. and Z. Yu, DoAI chatbots improve students learning outcomes? Evidence from a meta‐analysis. British Journal of Educational Technology, 2024. 55(1): p. 10-33.
Belda-Medina, J. and V. Kokošková, Integrating chatbots in education: insights from the Chatbot-Human Interaction Satisfaction Model (CHISM). International Journal of Educational Technology in Higher Education, 2023. 20(1).
Herodotou, C., et al., Empowering online teachers through predictive learning analytics. British Journal of Educational Technology, 2019. 50(6): p. 3064-3079.
Celik, I., Towards Intelligent-TPACK: An empirical study on teachers’ professional knowledge to ethically integrate artificial intelligence (AI)-based tools into education. Computers in Human Behavior, 2023. 138: p. 107468.
Derakhshan, A., F. Moradi and M. Nazari, Conceptualising the role of practice level in language teacher identity construction: An identities‐in‐practice study. International Journal of Applied Linguistics, 2024. 34(2): p. 797-811.
Cath, C., Governing artificial intelligence: ethical, legal and technical opportunities and challenges. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 2018. 376(2133): p. 20180080.
Embretson, S. and J. Gorin, Improving Construct Validity with Cognitive Psychology Principles. Journal of Educational Measurement, 2001. 38(4): p. 343-368.
Roediger, H.L. and M.A. Pyc, Inexpensive techniques to improve education: Applying cognitive psychology to enhance educational practice. Journal of Applied Research in Memory and Cognition, 2012. 1(4): p. 242-248.
Dekeyser, R. and R. Criado, Automatization, Skill Acquisition, and Practice in Second Language Acquisition. 2012.
O'Malley, J.M., A.U. Chamot and C. Walker, Some Applications of Cognitive Theory to Second Language Acquisition. Studies in Second Language Acquisition, 1987. 9(3): p. 287-306.
Printer, L., Positive emotions and intrinsic motivation: A self-determination theory perspective on using co-created stories in the language acquisition classroom. Language Teaching Research, 2023.
Wei, P., X. Wang and H. Dong, The impact of automated writing evaluation on second language writing skills of Chinese EFL learners: a randomized controlled trial. Frontiers in Psychology, 2023. 14.
McEown, M.S. and W.L.Q. Oga-Baldwin, Self-determination for all language learners: New applications for formal language education. System, 2019. 86: p. 102124.
Noels, K.A., et al., Why Are You Learning a Second Language? Motivational Orientations and Self‐Determination Theory. Language Learning, 2003. 53(S1): p. 33-64.
Liu, H., L. Wang and M.J. Koehler, Exploring the intention‐behavior gap in the technology acceptance model: A mixed‐methods study in the context of foreign‐language teaching in China. British Journal of Educational Technology, 2019. 50(5): p. 2536-2556.
Warner, J.A., X. Koufteros and A. Verghese, Learning Computerese. Educational and Psychological Measurement, 2014. 74(6): p. 991-1017.
Hess, T.J., A.L. McNab and K.A. Basoglu, Reliability Generalization of Perceived Ease of Use, Perceived Usefulness, and Behavioral Intentions. MIS Quarterly, 2014. 38(1): p. 1-28.
Liu, G. and C. Ma, Measuring EFL learners’ use of ChatGPT in informal digital learning of English based on the technology acceptance model. Innovation in Language Learning and Teaching, 2024. 18(2): p. 125-138.
Payne, M., Exploring Stephen Krashen's ‘i+1’ acquisition model in the classroom. Linguistics and Education, 2011. 22(4): p. 419-429.
Yuan, Y., An empirical study of the efficacy of AI chatbots for English as a foreign language learning in primary education. Interactive Learning Environments, 2023: p. 1-16.
Kim, J., K. Merrill Jr. and C. Collins, AI as a friend or assistant: The mediating role of perceived usefulness in social AI vs. functional AI. Telematics and Informatics, 2021. 64: p. 101694.
An, X., et al., Modeling students’ perceptions of artificial intelligence assisted language learning. Computer Assisted Language Learning, 2023: p. 1-22.
Huang, A.Y.Q., O.H.T. Lu and S.J.H. Yang, Effects of artificial Intelligence–Enabled personalized recommendations on learners’ learning engagement, motivation, and outcomes in a flipped classroom. Computers & Education, 2023. 194: p. 104684.
Lin, P., et al., Modeling the structural relationship among primary students’ motivation to learn artificial intelligence. Computers and Education: Artificial Intelligence, 2021. 2: p. 100006.
Criollo-C, S., et al., Exploring the technological acceptance of a mobile learning tool used in the teaching of an indigenous language. PeerJ Computer Science, 2021. 7: p. e550.
Derakhshan, A. and J. Fathi, Growth mindset, self-efficacy, and self-regulation: A symphony of success in L2 speaking. System, 2024. 123: p. 103320.
Lai, K.K. and H.H. Chen, A comparative study on the effects of a VR and PC visual novel game on vocabulary learning. Computer Assisted Language Learning, 2023. 36(3): p. 312-345.
Khazaie, S. and A. Derakhshan, Extending embodied cognition through robot's augmented reality in English for medical purposes classrooms. English for Specific Purposes, 2024. 75: p. 15-36.
Vincent-Lancrin, S. and R. van der Vlies, Trustworthy artificial intelligence (AI) in education. 2020.
Crawford, J., et al., When artificial intelligence substitutes humans in higher education: the cost of loneliness, student success, and retention. Studies in Higher Education, 2024. 49(5): p. 883-897.
Memarian, B. and T. Doleck, ChatGPT in education: Methods, potentials, and limitations. Computers in Human Behavior: Artificial Humans, 2023. 1(2): p. 100022.
Robinson, S.C., Trust, transparency, and openness: How inclusion of cultural values shapes Nordic national public policy strategies for artificial intelligence (AI). Technology in Society, 2020. 63: p. 101421.
Knox, J., Artificial intelligence and education in China. Learning, Media and Technology, 2020. 45(3): p. 298-311.

No competing interests reported.

The Role of GenAI in EFL: Impact on Learning Motivation and Outcome

Status:

Version 1

Abstract

I. Introduction

II. Literature review

III. Research Design and Data Analysis

IV. Results and Data Analysis

V. Discussion

VI. Conclusion

Declarations

References

Additional Declarations

Status:

Version 1