Thank you for using the Interlibrary Loan Services at the University of Illinois at Urbana-

Champaign Library. This service is free to you. Please ignore any dollar amounts that you
might see on the photocopy cover page.

IS SOMETHING WRONG WITH THIS PHOTOCOPY?

We hope not! But mistakes can happen. If there is a problem with this photocopy contact us and
we will fix it as quickly as possible. Please do not place a new request for the same item.

You may contact us by:

• Email: borrowing@library.illinois.edu
• Phone: (217) 333-0832

Provide the following information:

• The transaction number (TN #) of your request (preferred), or the full article title exactly
as you requested it
• Your full name or NetID
• Describe the problem clearly (i.e. are pages missing, pictures too dark, or is the text
illegible?)

WARNING CONCERNING COPYRIGHT RESTRICTIONS:

The copyright law of the United States (Title 17, United States Code) governs the making of photocopies
or other reproductions of copyrighted material.

Under certain conditions specified in the law, libraries and archives are authorized to furnish a photocopy
or other reproduction. One of these specified conditions is that the photocopy or reproduction is not to be
“used for any purpose other than private study, scholarship, or research.” If a user makes a request for, or
later uses, a photocopy or reproduction for purposes in excess of “fair use,” that user may be liable for
copyright infringement.

This institution reserves the right to refuse to accept a copying order if, in its judgment, fulfillment of the
order would involve violation of copyright law.

See http://www.copyright.gov/ for more information.

 Interactive Learning Environments

ISSN: 1049-4820 (Print) 1744-5191 (Online) Journal homepage: www.tandfonline.com/journals/nile20

An empirical study of the efficacy of AI chatbots

for English as a foreign language learning in
primary education

Yijia Yuan

To cite this article: Yijia Yuan (2024) An empirical study of the efficacy of AI chatbots for English
as a foreign language learning in primary education, Interactive Learning Environments, 32:10,
6774-6789, DOI: 10.1080/10494820.2023.2282112

To link to this article: https://doi.org/10.1080/10494820.2023.2282112

Published online: 13 Nov 2023.

Submit your article to this journal

Article views: 3635

View related articles

View Crossmark data

Citing articles: 28 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=nile20
INTERACTIVE LEARNING ENVIRONMENTS
2024, VOL. 32, NO. 10, 6774–6789
https://doi.org/10.1080/10494820.2023.2282112

An empirical study of the efficacy of AI chatbots for English as a

foreign language learning in primary education
Yijia Yuan
Faculty of Education, University of Cambridge, Cambridge, UK

ABSTRACT ARTICLE HISTORY

This experimental research examined the effectiveness of using chatbots Received 30 May 2023
in English as a Foreign Language (EFL) classrooms at a Chinese elementary Accepted 6 November 2023
school. Seventy-four students were divided into two groups: one
KEYWORDS
employing traditional methods, and the other using chatbots. Before Elementary education;
and after the 3-month teaching period, pre- and post-tests were used chatbots; artificial
to measure students’ oral English proficiency, and a questionnaire was intelligence; oral English
employed to assess their willingness to communicate (WTC). Data from teaching; willingness to
interviews with teachers and students provided insights into chatbot communicate; function
usage and modification. Data analysis involved using SPSS to identify optimization
significant differences in pre- and post-test scores and NVivo 11 to code
qualitative feedback from student and teacher interviews. The results
show that chatbot integration significantly improved oral English
proficiency and WTC in the experimental group compared with the
control group. Teachers can enhance instruction by adopting tailored
chatbot features to refine teaching methods, thus creating a more
adaptive learning path for students. Further, chatbot improvements
such as personalised error correction and user-centric features pave the
way for immersive and fruitful language learning. This study provides
valuable insights for integrating chatbots into language classrooms and
suggests opportunities for further chatbot enhancements..

1. Introduction
Artificial intelligence (AI), hailed as the ultimate future technology, is revolutionising multiple fields
in the era of Industry 4.0. Artificial intelligence has been recognised as a potentially game-changing
technology, and it is already penetrating a variety of fields, including EFL instruction. According to
UNESCO, AI has the potential to address some of the biggest challenges in education today, opening
up new avenues for teaching and learning practice in the classroom (UNESCO, 2021). AI can broadly
customise the educational experience for every learner and provide learners with authentic and
immersive experiences in a way that is easier than ever before (Bailey, 2019). It is more specific
and advanced than other technologies, enabling computers to perceive, associate, predict, and
plan what the human brain can do (Boden, 2016).
Within the field of AI-enhanced education, language education is a major area of interest. Few
domains may experience technology advancement as keenly as the second and foreign language
learning (Blake, 2013). Additionally, the essence of how technology could aid in language learning
is via conversational interaction with intelligent agents (Fryer et al., 2017). Therefore, chatbots,
already widely acknowledged as the future of the Internet (Fryer et al., 2020), are emerging gradually.
The chatbot is an artificial intelligence capable of interacting with users and processing their input

CONTACT Yijia Yuan layla.yj.yuan@outlook.com

© 2023 Informa UK Limited, trading as Taylor & Francis Group
 INTERACTIVE LEARNING ENVIRONMENTS 6775

using natural language (Huang et al., 2022). The chatbot is able to conduct conversations around a
specific topic through audio or text (Abdul-Kader & Woods, 2015; Shevat, 2017). ELIZA is considered
to be the first chatbot developed by a German computer scientist, Joseph Weizenbaum, in 1966.
Recent years have witnessed the application of chatbots in language education contexts, including
utilising chatbots for daily conversation practice (Hua et al., 2017), carrying out assessments and pro-
viding feedback (Read, 2000; Zhang et al., 2019), and answering language learning questions (Xu
et al., 2021). Several preceding investigations have suggested chatbots as a potentially powerful
tool for empowering learners’ language learning, providing a more engaging and genuine linguistic
setting, and improving students’ language learning outcomes (Lin & Chang, 2020; Wang et al., 2017).

1.1. Statement of the problem

Although the issue of AI language education has received considerable critical attention, most
studies in the field of AI chatbot-enhanced language education have only focused on higher edu-
cation (Huang et al., 2022). What is not yet clear is the impact of a chatbot on children receiving
primary language education who are in the critical period of language learning. The early develop-
ment of children’s language ability plays a decisive role in their lifelong development (Johnson &
Newport, 1989). Besides, much of the research up to now has tended to centre on AI chatbot’s
effect on students’ receptive skills, such as reading (Xu et al., 2021), vocabulary (Jia et al., 2012;
Kim, 2018), and grammar (Kim et al., 2019), rarely assessing students’ productive skills such as
oral English proficiency. Moreover, willingness to communicate (WTC) is acknowledged as a funda-
mental element that plays a substantial role in fostering effective foreign language learning (Zhang
et al., 2018). While increasing learners’ WTC has been an ongoing struggle for a large number of tea-
chers of second language (Peng, 2019). The WTC of L2 learners has been shown to be impacted by a
range of factors. However, it’s worth noting that the majority of L2 WTC studies have concentrated
on interpersonal interactions in traditional, non-digital settings, rather than examining WTC within
an AI-enhanced context.
It will thus be instructive to assess the impact of AI chatbots on primary-level kids’ WTC and their
spoken English skills. The purpose of this study is to compare the learning performance of experimen-
tal and control groups of primary school pupils in terms of spoken English competence and willingness
to communicate (WTC). This research offers multiple critical insights into promoting children’s
language learning in the era of artificial intelligence, as well as providing a more practical basis
through experiments for the large-scale implementation of intelligent education in the future. Fur-
thermore, this is the first study to undertake an empirical study to assess AI chatbots’ effects on
primary school students English learning performance in the Chinese educational context. Moreover,
with the emergence of AI chatbots, many teachers have yet to find a way to adapt to the new tech-
nology (Linh et al., 2022). Teachers also face great challenges in adapting to the new software and
adjusting teaching programs (Trust & Whalen, 2021). It is necessary to investigate how they can effec-
tively remodel themselves to integrate AI chatbots into classrooms more seamlessly. Last but not least,
while a multitude of AI language education products are readily available, there is often a lack of atten-
tion directed toward the modification of these applications. It’s important to examine how these tools
can be further enhanced to meet the needs of language learners more effectively.

2. Literature review
2.1. The application of chatbots in language learning
In general, recent years have witnessed increasingly more attention on chatbots for language learn-
ing within the AI-integrated education realm. Previous studies about AI-enhanced English learning
mainly focus on its efficacy on learners’ language ability and learning motivation. Overall, studies
have shown that chatbots can positively promote students’ skills in grammar, listening, vocabulary,
6776 Y. YUAN

and writing (Jia et al., 2012; Kim, 2018; Kim et al., 2019; Lin & Chang, 2020). However, previous studies
failed to prove chatbots’ efficacy on learners’ reading comprehension. Xu et al. (2021), for instance,
compared chatbot-assisted conversations (experimental group) with human conversations (control
group). The chatbot was tasked with guiding the youngsters in the experimental group through the
story by way of questions, and human teachers led the children to read in the control group. Results
have shown that there is a similar effect in facilitating children’s reading comprehension (p = .29)
between human teachers and chatbots. Similarly, Kim (2016) integrated chatbots into college stu-
dents’ schoolwork as language-practice companions. Students interact with the chatbot “Elbot”
via text or audible messages on their mobile devices. However, there was no obvious change in
their reading comprehension after the intervention.
Much of the current literature on chatbots and language learning also pays particular attention to
learners’ learning motivation. Jia and Chen (2008), for example, showed the teaching function of a
chatbot named CSIEC in English learning, demonstrating that CSIEC may inspire students to study
and practise English. Wang et al. (2013) carried out experiments on chatbots that they had built
and found that these chatbots had a positive impact on the learner’s engagement and motivation
in language learning, particularly for learners with low levels of competence. Thus, recent evidence
suggests that AI-chatbot could exert a positive impact on students’ learning motivation.

2.2. AI-chatbot and oral English speaking proficiency

In terms of oral English speaking, Sha (2009) created an AI-based chatbot and conducted classroom
surveys to assess the benefits of these chatbots as a tool for teaching spoken English. The limitation
is that it only measures the students’ motivation and does not indicate the usefulness of this
machine in improving oral English proficiency. Some pre-tests and post-tests can be added to evalu-
ate the practicality of chatbots. Green and O’Sullivan (2019) performed a large-scale study on the
improvement of students’ spoken English language ability among 746 adult users of Liulishuo, an
artificial intelligence programme created for English learners in China that functions similarly to Duo-
lingo. In around two months, 72 per cent of adult users displayed increased test performance,
according to the findings. Zou et al. (2020) conducted a survey on the use of AI-spoken English learn-
ing software by Chinese college students, including Liulishuo and Real Skills. They found that stu-
dents were satisfied with the scores and feedback given by this AI software for their oral practice,
which was helpful for extracurricular oral English learning. Other authors (Kim et al., 2019)
focused on the availability and fluency of those applications. When analysing the effectiveness of
two artificial intelligence chatbots, namely Google Assistant and Alexa, in secondary school
English classes, the researchers discovered that both chatbots had a high response rate of over 86
percent.
Above all, it is possible to conclude that chatbot has already been recognised as a welcoming
language-learning facilitator. However, until recently, there has been little reliable evidence from
empirical studies to show that it can exert a positive influence on children’s oral English proficiency,
especially those who are in the critical period of language learning. Just as Li and Min (2010) noted,
there is still a large gap to be bridged between AI and spoken English proficiency, including the accu-
racy of learners’ pronunciation and so on.

2.3. Willingness to communicate (WTC) and L2 learning

In today’s second language lessons, one of the most common challenges is that many students are
hesitant to speak in the target language because they are embarrassed by the possibility of making
errors in front of their classmates and worried about being judged negatively (Tai & Chen, 2023). As a
result, increasing learners’ WTC has been an ongoing struggle for a large number of teachers of
second language (Peng, 2019). WTC refers to the learner’s readiness to use the L2 to engage in com-
munication with one or more specific individuals at a precise moment (MacIntyre et al., 1998). WTC
 INTERACTIVE LEARNING ENVIRONMENTS 6777

has always been seen as an essential component of second language learning, with many studies
aimed at exploring effective ways to enhance WTC.
MacIntyre et al.’s (1998) multilayered pyramid model of willingness to communicate (WTC) serves
as the theoretical basis of this study. The model of a pyramid has six levels. The top (communication
behaviour, behavioural intention, and situated antecedents) shows how WTC is affected by the
current situation, while the bottom (motivational propensities, affective–cognitive context, and
social and individual context) reveals how the process is affected by stable, long-lasting factors.
Empirical studies of L2 WTC have shown that it is linked to many other factors. One of the most
important parts of L2 WTC is second language confidence, which is the belief that you can commu-
nicate in a second language in an adaptable and effective way (MacIntyre et al., 1998). Several
studies have indicated that the most essential component in determining L2 WTC is the speaker’s
level of self-confidence (Ghonsooly et al., 2012; Peng & Woodrow, 2010; Yashima, 2002). L2 WTC
has also been associated with anxiety (Cha & Kim, 2013; Yu, 2011), motivation (Hashimoto, 2002),
and so on.
As more and more students practise English in digital settings, work has been done recently to
make a connection between the L2 WTC and computer-assisted language learning (CALL). For
instance, 30 EFL university students in Thailand participated in research by Reinders and Sorada
(2014), where they spent nine hours per week for six weeks playing a multiplayer online role-
playing game. Participants in the intervention course not only had a decrease in L2 anxiety but
also saw improvements in L2 WTC and L2 confidence. The two researchers found that EFL students
tend to be quieter in classrooms when they are learning a second language because they are fre-
quently anxious about the possibility of making mistakes and receiving negative feedback from
both their instructors and their classmates. By contrast, in digital environments, they generally
feel “safe,” “relaxed,” and “open,” resulting in more interaction with other English users.
The WTC of L2 learners has been shown to be influenced by a variety of circumstances; however, it
should be emphasised that most L2 WTC research has focused on interpersonal interactions in non-
digital environments rather than WTC in a digital context. Because chatbots can react fast to requests
from users, some academics believe that the introduction of AI will greatly improve learners’ WTC
(Dizon, 2020; Underwood, 2017). The possible contribution of chatbots to L2 seems to be an impor-
tant but unexplored area of research. Therefore, this investigation aims to examine the effects of
chatbots on children’s L2 WTC. These results ought to help considerably toward a better understand-
ing of the potential of chatbots in second language learning.
In conclusion, from previous research, it can be found that AI has become a well-recognised tech-
nology, especially integrated with language learning or teaching. However, three major problems in
relevant studies can still be noticed:
First, in terms of the age distribution of subjects, nearly 80% of the studies were carried out in
universities, and few studies were carried out in the education of lower age groups (kindergarten/
primary/secondary school) (Huang et al., 2022). While children are in the critical period of language
learning, the early development of children’s language ability plays a decisive role in their lifelong
development (Johnson & Newport, 1989). Therefore, it is crucial to study the effectiveness of AI in
language education at this stage. Filling this gap can have a breakthrough significance for facilitating
children’s language learning in the era of big data.
Second, there are too few empirical studies on this subject, and most of these results are based on
single experimental studies, which should be interpreted cautiously. The effect of AI on language
education still requires more research, especially in terms of oral English proficiency (Zou & Wang,
2021). Also, the AI chatbot’s efficacy in the L2 WTC still remains unknown.
Third, although there are already numerous AI language education products, such as Duolingo,
Cleverbot, and Mondly, the modification of such applications is always overlooked. There is still
improvement room for AI to personalise language learning instruction. For instance, error correction
systems and adaptive learning systems can be developed. Teachers also face great challenges in
adapting to the new software and adjusting teaching programs.
6778 Y. YUAN

Therefore, this study aims to determine the extent to which an AI chatbot influences the oral
English competence and WTC of primary school children compared to conventional learning
methods. The following research questions serve as the basis for this study:

1. How effective are AI-chatbots in improving children’s oral English proficiency and willingness to
communicate?
2. How can teachers remodel themselves to integrate AI chatbots into classrooms better?
3. What are the ways to further modify the artificial intelligence chatbot?

3. Methodology
3.1. Participants
The research setting was one primary school in Huludao City in Northeast China, which is on a six-
year cycle. The students in the study started to learn English in Grade 3. This study employed the
purposeful sampling strategy in order to concentrate on students with specified qualities who will
contribute more to the relevant research (Etikan, 2016). Finally, two English teachers and 74 students
from Grade 5, approximately 12 years old (M = 11.89 years, SD = .35), participated in this study.
Included participants were native Chinese speakers studying English as beginners (with little to
no exposure beyond kindergarten), without previous AI chatbot exposure. All of them possessed
an intermediate level of English language proficiency (scored 80–95 on the final English exam).
Experimental (n = 39) and control (n = 35) groups of students were formed. An AI chatbot assisted
the experimental group in learning English, communicating with students for 10 min three times
a week in English classes. In comparison, the conventional teaching method was used by the
control group, which was mainly based on the textbook and teachers. The practising part of the
control group was mainly conducted between students and their peers instead of AI chatbots.
Before the study, permission was given by the school and the parents. Table 1 highlights the stu-
dents’ details.
These students’ use of spoken English was limited in English classes. Their English teachers found
that many students were reticent when communicating in English, which might be brought on by a
lack of confidence, a lack of English competence, and personality traits (e.g. introversion or shyness)
(Tai & Chen, 2023).

3.2. Instruments
3.2.1. Ai chatbot: mondly
In 2016, Mondly was launched as the first chatbot developed specifically for language learning,
unlike bots like Cleverbot, which was designed as a means to entertain users or Alexa/Siri, acting
as digital assistants for humans. Mondly operates as a smartphone-centred software, with over
100 million downloads from 190 countries. Mondly recognises millions of inputs and provides feed-
back according to its powerful phrase database.

Table 1. Students’ details.

Class A Class B
Age 11/12 11/12
Number of subjects 39 35
Total number 74
Students’ L1 Chinese
Students’ L2 English
Years of learning English 2.5
 INTERACTIVE LEARNING ENVIRONMENTS 6779

What distinguishes the Mondly chatbot from others is that it enables users to practise the
language in real-world situations, such as making restaurant orders, checking in the hotel, and
greeting other people. And it can provide recommended answers so that users are allowed to
efficiently master real-life conversations in diverse situations. The application interface is shown
in Figure 1. In conclusion, Mondly is specially developed for language learning, and the learning
content is practical and moderate in difficulty, which is suitable for primary school students
English level. Moreover, this chatbot is user-friendly, simply becoming operable after being down-
loaded on a mobile phone or tablet. Due to the reasons above, this experiment chose Mondly as the
experimental instrument.

3.2.2. Background survey

A background survey was distributed to students’ parents to select participants who were best
suited to the research. The background survey comprised three parts. Students’ basic demographic
data, including gender, age, and contact information, were provided in the first section. Part two’s
questions focused on participants’ English language scores on the final English exam and their
spoken English ability. Part three asked whether students had using history of AI-integrated edu-
cational software.

3.2.3. Oral English proficiency test

There are eight cards in the test, each with one question on it. Students took turns choosing a topic
from three cards to discuss. They won’t be compelled to discuss a subject about which they might
not have any opinions. Five criteria were used to evaluate their performance.: smoothness of utter-
ances, meaning construction, comprehensibility, pronunciation, and accuracy (Abd Al Galil & Abd Al
Galil, 2019). Pearson’s correlation coefficient was calculated to be 0.92, suggesting that the test was
highly reliable. On a four-point scale, each of the five components was rated, with “1” denoting poor
performance and a “4” indicating exceptional performance.

Figure 1. Examples of Mondly chatbot interface.

6780 Y. YUAN

3.2.4. Willingness to communicate scale

The L2 WTC scale was adapted from Lee and Chen Hsieh (2019) and Yashima (2009). It assesses L2
WTC in-class (4 items) and out-of-class (4 things) on a five-point Likert scale, ranging from 1
“Definitely not willing”, to 5 “Definitely willing.” Cronbach’s alpha coefficients for the two question-
naire portions were 0.91 and 0.86, showing strong internal consistency.

3.2.5. Semi-structured interview and focus group interview

In this research, semi-structured interviews and a focus group interview were employed to delve into
the perspectives of teachers and students regarding the utilisation of AI chatbots in English language
learning based on questions derived from relevant literature (e.g. Jeon, 2022; Yang et al., 2022; Zou
et al., 2020). The interviews followed four pivotal themes: “Perceived Efficacy,” “Usage Experience,”
“Improvement Suggestions,” and “Language Learning Impact,” all of which underwent rigorous vali-
dation by a linguistic expert and an AI education specialist to ensure their accuracy and reliability.
Spanning a 45-minute interview with teachers and a 60-minute focus group discussion with stu-
dents. The ten open-ended questions for teachers explored their perceptions of the benefits and
drawbacks, and extracted insightful suggestions pertaining to modifications of AI chatbots in
English teaching. Similarly, students were engaged with a set of 7 questions, modified to accommo-
date their comprehension levels, seeking to extract their perspectives and recommendations for
chatbot-aided English learning.

3.3. Data collection

This study employed a mixed-methods approach. On this basis, quantitative data were collected
through pre- and post-tests and data-collecting questionnaires. After that, qualitative data were
gathered via focus group interviews with students and through two semi-structured interviews
with two teachers. The data sources and collection methods are presented in Table 2.

3.4. Research procedure

The main study adopted a mixed-methods approach, and students were allocated to the experimen-
tal and control groups. The experiment lasted for three months. The experimental class adopted an
AI-integrated pedagogy, communicating with chatbots for 15 min three times a week apart from the
normal English teaching process in class. Each student spent around 10 min taking an oral English
competency test before and after the experiment. Participants were also required to answer the
WTC questionnaire. Finally, a focus group interview with five consenting participants was done fol-
lowing the 3-month experiment to get more insights from the students’ experiences using chatbots
for learning. The participants’ native language, Mandarin, was used for the interview. The general
procedure of the experiment is shown in Figure 2.

3.5. Data analysis

Data analysis of the quantitative data was performed using SPSS (v25.0). The results of the partici-
pants’ pre-tests and post-tests on their spoken English ability were compared in order to evaluate
whether or not AI chatbots had an effect on the participants’ English learning. The ANCOVA and

Table 2. Data sources and methods of data collection.

Data sources Data collection
Students Oral English proficiency Pre- and post-tests
Willingness to communicate Questionnaire
Attitudes Focus group interview
Teachers Attitudes Semi-structured Interview
 INTERACTIVE LEARNING ENVIRONMENTS 6781

Figure 2. The experimental procedure.

the paired-samples t-test were used to determine whether pre- and post-tests differences were sig-
nificant. For the quantitative data, NVivo 11 was used for coding after the focus group interview.

4. Results
4.1. Oral English proficiency
4.1.1. Quantitative results
Using a one-way ANCOVA with the post-test score as the dependent variable, the learning method
as the independent variable, and the pre-test score as the covariate, spoken English competency was
evaluated and compared between the two groups at the post-test. The regression homogeneity
hypothesis was fulfilled, with F = 1.935 (p = .168 > 0.05), showing that using ANCOVA to examine stu-
dents’ oral English competency was viable. Table 3 summarises the findings.
As shown, a significant between-group effect on oral English proficiency (F = 10.89, p = <0.05, η²
= 0.133 > 0.059) was found. Thus, the result indicated that the students (Mean = 2.32, Std. error =
.11) in the AI group outperformed those in the traditional group (Mean = 2.11, Std. error = .09). Fur-
thermore, the effect size of different learning methods on spoken English competence was medium,
according to Cohen’s illustration (η² > 0.059).

4.1.2. Qualitative results

To supplement these quantitative findings, participant reflections are explored. Their opinions are as
follows. The students, as well as teacher insights, illuminate the underlying factors which might have
contributed to the superior outcomes in the AI group.
Reducing Anxiety: My brain always turns blank when I speak English, so I never answer teachers’ questions in
class. But now, I feel that some ideas or words start to pop into my head when I want to express something

Table 3. The analysis of the ANCOVA on students’ oral English proficiency.

Variable Group N Mean SD Std. error F η² ig
Oral English proficiency Experimental group 39 2.32 0.69 0.11 10.89 0.133 .002
Control group 35 2.11 0.60 0.09
6782 Y. YUAN

in English. Because when I talk to the chatbot, it can gradually lessen the anxiety when I speak English and teach
me how to say something correctly. (Student No. 15)

Enhancing Fluency and Pronunciation: Talking with the chatbot is really an interesting experience, and I feel less
nervous than before when I talk with other people in English. And I like imitating what the chatbot says, like its
accent, which is also really helpful for my fluency and pronunciation. (Student No. 39)

Enriching Vocabulary and Facilitating Usage: I learned many new words when I talked to the chatbot. I used to
have a very limited vocabulary, so when I talked to others, I always spent a long time thinking about the proper
word. But now, things are getting better with the help of a chatbot. I even teach my parents how to order in the
restaurant in English when having dinner, which makes me more motivated to practice my oral English speaking.
(Student No. 43)

Implementing and Retaining Learned Vocabulary: The chatbot is really like a great teaching assistant for me. Stu-
dents talking to the chatbot can use the word they have learned in English class that day so as to deepen their
memorisation of the English word and put it to good use. (Teacher No. 1)

Student comments, like those from No. 15 and No. 39, show a clear comfort and confidence boost in
using English after interacting with chatbots. Particularly, the anecdote of ordering food in English
suggests a transfer of chatbot-assisted learning to practical, everyday application, reflecting not
merely an academic improvement but also a functional utility in real-world contexts. Teacher No.
1 also sees a positive change, highlighting that chatbots help solidify class learning by offering
extra practice in a stress-free way.

4.2. Willingness to communicate

4.2.1. Quantitative results
To assess if a significant difference occurred between the experimental and control groups, a one-
way ANCOVA was performed using the post-test score as the dependent variable, the learning
method as the independent variable, and the pre-test score as the covariate. The findings are
shown in Table 4.
A paired-samples t-test further indicated significantly higher WTC amongst the experimental
group (p = <0.01). Table 5 displays the results.
As shown in Table 4, the analysis found a significantly higher WTC in the experimental (AI) group
in the post-test, p = <.001. Results thus suggest that AI-assisted teaching significantly improves stu-
dents’ WTC compared with traditional methods. Opinions from Student No. 18, No. 22, and Teacher
No. 2 also reflect this outcome to a certain extent.

4.2.2. Qualitative results

To provide additional depth and context to these quantitative findings, experiences and perceptions
from several participants are highlighted. Their thoughts are shared below.
Boosting Confidence: I feel that I am not as afraid of making mistakes as before because I often talk with chatbots,
which makes me more confident. I pushed myself to say one sentence correctly at once, but now I allow myself
to think about it for a second or make some mistakes. Making some mistakes when speaking English is not a
shameful thing, right? (Student No. 18)

Maintaining Sustained Conversations: I really like talking to the chatbot because its English level is better than
mine, but I can still understand what it says. That makes me feel that we can keep a long talk. Interacting
with it is funny and helpful to my oral English. (Student No. 22)

Table 4. ANCOVA for WTC Results between two groups.

Class/Group N Mean F Asymp. Sig. (2-tailed) η²
Experimental (Artificial Intelligence) 39 3.57 103.489 .000 .593
Control (Traditional) 35 2.31
 INTERACTIVE LEARNING ENVIRONMENTS 6783

Table 5. A paired-samples t-test of the experimental group (AI).

Class/Group N t df Asymp. Sig. (2-tailed)
Experimental (Artificial Intelligence) 39 −10.15 38 .000

Enhancing Classroom Dynamics: Before using the chatbot, my questions often encountered silence in English
class, especially when I said them in English. They are all shy and reluctant to use English. So I have to force
one or two students to answer it. But now, more and more students are willing to raise their hands, and the
classroom has become more active now. (Teacher No. 2)

The narrative of Student No. 18 and Student No. 22 reveals that the non-judgmental interaction with
chatbots provides a secure and encouraging environment, potentially reducing anxiety and foster-
ing a heightened WTC in English. This is mirrored in the changed classroom dynamics observed by
Teacher No. 2, where an increase in hand-raising and active participation notably contrasts with the
prior hesitancy. In conclusion, through the lens of both quantitative data and rich student-teacher
narratives, it becomes palpable that integrating AI in English learning not only statistically improves
oral proficiency but also naturally nurtures a conducive learning environment, fostering heightened
willingness and confidence among students to communicate in English.

5. Discussion
This study sought to investigate the effect of AI chatbots on Chinese primary school students’ oral
English proficiency and willingness to communicate. Gathered data were analysed to determine the
effect of the intervention on the spoken skills and the WTC of the students. The results revealed that
the students learning with chatbots obtained higher post-test scores on oral proficiency and the
WTC questionnaire relative to their pre-test scores.

5.1. Q1: how effective are AI-chatbots in improving children’s oral English proficiency and
willingness to communicate compared with traditional teaching methods?
In terms of oral English ability, the current findings appear to be in line with Green and O’Sullivan’s
(2019) research, proving AI chatbot’s efficacy in improving learners’ oral English proficiency. In the
aspect of L2 WTC, This finding supports previous research into this area, which links WTC and AI
(Ayedoun et al., 2015; Tai & Chen, 2023). Specifically, this study bridged the gap in AI’s usefulness
to children in the critical period, which provides sufficient practical data for the future implemen-
tation of intelligent education in the Chinese primary education system.
Learners require an immersive social context as well as a high quantity of output in order to feel
satisfied and confident, lower their anxiety levels, and ultimately increase their L2 WTC (Tai & Chen,
2023). As shown in the WTC pyramid, situational antecedents are thought to have an impact on the
WTC. Baker and MacIntyre (2000) investigated the nonverbal outcomes of French immersion and
non-immersion programmes. Immersion students exhibited stronger WTC and more frequent
French conversation, as predicted. However, the majority of L2 students lack access to such an
immersed social context. Due to the lack of a suitable dialogue environment, improving learners’
WTC in L2 is a difficult problem to overcome. Nonetheless, several studies have demonstrated
that AI-chatbot learning settings have the potential to be effective substitutes for real human inter-
action, providing learners with an immersive environment. This research, together with several
related ones, lays a solid practical foundation for the future development of the AI chatbot.
Furthermore, researchers usually approach AI-integrated education from two perspectives, which
are inherited from the wider field of second language acquisition (SLA) and development: cognitive
and sociocultural perspectives. This research adopted a sociocognitive perspective (Atkinson, 2002),
trying to transcend their difference and fruitfully taking advantage of both of them. Two main differ-
ences between the two theories mainly distinguish them. In terms of the research object, the
6784 Y. YUAN

cognitive approach (Gass et al., 2007; Long, 1997) maintained that the research object was second
language acquisition rather than second language (L2) use. These researchers all clearly stated that
the primary task of SLA research is to investigate how learning occurs, especially to find out which
psychological processes and learner factors contribute to language acquisition and how to make
learning happen through the use of language. They opposed expanding the scope of the study
of SLA to the use of L2 in real contexts. In other words, this theory advocates that SLA should not
be affected by the learning environment and individual differences of learners. However, the socio-
cultural approach advocates that the research object of SLA is the use of L2 rather than language
acquisition. Firth and Wagner (2007) argued that language acquisition must be usage-based. The
study of L2 use is to investigate how L2 is used in social contexts. Learning and using are an insepar-
able continuum. In the aspect of the research method, the cognitive approach adopts the quantitat-
ive method, emphasising objectivity and justice, and opposes researchers’ personal views. The
sociocultural approach mostly uses qualitative methods, emphasising the interaction between
researchers and subjects. It requires researchers to understand and explain social communication
events from the perspective of the subjects.
This research adopted the combination of the two major perspectives – the sociocognitive
approach to analyse the outcome. The sociocognitive model assumes that input, output, and
uptake largely depend on situational contexts and the kind of activity involved in learners’ inter-
action. In the sociocognitive model of language learning, the AI chatbot acts as the mediator, facil-
itating social interaction with learners, generating input, and eliciting outputs from learners. In
addition, the AI chatbot provides a learner-centred approach to L2 learning (Ko, 2012). Therefore,
learners in AI environments have more time to reflect on their output and monitor their speech,
which in turn promotes the coherence of discourse.

5.2. Q2: how can teachers remodel themselves to integrate chatbots into classrooms
better?
From the semi-structured interview with two teachers, it was found that they generally hold a scep-
tical and slightly resistant attitude toward the integration of AI chatbots with their teaching. That
might be due to the reason that they are not familiar with its operation and have little idea about
AI and its efficacy. Incorporating chatbots into the classroom requires the teachers to find the
right fit with chatbots, taking the initiative to understand how chatbots are used and the
different functions that different chatbots have.
Existing chatbots fall into two categories. One is that none of the Settings can be changed, and
the user can only passively accept the topics discussed with the chatbot. The other is the ability to
customise the bot’s responses as needed. AI chatbot developers have recently sought to improve
their capacity to respond to unstructured input from end users. Google’s Dialogflow, for example,
allows users to customise conversations by adding a default database. Users may drag and drop con-
versation streams using the online chatbot platform BotStar, using a dashboard that enables tea-
chers to script their students’ learning experiences and tailor their lesson plans depending on
desired learning objectives. With this software, lesson plans and chatbots can be better integrated.
Teachers can choose their adoption of the software according to their needs.
Commentators on chatbots’ current ability to communicate often argue that chatbots communi-
cate beyond a few times (Höhn, 2017; Knight, 2017), have not yet developed coherent communi-
cation, and usually, only one question-and-answer exchange takes place. User attention may drop
off quickly when a chatbot doesn’t react effectively or doesn’t continue to think (Fryer et al.,
2020). More potent chatbots will undoubtedly be the solution to this issue, but until then, deploying
numerous chatbots simultaneously may be able to cover the gap (Vasconcelos et al., 2017). Learners
may benefit from using multiple chatbots that each offer unique responses and pose unique ques-
tions to ensure they receive adequate information (which is crucial for language learning). Utilising
the ideal blend of chatbot personalities and creating a clear mechanism for effective team
 INTERACTIVE LEARNING ENVIRONMENTS 6785

interaction are two important factors in the success of this approach. This demonstrates how con-
temporary technologies may be used more effectively.

5.3. Q3: what are the ways to further modify the artificial intelligence chatbot?
5.3.1. Adaptive and personalised learning system
The main objectives of an intelligent learning environment have always been personalised learning
and adaptive learning (Peng, 2019). Confucius’ suggestion to “teach pupils according to their ability”
and Socrates’ heuristic teaching approach can be linked to the wisdom of personalised learning. The
goal of differentiated teaching is to focus on individual differences through customised learning and
adaptive learning. The American Institute for Supervision and Curriculum Development (ASCD)
explains this approach: “Differentiated instruction is an instruction in which educators actively
plan students’ differences so that all students can learn best.” To effectively educate, teachers in
a classroom with varied instruction devote time, resources, and effort to students with various back-
grounds, preparations, skill levels, and interests. It customises instruction to fit each learner’s unique
learning requirements and circumstances while also tailoring learning activities and information to
each individual.
Unquestionably, personalised learning is also the goal that AI chatbots have been pursuing in
development. In practice, there is still great room for improvement for chatbots. Establishing an
error correction memorisation system is one of the personalised function-modification alternatives.
The bot is able to remember what mistakes users are prone to make and often mentions them in the
following learning process. For example, if a user frequently mistakes the masculine or feminine
property of the word “flower” in German, the chatbot could reinforce its memory by reminding
users of situations in which the word might be used so as to deepen the memorisation of the
word property.

5.3.2. User archive

It is also useful if the user archives can be read at any time when users use them. In that way,
advanced courses can be set up to match communication difficulty and complexity based on the
user level. Stephen Krashen termed comprehensible input as input that is slightly beyond the
current level of competence of the language learner (Krashen, 1982). Therefore, if the objective is
to aid the advancement of the language learner, it is vital to supply comprehensible input “i + 1.”

5.3.3. Celebrity chatbot

A major problem of educational technology is attracting and maintaining learners’ interest (Fryer
et al., 2017). The novelty effect and the danger that language learners quickly lose interest is a
serious problem. Learners may be more likely to receive the necessary language practice if they
can speak with their favourite public figures.

5.3.4. Extensive knowledge of a specific topic

Chatbots with in-depth knowledge of particular subjects that users would find valuable. These sub-
jects might include certain sports, nations, movie genres, health, or automobiles. Learners are more
inclined to overlook the chatbot’s shortcomings if they discuss subjects that they are personally
interested in.

5.3.5. Elements that mimic real people

Users learn better when engaging with a genuine voice, according to Mayer (2017); therefore, devel-
opers may design their own speech synthesis instead of depending on the built-in speech synthesis
on platforms like Android and iOS. It could be crucial for chatbots to have a distinctive voice that fits
their identity and age. It may be important for chatbots to have a unique voice that suits their age
and personality. In addition, the chatbot can change its appearance and clothes according to the
6786 Y. YUAN

user’s age to create a sense of realism for the user. And adding gestures and expressions to the bot
will also help add realism and fun. While adults might find it tedious, youngsters might find this
amusing.

5.4. Implication and limitation

In general, the study’s findings can offer pedagogical suggestions for effective L2 teaching. The
examination of the sociocognitive approach to SLA was conceptually expanded by the study. Fur-
thermore, empirical research demonstrated the chatbot’s potential to improve learners’ spoken
English proficiency. This would enrich AI-integrated educational studies and provide some sugges-
tions for future studies on the usage of chatbots in English learning. However, due to the relatively
small sample size, the findings may not be generalisable to all elementary schools. Second, this study
does not take the novelty effects into consideration. A longitudinal study might yield more solid
findings because the students will become more familiar with the technology after using it for a
long time, and the novelty effect will disappear. Further research regarding parents’ attitudes
toward the application of AI chatbots would be worthwhile. As is the case with any research that
is carried out in a specific setting, the external validity of my findings will not be established until
they have been duplicated in other educational settings, at other levels, and in fields that are not
related to foreign language learning.

6. Conclusion
This experimental study looked into the effectiveness of employing AI chatbots to help primary
school kids learn English. The research included seventy-four primary school pupils and two
English teachers. There were two groups of participants: experimental and control. The experimen-
tal group adopted an AI chatbot to assist students’ English learning, whereas the control group
used conventional pedagogy. After data analysis by SPSS and NVivo, the quantitative findings, sub-
stantiated by the qualitative insights gleaned from student and teacher reflections, converged on a
shared outcome: students in the AI group showed much better oral English proficiency and will-
ingness to communicate than those in the conventional group. Chatbots can function as a
potent facilitative tool, not only enhancing students’ English proficiency but also reducing their
language anxiety and boosting confidence by offering a supportive and non-critical practice
environment.
Given these results, the study speculates on why an AI chatbot may enhance primary school chil-
dren’s spoken English abilities, while also making suggestions for better AI chatbot modification and
broad use of AI chatbot in classrooms. However, limitations include the specific student demo-
graphic and singular school setting, possibly affecting the broader applicability of findings. Future
research could explore chatbot efficacy across different educational stages and age groups, as
well as delve into varied language learning aspects.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Notes on contributor
Yijia Yuan is an MPhil student at the Faculty of Education at the University of Cambridge. She received a bachelor’s
degree in English from Zhengzhou University. Her current research interests are educational technology and second
language education. She has published four articles in this field. She is about to continue her studies in the doctoral
field.
 INTERACTIVE LEARNING ENVIRONMENTS 6787

ORCID
Yijia Yuan http://orcid.org/0009-0000-8897-2119

References
Abd Al Galil, H., & Abd Al Galil, M. (2019). The effect of using reflective listening on developing EFL adults’ oral fluency.
Online Submission. https://eric.ed.gov/?id=ED592977
Abdul-Kader, S. A., & Woods, J. C. (2015). Survey on chatbot design techniques in speech conversation systems.
International Journal of Advanced Computer Science and Applications, 6(7). https://doi.org/10.14569/IJACSA.2015.
060712
Atkinson, D. (2002). Toward a sociocognitive approach to second language acquisition. The Modern Language Journal,
86(4), 525–545. https://doi.org/10.1111/1540-4781.00159
Ayedoun, E., Hayashi, Y., & Seta, K. (2015). A conversational agent to encourage willingness to communicate in the
context of English as a foreign language. Procedia Computer Science, 60, 1433–1442. https://doi.org/10.1016/j.
procs.2015.08.219
Bailey, L. W. (2019). New technology for the classroom: Mobile devices, artificial intelligence, tutoring systems, and
robotics. In Educational technology and the new world of persistent learning (pp. 1–11). IGI Global.
Baker, S. C., & MacIntyre, P. D. (2000). The role of gender and immersion in communication and second language orien-
tations. Language Learning, 50(2), 311–341. https://doi.org/10.1111/0023-8333.00119
Blake, R. J. (2013). Brave new digital classroom: Technology and foreign language learning. Georgetown University Press.
Boden, M. A. (2016). AI: Its nature and future. Oxford University Press.
Cha, J.-S., & Kim, T.-Y. (2013). Effects of English-learning motivation and language anxiety of the elementary school stu-
dents on willingness to communicate in English and English speaking. Primary English Education, 19(1), 271–294.
https://www.kci.go.kr/kciportal/ci/sereArticleSearch/ciSereArtiView.kci?sereArticleSearchBean.artiId=ART001754205
Dizon, G. (2020). Evaluating intelligent personal assistants for L2 listening and speaking development. Language
Learning & Technology, 24(1), 16–26. https://doi.org/10.125/44705
Etikan, I. (2016). Comparison of convenience sampling and purposive sampling. American Journal of Theoretical and
Applied Statistics, 5(1), 1. https://doi.org/10.11648/j.ajtas.20160501.11
Firth, A., & Wagner, J. (2007). Second/Foreign language learning as a social accomplishment: Elaborations on a recon-
ceptualized SLA. The Modern Language Journal, 91(s1), 800–819. https://doi.org/10.1111/j.1540-4781.2007.00670.x
Fryer, L., Ainley, M., Thompson, A., Gibson, A., & Sherlock, Z. (2017). Stimulating and sustaining interest in a language
course: An experimental comparison of Chatbot and Human task partners. Computers in Human Behavior, 75,
461–468. https://doi.org/10.1016/j.chb.2017.05.045
Fryer, L., Coniam, D., Carpenter, R., & Lăpușneanu, D. (2020). Bots for language learning now: Current and future direc-
tions. Language Learning & Technology, 24(2), 8–22. http://hdl.handle.net/10125/44719
Gass, S., Lee, J., & Roots, R. (2007). Firth and Wagner: New ideas or a new articulation?. Modern Language Journal, 91,
788–799. https://doi.org/10.1111/j.1540-4781.2007.00669.x
Ghonsooly, B., Khajavy, G. H., & Asadpour, S. F. (2012). Willingness to communicate in English among Iranian non–
English major university students. Journal of Language and Social Psychology, 31(2), 197–211. https://doi.org/10.
1177/0261927X12438538
Green, A., & O’Sullivan, B. (2019). Language learning gains among users of English Liulishuo. LAIX. http://hdl.handle.net/
10547/623198
Hashimoto, Y. (2002). Motivation and willingness to communicate as predictors of reported L2 use: The Japanese ESL
context. University of Hawai’I Second Langauge Studies Paper 20 (2). https://www.hawaii.edu/sls/wp-content/
uploads/2014/09/Hashimoto.pdf
Höhn, S. (2017). A data-driven model of explanations for a chatbot that helps to practice conversation in a foreign
language. In K. Jokinen, M. Stede, D. DeVault, & A. Louis (Eds.), Proceedings of the 18th Annual SIGdial Meeting on
Discourse and Dialogue (pp. 395–405). Association for Computational Linguistics. https://doi.org/10.18653/v1/
W17-5547
Hua, L. L., Chen, L., & Sun, M. M. (2017). 人工智能促进英语学习变革研究 [Research on AI Facilitating English Learning
Reform]. Modern Distance Education, 6, 27–31. https://doi.org/10.13927/j.cnki.yuan.2017.0054
Huang, W., Hew, K. F., & Fryer, L. K. (2022). Chatbots for language learning—Are they really useful? A systematic review
of chatbot-supported language learning. Journal of Computer Assisted Learning, 38(1), 237–257. https://doi.org/10.
1111/jcal.12610
Jeon, J. (2022). Exploring AI chatbot affordances in the EFL classroom: Young learners’ experiences and perspectives.
Computer Assisted Language Learning, 1–26. https://doi.org/10.1080/09588221.2021.2021241
Jia, J., & Chen, W. (2008). Motivate the learners to practice English through playing with chatbot CSIEC. International
Conference on Technologies for E-Learning and Digital Entertainment, 180–191. https://doi.org/10.1007/978-3-540-
69736-7_20
6788 Y. YUAN

Jia, J., Chen, Y., Ding, Z., & Ruan, M. (2012). Effects of a vocabulary acquisition and assessment system on students’ per-
formance in a blended learning class for English subject. Computers & Education, 58(1), 63–76. https://doi.org/10.
1016/j.compedu.2011.08.002
Johnson, J. S., & Newport, E. L. (1989). Critical period effects in second language learning: The influence of maturational
state on the acquisition of English as a second language. Cognitive Psychology, 21(1), 60–99. https://doi.org/10.1016/
0010-0285(89)90003-0
Kim, N.-Y. (2016). Effects of voice chat on EFL learners’ speaking ability according to proficiency levels. Multimedia-
Assisted Language Learning, 19(4), 63–88. https://doi.org/10.15702/mall.2016.19.4.63
Kim, N.-Y. (2018). Chatbots and Korean EFL students’ English vocabulary learning. Journal of Digital Convergence, 16(2),
1–7. https://doi.org/10.14400/JDC.2018.16.2.001
Kim, N.-Y., Kim, H.-S., & Cha, Y. J. (2019). Future English learning: Chatbots and artificial intelligence. Multimedia-Assisted
Language Learning, 22(3), 32–53. https://doi.org/10.15702/mall.2019.22.3.32
Knight, W. (2017, June, 31). To build a smarter chatbot, first teach it a second language. PublisherLoc: MIT Technology
Review. https://www.technologyreview.com/2017/07/31/150235/to-build-a-smarter-chatbot-first-teach-it-a-
second-language/
Ko, C.-J. (2012). Can synchronous computer-mediated communication (CMC) help beginning-level foreign language
learners speak? Computer Assisted Language Learning, 25(3), 217–236. https://doi.org/10.1080/09588221.2011.
649483
Krashen, S. (1982). Principles and practice in second language acquisition. Pergamon Press Inc.
Lee, J. S., & Chen Hsieh, J. (2019). Affective variables and willingness to communicate of EFL learners in in-class, out-of-
class, and digital contexts. System, 82, 63–73. https://doi.org/10.1016/j.system.2019.03.002
Li, Y., & Min, S. C. (2010). 国内英语口语研究现状及发展趋势 [Current Status and Development Trends of Domestic
English Oral Research]. Chinese Foreign Languages, 7(06), 85–91. https://doi.org/10.13564/j.cnki.issn.1672-9382.
2010.06.014
Lin, M. P.-C., & Chang, D. (2020). Enhancing post-secondary writers’ writing skills with a chatbot. Journal of Educational
Technology & Society, 23(1), 78–92.
Linh, P. M., Starčič, A. I., & Wu, T.-T. (2022). Challenges and opportunities of education in the COVID-19 pandemic:
Teacher perception on applying AI chatbot for online language learning. International Conference on Innovative
Technologies and Learning, 501–513. https://doi.org/10.1007/978-3-031-15273-3_55
Long, M. H. (1997). Construct validity in SLA research: A response to Firth and Wagner. The Modern Language Journal, 81
(3), 318–323. https://doi.org/10.2307/329306
MacIntyre, P. D., Clément, R., Dörnyei, Z., & Noels, K. A. (1998). Conceptualizing willingness to communicate in a L2: A
situational model of L2 confidence and affiliation. The Modern Language Journal, 82(4), 545–562. https://doi.org/10.
1111/j.1540-4781.1998.tb05543.x
Mayer, R. E. (2017). Using multimedia for e-learning. Journal of Computer Assisted Learning, 33(5), 403–423. https://doi.
org/10.1111/jcal.12197
Peng, J.-E. (2019). The roles of multimodal pedagogic effects and classroom environment in willingness to communicate
in English. System, 82, 161–173. https://doi.org/10.1016/j.system.2019.04.006
Peng, J.-E., & Woodrow, L. (2010). Willingness to communicate in English: A model in the Chinese EFL classroom context.
Language Learning, 60(4), 834–876. https://doi.org/10.1111/j.1467-9922.2010.00576.x
Read, J. (2000). Assessing vocabulary. Cambridge university press.
Reinders, H., & Sorada, W. (2014). Can i say something? The effects of digital game play on willingness to communicate.
Language Learning & Technology, 18(2), 101–123. http://llt.msu.edu/issues/june2014/reinderswattana.pdf
Sha, G. (2009). AI-based chatterbots and spoken English teaching: A critical analysis. Computer Assisted Language
Learning, 22(3), 269–281. https://doi.org/10.1080/09588220902920284
Shevat, A. (2017). Designing bots: Creating conversational experiences. O’Reilly Media, Inc.
Tai, T.-Y., & Chen, H. H.-J. (2023). The impact of Google Assistant on adolescent EFL learners’ willingness to communi-
cate. Interactive Learning Environments, 31(3), 1485–1502. https://doi.org/10.1080/10494820.2020.1841801
Trust, T., & Whalen, J. (2021). K-12 teachers’ experiences and challenges with using technology for emergency remote
teaching during the COVID-19 pandemic. Italian Journal of Educational Technology, IJET-ONLINE FIRST. https://doi.
org/10.17471/2499-4324/1192
Underwood, J. (2017). Exploring AI language assistants with primary EFL students. CALL in a Climate of Change: Adapting
to Turbulent Global Conditions-Short Papers from EUROCALL, 317–321. https://doi.org/10.14705/rpnet.2017.
eurocall2017.733
UNESCO. (2021). AI and education: Guidance for policy-makers (pp. 1–45). United Nations Educational, Scientific and
Cultural Organization. http://creativecommons.org/licenses/by-sa/3.0/igo/
Vasconcelos, M., Candello, H., Pinhanez, C., & dos Santos, T. (2017). Bottester: Testing conversational systems with simu-
lated users. Proceedings of the XVI Brazilian Symposium on Human Factors in Computing Systems, 1–4. https://doi.org/
10.1145/3160504.3160584
 INTERACTIVE LEARNING ENVIRONMENTS 6789

Wang, Y. F., Petrina, S., & Feng, F. (2017). VILLAGE—Virtual immersive language learning and gaming environment:
Immersion and presence. British Journal of Educational Technology, 48(2), 431–450. https://doi.org/10.1111/bjet.
12388
Wang, Y. H., Young, S. S.-C., & Jang, J.-S. R. (2013). Using tangible companions for enhancing learning English conversa-
tion. Journal of Educational Technology & Society, 16(2), 296–309. https://www.jstor.org/stable/10.2307jeductechsoci.
16.2.296
Xu, Y., Wang, D., Collins, P., Lee, H., & Warschauer, M. (2021). Same benefits, different communication patterns:
Comparing Children’s Reading with a conversational agent vs. A human partner. Computers & Education, 161,
104059. https://doi.org/10.1016/j.compedu.2020.104059
Yang, H., Kim, H., Lee, J. H., & Shin, D. (2022). Implementation of an AI chatbot as an English conversation partner in EFL
speaking classes. ReCALL, 34(3), 327–343. https://doi.org/10.1017/S0958344022000039
Yashima, T. (2002). Willingness to communicate in a second language: The Japanese EFL context. The Modern Language
Journal, 86(1), 54–66. https://doi.org/10.1111/1540-4781.00136
Yashima, T. (2009). 7. International posture and the ideal L2 self in the Japanese EFL context. Motivation, Language
Identity and the L2 Self, 86(1), 144–163. https://doi.org/10.21832/9781847691293-008
Yu, M. (2011). Effect of communication variables, affective variables, and teacher immediacy on willingness to commu-
nicate of foreign language learners. Chinese Journal of Communication, 4(02), 218–236. https://doi.org/10.1080/
17544750.2011.565678
Zhang, F. H., Li, W. T., Long, M. Y., & Gao, Y. (2019). 基于三个技术平台的自我调节性写作学习效果对比研究
[Comparative study of self-regulated writing learning effects based on three technology platforms]. Foreign
Language Electronic Teaching, 5(5), 22–26.
Zhang, J., Beckmann, N., & Beckmann, J. F. (2018). To talk or not to talk: A review of situational antecedents of willingness
to communicate in the second language classroom. System, 72, 226–239. https://doi.org/10.1016/j.system.2018.01.
003
Zou, B., Liviero, S., Hao, M., & Wei, C. (2020). Artificial intelligence technology for EAP speaking skills: Student perceptions
of opportunities and challenges. In Technology and the psychology of second language learners and users (pp. 433–
463). Springer. https://link.springer.com/chapter/10.1007978-3-030-34212-8_17
Zou, B., & Wang, M. J. (2021). 人工智能技术与英语教学⍰现状与展望 [Artificial Intelligence Technology and English
Teaching: Current Status and Prospects]. Foreign Languages and Literature, 37(3), 124–130. https://doi.org/10.
3969/j.issn.1674-6414.2021.03.013