You are on page 1of 27

Computer Assisted Language Learning

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/ncal20

Exploring AI chatbot affordances in the EFL


classroom: young learners’ experiences and
perspectives

Jaeho Jeon

To cite this article: Jaeho Jeon (2022): Exploring AI chatbot affordances in the EFL classroom:
young learners’ experiences and perspectives, Computer Assisted Language Learning, DOI:
10.1080/09588221.2021.2021241

To link to this article: https://doi.org/10.1080/09588221.2021.2021241

View supplementary material

Published online: 10 Jan 2022.

Submit your article to this journal

View related articles

View Crossmark data

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


https://www.tandfonline.com/action/journalInformation?journalCode=ncal20
Computer Assisted Language Learning
https://doi.org/10.1080/09588221.2021.2021241

Exploring AI chatbot affordances in the EFL


classroom: young learners’ experiences and
perspectives
Jaeho Jeon
Department of English Education, Seoul National University of Education, Seoul, Republic of Korea

ABSTRACT KEYWORDS
Professionals within the field of language learning have pre- chatbots; conversational
dicted that chatbots would provide new opportunities for agents; artificial
the teaching and learning of language. Despite the assumed intelligence; affordances;
benefits of utilizing chatbots in language classrooms, such learner motivation; EFL
learners; Dialogflow
as providing interactional chances or helping to create an
anxiety-free atmosphere, little is known about learners’ actual
use of chatbots during language classes or how chatbots
affect their motivation to learn a language. To address these
gaps, this exploratory study aimed to create an inventory
of affordances that chatbots provide in the primary English
as a foreign language (EFL) classroom and to explore how
the affordances affect psychological aspects in language
learners, particularly regarding their motivation to learn
English through chatbots. Thirty-six Korean primary school
learners participated in a 16-week EFL course that utilized
customized chatbots. These chatbots were created using
Google’s Dialogflow. After the course, individual in-depth
interviews were conducted regarding the participants’ expe-
riences and perceptions of the chatbots. Student-chatbot
interaction logs produced during the course were also col-
lected to supplement the interview data. Qualitative analysis
of the interview transcripts and interaction logs revealed
the presence of pedagogical, technological, and social affor-
dances. Depending on the learner, the chatbot affordances
were perceived differently; thus, each affordance acted as
either an opportunity or a constraint for English language
learning. In addition, this study specifically discussed how
these chatbot affordances might have affected psychological
states in language learners. Future recommendations regard-
ing the use of chatbots in language classrooms were sug-
gested from both pedagogical and technological
perspectives.

CONTACT Jaeho Jeon jaehojeon21@gmail.com Department of English Education, Seoul National


University of Education, Seochojungang-ro, Seocho-gu, Seoul 06639, Republic of Korea
Supplemental data for this article can be accessed here.
© 2022 Informa UK Limited, trading as Taylor & Francis Group
2 J. JEON

Introduction
Language educators and researchers have recommended chatbots as
powerful tools for the facilitation of language learning by emphasizing
the value of their role as conversation partners (Fryer et al., 2020). With
the development of artificial intelligence (AI) technology, such as natural
language processing or machine learning, the relevance of chatbot tech-
nology to language learning has become more prevalent. Reflecting on
this change, various scholarly attempts to integrate chatbot technology
into language learning have been made. Previous research regarding
chatbot technology for language learning has revealed that chatbots are
effective in increasing language learning motivation (Lee et al., 2011),
facilitating specific language skills such as lexical inferencing (Jeon,
2021), and improving overall L2 proficiency (Bibauw et al., 2019;
Divekar* et al., 2021). Furthermore, research has indicated that chatbots
act as tools for providing abundant conversation opportunities or practice
(Timpe-Laughlin et al., 2020; Xu et al., 2021). On the other hand, some
concerns have been raised regarding the efficacy of using chatbots in
language learning. These include concerns that chatbots are not capable
of sustaining goal-oriented conversations necessary for the purpose of
language learning (Coniam, 2014); that learners quickly lose interest
after the early stages of language learning using chatbots (Fryer et al.,
2017, 2019); or that learners do not conduct themselves in a manner
appropriate for conversations with human partners when they commu-
nicate with chatbots (Hill et al., 2015). In addition, Divekar et al. (2018)
indicated that chatbots should possess help features for learners who
lack L2 proficiency, such as providing feedback in L1 or vocabulary
clarification, to overcome interaction challenges.
Despite the growing popularity of chatbot technology for language
learning, it is still unclear how chatbots could be integrated into lan-
guage classrooms. One reason for this lack of knowledge is that little
research has been conducted on how chatbots are used in language
classrooms or on how chatbots in such settings affect learners’ motiva-
tion to learn languages, although other factors, such as technological
immaturity, a lack of infrastructure, or teachers’ reluctance to use tech-
nology, can also contribute. Furthermore, as noted in review studies on
chatbots (Huang et al., 2021; Hwang & Chang, 2021), chatbot research
in education has mainly used quantitative methods in design or has
used relatively short-term experiments composed of a small number of
tasks. Before we can determine whether chatbots are beneficial tools for
use in the language classroom, it is of particular importance to gain
specific knowledge about how chatbots can be incorporated into the
language classroom and about their affective influences on language
Computer Assisted Language Learning 3

learners, especially through longitudinal and interpretive work that can


deepen our contextualized understanding of chatbots.
To address these concerns, the researcher first built educational chat-
bots capable of sustaining goal-oriented conversations by using Google’s
Dialogflow, where teachers can make customized chatbots, and then
implemented a set of language activities using the chatbots during a
16-week English as a foreign language (EFL) course for beginner-level
primary school students. By qualitatively analyzing student-chatbot inter-
action records produced during the course and post-course individual
interview data, this exploratory study aimed to identify and inventory
affordances that chatbots provide within primary EFL classrooms. In
addition, this research discussed both the opportunities and constraints
of using chatbots in regard to English learning while also revealing how
chatbots affect psychological aspects in language learners, particularly
regarding motivation to learn English through chatbots. The following
question guided this research:
What are enabling and constraining affordances provided by chatbots in the EFL
classroom, and how do those affordances affect students’ motivation to learn English?

Literature review
Chatbot technology for language learning

Chatbot technology has garnered a great deal of attention from language


learning professionals, mainly due to its chat functionality. Suggesting
a vision for a conversation practice machine, Atwell (1999) paid attention
to the role of chatbots for language learning; this was because of their
potential to provide students with abundant opportunities for practicing
a new language. Over time, chatbots have progressed alongside the
development of technology such as natural language processing and
machine learning, thus increasing the relevance of chatbots within the
field of language learning (Fryer et al., 2020).
Initial attempts to use chatbots for language learning were made using
rule-based chatbots that were implemented with either restricted pattern
or keyword matching. For example, Jia (2003) analyzed 1,256 users’
chatbot interaction logs. The chatbot system employed restricted key-
words and pattern-matching mechanisms. The paper concluded that the
chatbot system that operated on these limited mechanisms could not
work effectively as a teaching assistant program in language learning
because it was not equipped with “the ability to process meaning (nat-
ural language understanding)” (Bibauw et al., 2019, p. 853). Recent
advances in natural language processing and machine learning technology
have made human-chatbot interaction more authentic and suitable for
4 J. JEON

language learning and thus have created new possibilities for language
teachers who desire to use chatbots (Kim & Jang, 2020). Reflecting on
this development, language scholars in chatbot research have primarily
employed two types of chatbots: 1) general-purpose chatbots that can
conduct a daily conversation with users, such as Cleverbot (e.g., Fryer
et al., 2017, 2019) and 2) purposeful chatbots that can conduct specific
tasks with learners of a particular language proficiency (e.g., Jeon, 2021;
Sydorenko et al., 2019; Timpe-Laughlin et al., 2020).
Despite these efforts, there may be limitations to using these two
types of chatbots in long-term language courses, particularly beginner-level
EFL courses. Regarding general-purpose chatbots, first, dialogues with
these chatbots tend to be strictly user-initiated and mainly reactive
(Bibauw et al., 2019). Novice EFL students might find it difficult to
initiate or actively lead conversations with the chatbots in the target
language, given their limited language proficiency (Divekar et al., 2018,
2021). Next, beginner-level EFL classes inevitably encompass specific
learning contexts, learning contents, and particularly structured inter-
actions (DeKeyser, 2010). This means that these classes need to utilize
chatbots capable of sustaining goal-oriented conversations within an
appropriate educational boundary; it is questionable how well English
teachers within their own specific teaching contexts could employ
general-purpose chatbots for their educational purposes. In addition, as
for existing purposeful chatbots, since these chatbots were created mainly
reflecting the context of a specific class, the chatbots are unlikely to
cover much of the content of any other long-term language course,
although the chatbots provide useful insight into how chatbots could
be created and used to attain specific classroom objectives (AlKhayat,
2017; Divekar* et al., 2021).
To address these limitations, language teachers can create customized
chatbots that are able to sustain goal-oriented conversations with students
by using chatbot builders such as Botstify, Chatfuel, or Dialogflow1.
Generally, the chatbots created through these platforms are employed
to perform functions such as help-desk, triage, lead generation, or
E-commerce assistance for business purposes (Sabharwal & Agrawal,
2020). Scholars in the field of language learning have also noted the
potential usefulness of these platforms (e.g., Jeon, 2021; Smutny &
Schreiberova, 2020). For example, Lee et al. (2020) noted that customized
chatbots created through these platforms could cater to teachers’ ped-
agogical needs, although some degree of training would be required.
Kessler (2018) also highlighted the relative easiness of chatbot creation
through the use of these platforms compared to coding software for
professionals. Furthermore, most of the platforms provide an automatic
transcript function that can be significantly useful in language classrooms
Computer Assisted Language Learning 5

consisting of multiple students because teachers can monitor each inter-


action through the records and accordingly prepare different instructions
for each student (Jeon, 2021). The researcher considered the benefits of
the various platforms, and for this study, selected Dialogflow because
of its compatibility with devices used in this study. In sum, by using
this platform, the researcher was able to create customized chatbots
capable of orally interacting with young students and of performing
goal-oriented conversations during a one-semester-long EFL course.

Affordances and learner psychology

The term affordance was first introduced by Gibson (1986) and was
later elaborated on by Stoffregen (2003) to describe action possibil-
ities that can emerge from the relation between an agent and envi-
ronment. Recognizing the value of the term, a body of scholars from
various fields have adopted the concept of affordance by modifying
it to their own contexts. In the field of language learning, scholars
have used the term to describe the language learning possibilities
inherent in a learning environment. For example, van Lier (2004)
preferred this term to input because it better helped teachers take
both the positive and negative potentials of the language learning
environment into consideration. Anderson (2015) also emphasized
the importance of considering the concept of affordance in language
learning and stated that teachers must be prepared for the emergence
of affordance to maximize learning opportunities. In this vein, the
role of language teachers could be considered one that understands
how affordances affect learners and allows for the creation of an
environment wherein learners can actualize positive affordances
(Plonsky & Ziegler, 2016).
Rapid development in technology has provided new affordances for
language learners. Scholars in the field of computer-assisted language
learning (CALL) have attempted to examine psychological effects that the
affordances of technology have on learners; the examination of psycho-
logical states when learners are using technology provides detailed insight
into the necessary conditions for effective learning to occur (Dörnyei,
2014). This line of research has also assisted teachers in making the
language classroom both technologically and pedagogically more relevant
to learners (Freiermuth, 2020). Therefore, for new technology (e.g., AI
tools) to be recommended to learners as a means of supporting their
learning, it is necessary to explore the affordances of the technology and
how the affordances affect the psychological state of the learners.
Previous literature has mainly focused on language learner psychology,
especially regarding motivation to learn languages in environments where
6 J. JEON

L2 interactions occur via the use of technology such as communication


apps or collaborative websites (e.g., Freiermuth & Huang, 2012; Zou
et al., 2018). These studies are meaningful because they have identified
how incorporating technology affects learner motivation by exploring
such factors as anxiety, willingness to communicate, or the need to use
a target language. For example, Freiermuth and Huang (2012) revealed
that online synchronous chat could be an effective mode for language
learning by showing that the technology provided psychological benefits
such as alleviating speaking anxiety and making learners more willing
to communicate in a target language.
However, chatbot technology creates a unique learning environment
in that learners directly interact with AI agents rather than with other
human interlocutors, as is the case when using the communication
technology mentioned above (Bower, 2019). Freiermuth (2020) stated,
“This independence from teachers should not divert our attention away
from the students; rather, it should heighten our attention to the ways
such [AI] applications are affecting the psychological underpinnings of
our students” (p. 18). Recently, Zou et al. (2020) explored university
learners’ perceptions toward AI apps for language learning, specifically
English for academic purposes (EAP) regarding speaking. Although it
was indicated that the apps lowered speaking anxiety and that learners
were satisfied with the apps’ immediate feedback on their overall lan-
guage performance, they also stated that the technology could not replace
language teachers because there were still limitations such as an inability
to recognize users’ local pronunciation or to provide tailored feedback.
AI chatbots are different from other AI tools, such as the EAP apps
mentioned above, in that chatbots are mainly utilized to provide inter-
actional chances to learners. Within an interactive environment, we can
assume that chatbots might generate different psychological effects on
language learners. However, only a few studies have explored how chat-
bots affect psychological states in learners, particularly in relation to
their motivation to learn a language (Bibauw et al., 2019). This line of
research has yielded some quantitative data, but the results have been
inconsistent. For example, Fryer et al. (2017, 2019) revealed that learners
quickly lost their interest in using chatbots to learn a language after
the early stages during a course. In contrast, some reported positive
chatbot effects on psychological factors such as motivation, interest, and
confidence (e.g., Lee et al., 2011). This inconsistency may be partly due
to the complex and multifaceted nature of learner psychology when
learners interact with chatbots. Clearly, research is needed to further
explore the complex nature of this inquiry, especially in the form of an
approach that can provide a contextualized understanding of the com-
plicated aspects of chatbots’ affective effects.
Computer Assisted Language Learning 7

Although chatbots might offer various affordances as a form of tech-


nology themselves, this paper specifically aims to identify and inventory
chatbot affordances in terms of language learning in primary EFL class-
rooms. Furthermore, to address the gap in the literature, and considering
that our understanding of AI chatbots is in its initial stages of devel-
opment, this research seeks to qualitatively explain how the affordances
affect psychological aspects in language learners, particularly regarding
their motivation to learn English.

Method
Participants and classroom context

The settings for this research were two primary classrooms in South
Korea. Thirty-six Korean students, all of them being 12 years old at the
time of the study, participated in a 16-week EFL course that utilized
chatbots. Students in South Korea start to learn English in public schools
at the age of nine and take lessons two to three times a week every
semester; this means that the participants in the research had four years
of EFL learning experience. The results of diagnostic tests administered
at the beginning of the semester were used to decide the participants’
English proficiency; all of them belonged to the novice level according
to American council on the teaching of foreign languages (ACTFL)
guidelines for speaking and listening. In addition, student observation
records offered by an English teacher instructing the participants at the
time of the research provided further information about their overall
spoken English proficiency; the participants were able to perform basic
functions in English, such as talking about hobbies or greeting friends;
but misunderstandings often arose; they frequently paused as they
attempted to recall vocabulary during those types of conversations.
During a 16-week EFL course, lessons were organized on a weekly
basis, with three 40-minute-long classes being covered every week. The
first two weeks were spent familiarizing students with the chatbots. For
the remaining 14 weeks, a different topic was taught each week, following
the same weekly lesson structure consisting of three sequential config-
urations: whole group, small group, and individual. Specifically, the
students first participated in a whole class configuration where they
were presented with target words and sentences and participated in
target dialogue practice through a combination of games, songs, and
chants depending on the lesson. In the second class of the week, they
participated in information-gap tasks that were adapted by the researcher
from the tasks suggested in Ur (1990). The students performed the tasks
in a group configuration by communicating with their peers. Next, as
8 J. JEON

the final class of the weekly lessons, another opportunity was given to
perform the information-gap tasks with the chatbots replacing their
peers; the students spent most of the allotted class time having inde-
pendent conversations with only their chatbots to complete the tasks.
Worksheets were supplied to the students to provide information nec-
essary to complete the tasks. The task topics were related to real-life
situations, such as how they could reach or travel to specific buildings
and locations (see Supplementary Appendices 1 and 2 for more infor-
mation regarding weekly topics and a worksheet sample). When the
students asked for help regarding technical issues or task performance,
the researcher, who also performed the role of a teacher, helped the
students perform tasks with the chatbots.

Chatbots for this study

The researcher created chatbots using Dialogflow to ensure that the chat-
bots could perform both extended and goal-oriented conversations with
the students regarding topics in the context of beginner-level EFL classes.
Specifically, the chatbots were designed to provide four different sets of
responses: corrective feedback, prompts, fallback intent, and evaluation,
all of which were given based on the type of student input. Conditions
and examples for each chatbot response are presented in Figure 1. The
dialogues coded into the chatbots are presented in Supplementary Appendix
1. To finish one set of interactions, the minimum number of interactional
turns (i.e., a pair of student-chatbot utterances) was two; however, depend-
ing on the type of student input, the interaction could continue until the
students provided the target response to the chatbot question.
The chatbots were then integrated with the Google Assistant interface
that provided speech-to-text and text-to-speech technologies. Students
accessed the chatbots by using the interface through tablet PCs, which
means that it was possible for the students to not only speak and listen
but also to read and type during their interactions with the chatbots.
This audio-based interaction makes the chatbots more relevant to lan-
guage learning (van Doremalen et al., 2016).

Data collection and analysis

The researcher adopted a qualitative approach in order to identify poten-


tial chatbot affordances and to reveal how the affordances might have
affected language learner psychology. To this aim, data were collected
from two sources: individual in-depth interviews as a primary source
and student-chatbot interaction logs to supplement student comments,
especially on their interaction experiences with the chatbots.
Computer Assisted Language Learning 9

Figure 1. A chatbot decision tree diagram example.

Regarding the interview process, students were first asked to write


reflections about their English learning experiences with the chatbots
immediately following the end of the course. As a prompt, the researcher
clarified that they should describe memorable or challenging moments
experienced when using the chatbots during the classes. Next, questions
for semi-structured interviews were designed, and the interviews were
conducted with each student about a week after the last class. The
researcher structured interviews around the following questions.

• Can you describe how you used the chatbots?


• Did you find using chatbots helpful when studying English? Why?
• What difficulties did you have when studying with the chatbots?
• How did you overcome these difficulties?
• Do you want to continue to use chatbots in English class? Why?

Given the exploratory aim of the study, the researcher initially


attempted to investigate what was possible through the chatbots in terms
of English language learning. Next, based on each student’s reflection
log, follow-up questions were carried out focusing on student perceptions
of the chatbots, specifically focusing more on the aspects of the chatbots
that could facilitate or constrain English language learning. The length
of these interviews ranged from 32 to 53 minutes. All interviews were
conducted in students’ L1 while being audio-taped. The interviews were
then transcribed verbatim.
10 J. JEON

Next, the transcripts of the interview data were qualitatively analyzed


to discover repeated patterns of meaning in relation to the research
question, following a thematic analytical approach (Braun & Clarke,
2006). Accordingly, an initial coding scheme was developed by the
researcher as follows: first, after the researcher gained a general under-
standing of the transcripts, all parts related to language learning were
extracted for coding; next, each code that shared a similar underlying
meaning was then grouped into a common theme; finally, themes were
categorized into different affordances, and three affordance categories
were identified. Next, a third of the transcripts that were randomly
derived from the interview data were given to another language learn-
ing professional. This professional was also asked to follow the same
coding scheme and to classify each part into the identified affordance
categories. Afterward, a discussion between the two coders was carried
out in which a new modified coding scheme was developed by refining
the initial one (see Supplementary Appendix 3). Last, guided by the
refined coding scheme, the two coders independently coded all tran-
scripts. Ninety-four percent intercoder reliability was achieved, and
another discussion was carried out to resolve disagreements. Themes
that consisted of less than five codes were excluded from the analysis.
In the end, 385 parts were divided into three different affordance
categories. Representative answers were extracted as a method of doc-
umenting and capturing repeated patterns of the meaning behind the
students’ words. Student names were coded with numbers (e.g., S1,
S2, …) to ensure confidentiality.
Student-chatbot interaction logs were then examined in an effort to
supplement the interview data, particularly regarding the students’ inter-
action experiences with the chatbots. The researcher read all of the
interaction transcripts that were saved in the Dialogflow system in a
line-by-line fashion to find any supplementary clues to the interview
excerpts and to ensure the objectivity and representativeness of the
student comments. This analysis process occurred in an iterative manner
as the researcher moved between the interaction logs, student interview
data, and relevant research reviewed in the Literature review. As a result,
some interaction logs were selected when considered representative of
the interview themes and presented along with interview excerpts to
describe chatbot affordances in an objective manner.
This type of qualitative approach is considered the most appropriate
for identifying any revelations in a research topic that a study might
provide, as is the case with the exploratory nature of the current study
(Braun & Clarke, 2006; Johnson, 1992). Particularly, by basing the cur-
rent study on students’ actual interaction logs and their reflections after
extended exposure to chatbots, the researcher gained a greater
Computer Assisted Language Learning 11

understanding of what affordances chatbots provided in the EFL class-


room and of psychological aspects in learners, particularly regarding
their motivation to learn English through chatbots.

Results
As a result of data analysis, three categories of affordances were iden-
tified: pedagogical, technological, and social. Pedagogical affordances
emerged from the direct interactions between the chatbots and the
students. These affordances are associated with the primary purpose of
employing chatbots in the language classroom, namely, to provide inter-
actional chances to students. Second, technological affordances are related
to the use of technology derived from the medium used for chatbot
operation or associated with the embedded technological features of the
chatbots. Last, social affordances are related to how the chatbots influ-
enced attitudes toward language class. These affordances were generated
because the use of chatbots created a socially unique learning configu-
ration that differed from traditional classrooms. An overview of these
affordances is presented in Table 1.

Pedagogical affordances

One major strength of the chatbots perceived by the students was related
to the chatbots’ interactional features. Figure 2 is an interaction tran-
script between S4 and the chatbot that was automatically saved in the
Dialogflow system during week 7 (see Supplementary Appendix 2 for
the worksheet used for this task). This interaction log is representative

Table 1. Overview of chatbot affordances.


Affordances Themes
Pedagogical Opportunity
Receiving speaking and listening chances
Receiving immediate feedback
Knowing when and how to use target speech
Constraint
Lacking skills to perform interactions
Technological Opportunity
Using the stop button to think more
Using search engines to find information
Using online dictionaries as vocabulary sources
Focusing on pronunciation when input is unrecognized
Constraint
Feeling anxiety about misrecognition
Repeating a conversation from the beginning when input is unrecognized
Social Opportunity
Being free from anxiety in front of others regarding speech mistakes
Controlling one’s own learning pace independently without peer-pressure
Constraint
Desiring to study with peers
12 J. JEON

Figure 2. Interaction transcript between S4 and the chatbot.

of a successful conversation that students could have with the chatbots.


These types of successful interaction experiences with the chatbots
allowed the students to benefit from the chatbots’ pedagogical affordances
in the EFL classroom.
According to 18 students, the chatbots were considered interlocutors
with whom the students could interact during the performance of the
tasks. This perception of the chatbots facilitated the students’ motivation
to utilize the chatbots, which led to active engagement in speaking
English. For example, 12 students mentioned that they felt the chatbots
were real human-like conversation partners and could practice listening
and speaking English with them. S4 stated, "It was great because I could
talk as much as I wanted, and the chatbot understood what I said. It
is like I was talking to a real person." In addition, the students reported
that they enjoyed conversations with the chatbots and were more engaged
in class compared to when they studied either with a whole class or a
group configuration. Similarly, S2 said, “I did not have as many speaking
and listening opportunities like this when I took class without the
chatbot.” Furthermore, as six students reported, tasks with the chatbots
created a more authentic interactional atmosphere, and they learned
Computer Assisted Language Learning 13

when and how they should use new target words and sentences appro-
priately. S14, for example, found that through interacting with the chat-
bot, he could not only listen to but also speak target sentences and
could understand whether his speech was appropriate or not in a given
context thanks to the immediate feedback the chatbot provided. Similar
comments were made:
I think it is very practical because I felt like I was talking in a real situation,
so I could learn how to use new expressions. (S3)

Whenever I made a mistake, the chatbot gave me feedback right away. It was
helpful because I could learn how to use expressions right. (S9)
I think I could use new expressions when I talk to a real foreigner because I
already had similar conversations with the chatbot, and now I know it. (S12)

However, while functioning as a conversational partner for some


students, there were also constraints caused by the interactional affor-
dance. While many of the students who were able to have a conversation
with the chatbots reported on the value of interactional opportunities,
eight students whose English skills were not advanced enough to have
a spoken conversation with the chatbots mentioned the difficulties they
encountered. It was observed that those students did not finish their
tasks. Figure 3 is an interaction transcript between S18 and the chatbot,
which represents a sample of an unsuccessful interaction.
S18, whose English proficiency was not sufficient to fully interact with
the chatbot, tried to speak in his L1. However, as shown in Figure 3, the
chatbot system could not recognize his L1 utterances and thus categorized
those utterances as unrelated, even though the L1 utterances might have
been related to the topic. Comments from the students provide informa-
tion regarding how these experiences might have affected their motivation
to use the chatbots as follows:
I wish the chatbot could speak Korean. I needed help to interact with the chat-
bot. The chatbot just kept saying I was wrong. I completed the activity with the
teacher’s help, but it was discouraging. (S18)

Since I am not good at English, it was very difficult. I need a teacher, not a
chatbot, who can help me step by step in Korean. (S10)

It is evident in the interview data that the interactive feature of the


chatbots offered plenty of interaction opportunities to students; some
students enjoyed these opportunities. However, there was also the sense
that the chatbots did not provide the benefit of being able to assist the
students as a teacher would during conversation practice (Divekar et al.,
2018). It was observed that this perception toward the chatbots’ inability
hampered some students’ motivation to use the tools for language
14 J. JEON

Figure 3. Interaction transcript between S18 and the chatbot.

learning (Zou et al., 2020); this is supported by comments that the


students made, such as S10 and S18, whose English skills were not
advanced enough to perform interactions with the chatbots in English only.

Technological affordances

The chatbots employed in this study were uploaded to tablet PCs; there-
fore, the chatbots provided several technological affordances that were
not only created by chatbot technology itself but also by the tablet PCs.
First, regarding the strengths of the chatbots, 10 students mentioned
the technological affordances of the chatbots concerning tablet PC func-
tionalities. For the most part, these affordances were described by using
positive words such as convenient, easy, and useful; the affordances
helped sustain the students’ willingness to interact with the chatbots
using English (Soon et al., 2020). For example, S11 mentioned that he
could stop the chatbot when he needed to think more. S9 said, “When
I could not figure out what to say, I stopped the chatbot and prepared
what to say. It was convenient because all I had to do was just push
the stop button on the screen.” S15 said, “I could easily go back and
forth from the chatbot to an online dictionary to find the meanings of
Computer Assisted Language Learning 15

words used in tasks.” S16 reported, “I could just stop and go to the
internet to search for some information about tasks.”
In addition, several comments were reported regarding speech rec-
ognition problems that the students experienced when they attempted
to talk to the chatbots. Thirty-three students reported that they expe-
rienced this type of speech recognition problem; however, what should
be noted is that these recognition issues were perceived differently
depending on the student. First, a total of 13 students described the
recognition problems as anxious, worrisome, or exhausting. This per-
ception made the students unwilling to speak English to the chatbots.
The comments included:
When the chatbot did not recognize my speech, it was very disappointing. I felt
anxious if the chatbot could not recognize my pronunciation. (S7)
The chatbot often misunderstood what I said. I was quite worried about my
speaking. Several times, I just typed instead of speaking. (S6)
It did not recognize what I said well. When I failed to make the chatbot under-
stand what I said several times, the chatbot just left the chatroom, and I had to
start the conversation all over again. It was exhausting. (S5)

However, in the case of 20 students, these recognition issues were


viewed as an opportunity to practice English speaking with more accurate
pronunciation. It was frequently observed that those students attempted
to pronounce words more clearly and to fix their errors in a repetitive
manner when the chatbot did not understand their English utterances.
Figure 4 is an interaction transcript between S19 and the chatbot, cap-
turing one of these students’ responses to the chatbot’s misrecognition.
The interview data revealed that, even with the technical limitations,
the chatbots enhanced some students’ motivation to learn English
through the chatbots. The students attempted to overcome these lim-
itations by repeatedly adjusting their English utterances because they
attributed misrecognition to themselves, not to the chatbots. Furthermore,
these sorts of obstacles were even viewed as game-like activities where
the students should complete missions.
It was hard to make the chatbot recognize the word florist. I found out that I
pronounced the word wrong. I tried to pronounce florist differently and finally
came to know how to pronounce the word right. (S19)

I concentrated on my pronunciation more while speaking to my chatbot. I repeat-


edly checked my pronunciation. I studied pronunciation in a more focused way
than before with the help of the chatbot. (S2)

It was like a game. My mission was to make the chatbot understand my speech.
When it recognized my speech, I felt so satisfied. I wondered if the chatbot also
recognized other sentences. (S21)
16 J. JEON

Figure 4. Interaction transcript between S19 and the chatbot.

As indicated in the interview data, students could employ other digital


tools such as online dictionaries or search engines while using the
chatbots, which helped sustain their willingness to interact with the
chatbots using English. Speech recognition problems made some par-
ticipants feel anxious about their utterances and lowered their motivation
to use English with the chatbots (Freiermuth, 2001; Freiermuth & Huang,
2012). In contrast, many students perceived this technological limitation
in the chatbots as a learning opportunity in which they could improve
their pronunciation accuracy in an engaging way. In other words, many
of the students recognized the chatbots as helpful tools from which they
could learn more; this perception ultimately led to increased motivation
to interact with the chatbots using English.

Social affordances

Students also described ways in which the learning configuration created


by the use of the chatbots influenced their attitude toward English class.
Most of the students indicated that they developed a positive attitude
toward their English class because of the chatbots. The students men-
tioned that a learning configuration where they practiced English with
only the chatbots was very beneficial because they did not worry about
Computer Assisted Language Learning 17

other interlocutors such as friends or teachers. Twenty-six students stated


that they felt comfortable talking to the chatbots because they did not
feel pressure from having to worry about human partners, specifically,
about their speech mistakes; that is, there was no sense of being judged
by their peers or teacher. Among the 26 students, eight students, includ-
ing S2 and S11, elaborated more on their preference for the chatbots
by citing the independent learning opportunities that the chatbots cre-
ated. Comments related to these pressure-free feelings included:
Even if I make countless mistakes, the chatbots do not get annoyed. So, I can
try as much as I want with only the chatbot without being worried about my
partner. (S2)
I could practice what I wanted at my own pace because the chatbot was the
only partner that I had. I am not good at English, so when studying English
with friends in a group, I felt pressured because I thought I might slow down
the group work. (S11)

As the comments reveal, the learning configuration provided by the


use of the chatbots was facilitative to their engagement in speaking
English. That is, having the chatbots as the only interlocutor while
performing the task helped reduce anxiety related to speaking, which
ultimately facilitated the students’ willingness to communicate in English
(Freiermuth & Huang, 2012; Freiermuth & Jarrell, 2006). However, it
was also observed that 10 students did not enjoy this learning config-
uration. The students considered the chatbot tasks in the configuration
not as attractive as tasks with human partners. This negative perception
toward the chatbots lowered their motivation to conduct language tasks
with the chatbots; two main reasons were identified regarding this neg-
ative perception. First, the students preferred human partners to the
chatbots because they enjoyed doing language activities more actively
with human partners than practicing alone. These comments included:
When we use the chatbots, this means that we cannot play an English game. I
like it when I can play an English activity with friends in the class. When I have
a chance to move my body or to compete with friends, I think class becomes
more engaging. (S9)

Second, the students also expressed their desire to collaborate with


human partners rather than with the chatbots. For example, S1, whose
English level was the highest among the participants, indicated that he
also did not like using the chatbots because he could not help or col-
laborate with other friends by using his already established English skills.
The comment was:

It was boring because I talked only to the chatbot that was just a machine. I
want to study English by talking to my friends. … I finished my tasks with the
18 J. JEON

chatbot, but it was not interesting. English class becomes more fun when I can
help my friends. (S1)

The interview data confirmed that a learning configuration that uti-


lized chatbots was perceived differently depending on the student, result-
ing in different attitudes toward English classes using that configuration.
Most of the participants expressed enjoyment about learning with only
the chatbots, but for some students, learning with the chatbots was
demotivating because they thought the chatbots could not replace human
partners (Fryer et al., 2017, 2019).

Discussion
It is important to recall that an individual student’s perceived affordances
did not occur in a decontextualized manner. Rather, each affordance
was perceived and determined by the individual student’s ability to
recognize and react to a particular feature within the technology-enhanced
learning environment (Chun et al., 2016; Freiermuth, 2020). Furthermore,
the affordances also affected the psychological aspects of the students,
which ultimately facilitated or decreased their motivation to learn a
language through the chatbots. The psychological aspects identified in
this study consisted of students’ own perceptions of their L2 competency
or technology control level, awareness of the chatbot’s pedagogical value,
perceptions of chatbots as authentic speakers, and anxiety regarding
English speaking. These aspects provided insight into how chatbots
should be incorporated into the language classroom.
First, student perceptions of their English language competencies
affected their motivation to learn English through chatbots. During the
16-week course that consisted of multiple topics, each topic varied in
the difficulty level of English words and sentences required. When stu-
dents perceived that a chatbot task could be resolved with their prior
language competency, they became more willing to communicate with
the chatbots using English (AlKhayat, 2017). For example, S18, who
described her overall experience with the chatbots as discouraging and
attributed this to her low English competency, stated that she particularly
enjoyed the direction-finding task with the chatbots during the second
week. She indicated that this task was within her competency level, as
could be observed in her reflection where she mentioned, “It was good
to do the direction task with a partner like the chatbot because I could
understand the chatbot well during this task and it was intelligent
enough to communicate with me” (S18). In this vein, language teachers
who desire to employ chatbots in class should first consider students’
prior language competencies when preparing chatbot tasks (Fryer
et al., 2019).
Computer Assisted Language Learning 19

However, not all students reported negative perceptions toward tasks


beyond their prior language competency. The interview data indicated
that technological affordances compensated for interactional constraints.
That is, some students utilized the technological aspects or functionalities
of the chatbots to compensate for the skills, knowledge, and attitudes
they lacked when interacting with the chatbots, which helped the stu-
dents sustain their motivation to interact with the chatbots in English
(Soon et al., 2020). These students also autonomously controlled the
chatbots, which further enhanced their motivation (Benson, 2011). For
example, as shown in the comment made by S9, when the student
needed more time to prepare for interacting with the chatbot in English,
he used the stop button and refined his English utterance. S6 controlled
his speaking anxiety by selecting English typing instead of speaking as
the main mode of communication with the chatbot. The interplay
between perceived chatbot control level and students’ willingness to
learn a language through chatbots may need to be investigated more
systematically in future research. Furthermore, language teachers need
to consider how technological affordances can compensate for interac-
tional constraints to maximize the educational effectiveness of chatbots.
Active use of the chatbots for language learning was also promoted
when the chatbots were perceived as pedagogically valuable (Fryer et al.,
2019). To be more specific, when the chatbots were seen as valuable by
the students, the possible technical limitations of the chatbots were
perceived less; these positive perceptions toward the chatbots facilitated
students’ intrinsic motivation to perform the tasks (Dörnyei, 2002;
Freiermuth & Huang, 2012). The chatbots utilized in this study were
not free from technical limitations such as speech recognition problems,
as was the case with chatbots in previous studies. However, as S19
indicated when he stated, “I could learn from the chatbot,” students
who perceived the pedagogical value of chatbots for language learning
maintained their willingness to use the chatbots even when faced with
technical limitations. That is, the perceptions about the chatbots as
learning tools outweighed the possible negative impressions that could
result from the chatbots’ technical limitations. Some students even con-
sidered recognition difficulties that they encountered as opportunities
for pronunciation practice and not as constraints for learning; rather,
this difficulty worked as a form of feedback with which the student
came to realize how to pronounce a word appropriately.
In addition, when perceived as authentic speakers, the chatbots facil-
itated students’ willingness to utilize the chatbots for language learning,
a finding in line with Timpe-Laughlin et al. (2020), who revealed that
teacher perceptions of chatbots as authentic speakers influenced their
willingness to employ the tools in language classes. When students
20 J. JEON

perceived the chatbots as authentic interlocutors with which they could


interact, they were more willing to talk to the chatbots and took advan-
tage of the opportunities that the chatbots provided. In contrast, students
such as S1, who considered the chatbots as just machines, held a negative
attitude toward English classes that utilized the chatbots; the chatbots,
when perceived as just machines, made students unwilling to speak
English. These students ultimately expressed their desire to study with
human interlocutors such as their peers. Hence, methods to make chat-
bots pedagogically more salient and authentic in terms of language
learning may need to be further investigated in future studies.
Finally, the unique configuration of chatbot tasks in which students
mainly interacted with the agents rather than human interlocutors pro-
vided psychological benefits for language learning. Most of the students
revealed positive perceptions toward only interacting with the chatbots
because this configuration created a learning atmosphere that helped to
reduce social anxiety such as peer pressure. This reduced anxiety level
allowed students to speak English more actively (Freiermuth & Huang,
2012; Hill et al., 2015). For example, as was revealed in the interview
data, S2 and S11 directly mentioned that they could comfortably perform
conversations in English because they did not have to worry about
their peers.
Despite the benefits of using chatbots mentioned above, the interview
data also revealed some constraints caused by the chatbots: some stu-
dents were not prepared to speak English to the chatbots; some needed
teacher assistance to perform interactions with the chatbots; some
expressed a stronger desire to study with their peers rather than with
the chatbots. Language teachers could diversify the forms and types of
chatbot activities to address these students’ concerns. For example, mul-
tiple chatbots could be utilized simultaneously during an activity (Fryer
et al., 2020). When students have difficulty understanding the target
language or need L1 instruction, two chatbots that have different lan-
guage settings might help students receive comprehensible guidance in
the language most appropriate at the moment. Furthermore, multiple
chatbots with different language difficulty levels could be created and
employed in a gradual way to provide language practice according to
students’ language ability. Specifically, chatbots that provide vocabulary
practice could be utilized first, and then, students could move on to
using chatbots with which they could have a more open-ended conver-
sation using the vocabulary; this will ultimately lead to the students
acquiring sufficient experience to then be able to perform the conver-
sation with human partners. In addition, language teachers might be
able to design chatbot group activities in which groups of students and
a chatbot interact with one another, a multi-faceted approach that
Computer Assisted Language Learning 21

incorporates both student-chatbot and student-student interaction


(Divekar* et al., 2021). Through this approach, students can take advan-
tage of some benefits that chatbots provide, including receiving imme-
diate feedback or focusing on pronunciation, while realizing their desire
to study with peers. Students may also perceive the use of chatbots
more positively and feel less speaking anxiety (Divekar* et al., 2021).
Along with the pedagogical solutions mentioned above, there also
exist some technical limitations for applying Dialogflow in language
classes; these limitations should be examined by either chatbot technol-
ogy or platform developers. First, multilingual chatbots that can com-
municate using more than two languages currently only work naturally
in the text mode. A chatbot that can communicate both in L1 and L2
using the audio mode might serve as a better teaching tool for
beginner-level students such as young EFL students in this study (Divekar
et al., 2018; Jeon, 2021). In addition, the automatic transcription function
in Dialogflow needs to provide more linguistic information about student
utterances through audio recordings. As presented in Figures 2–4, the
written records do not reveal details regarding student pronunciation
or speaking fluency. An audio record could provide more accurate
information regarding students; thus, teachers might be able to prepare
subsequent learning that is more finely focused on an individual’s prob-
lem areas (Jeon, 2021). Third, as students advance in language profi-
ciency, it will be difficult for teachers to create chatbots that can provide
students with the multitude of potential answers that might occur. The
first step in solving this problem would be for developers to provide
initial coding models designed for the purpose of language learning;
thus, teachers could build on the models to create their chatbots with
more complicated interaction capabilities. This effort should be imple-
mented through close collaboration between developers and language
education experts. Furthermore, teachers might be able to share their
chatbot programs with other teachers, which will eventually contribute
to quality enhancements in customized chatbots. Last, other technical
issues such as more accurate speech recognition for non-native speech
or more authentic speech synthesis functions should be addressed further
(Sydorenko et al., 2019).

Limitations
There are several noticeable limitations in this study. First, this study
examined only a small number of Korean EFL students in the classroom
setting; therefore, the findings may not be broadly generalizable to other
contexts. Future studies need to examine language learners from a diverse
range of proficiency levels, ages, and cultural backgrounds for further
22 J. JEON

theorizing and empirical examination of chatbot affordances and their


effects. Second, the type of conversation topic used for the chatbot tasks
may have affected the emergence of affordance. Future studies need to
consider how affordances emerge differently depending on the topic. In
addition, it would be an interesting topic to explore how students’
experiences or opinions evolve over time. Last, the researcher exclusively
used a type of chatbot that utilized both audio and text modes in English
only. Chatbots with different modalities and language settings might
generate different affordances.

Conclusion
This study aimed to identify and inventory chatbot affordances in pri-
mary EFL classrooms while giving implications regarding how chatbot
technology could be utilized for young EFL students. Various types of
possibilities created by the use of chatbots were confirmed through the
analysis of student interview data and student-chatbot interaction logs.
The analysis has revealed the presence of pedagogical, technological,
and social affordances. In addition, although the chatbot affordances
were provided to all students in the classroom, each affordance was
perceived differently depending on the student. This study also identified
how the chatbot affordances affected various aspects of psychological
states in language learners, particularly regarding their motivation to
learn English through chatbots. By considering the different types of
affordances and their influences on language learners, teachers will be
able to facilitate language learning as professionals who effectively take
advantage of the privileges provided by technological development.
The current study also provides important implications regarding the
future direction of language learning that utilizes chatbots. First, cus-
tomized chatbots created using a chatbot builder were used for the
language course. As shown in this paper, teachers might be able to take
advantage of chatbot builders such as Dialogflow as an effective way of
incorporating chatbots into their language classes (Jeon, 2021; Lee et al.,
2020). To support this effort, teacher training courses in the CALL field
may need to cover both technical and pedagogical elements of chatbot
creation and use. Next, these chatbots have the potential to be utilized
in a considerably broader range within the field of language learning
(Jeon, 2021). Although the chatbots built for this paper were exclusively
administered to only 36 students, some teachers would be able to simul-
taneously apply their own chatbots to much larger numbers of students
once they begin to create their own chatbots. Last, students might be
able to interact with chatbots ubiquitously (Fryer et al., 2020). It would
be possible for students to interact with chatbots anywhere and at any
Computer Assisted Language Learning 23

time if teachers digitally share their chatbot software and have students
download the software to their own devices, such as smartphones.
Chatbots in this environment may yield different affordances and effects
for language learning and teaching.

Notes
1. Websites related to chatbot builders mentioned in this article
-Botsify (developed by Botsify) https://botsify.com/chatbot-for-education
-Chatfuel (developed by Chatfuel) https://chatfuel.com
-Dialogflow (developed by Google) https://dialogflow.com

Disclosure statement
No potential conflict of interest was reported by the authors.

Notes on Contributor
Jaeho Jeon is a doctoral student of English Education at Seoul National University of
Education and a primary school English teacher in South Korea. His research interests
include dynamic assessment, teacher education, and computer-assisted language learning.

ORCID
Jaeho Jeon http://orcid.org/0000-0002-1161-3676

References
AlKhayat, A. (2017). Exploring the effectiveness of using chatbots in the EFL classroom.
In P. Hubbard & S. Ioannou-Georgiou (Eds.),Teaching English reflectively with tech-
nology (pp. 20–36). IATEFL.
Anderson, J. (2015). Affordance, learning opportunities, and the lesson plan pro forma.
ELT Journal, 69(3), 228–238. https://doi.org/10.1093/elt/ccv008
Atwell, E. (1999). The language machine: The impact of speech and language technologies
on English language teaching. British Council.
Benson, P. (2011). Teaching and researching autonomy in language learning (2nd ed.).
Longman. https://doi.org/10.4324/9781315833767
Bibauw, S., François, T., & Desmet, P. (2019). Discussing with a computer to practice
a foreign language: Research synthesis and conceptual framework of dialogue-based
CALL. Computer Assisted Language Learning, 32(8), 827–877. https://doi.org/10.108
0/09588221.2018.1535508
Bower, M. (2019). Technology-mediated learning theory. British Journal of Educational
Technology, 50(3), 1035–1048. https://doi.org/10.1111/bjet.12771
Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative
Research in Psychology, 3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa
Chun, D., Smith, B., & Kern, R. (2016). Technology in language use, language teaching,
and language learning. The Modern Language Journal, 100, 64–80. https://doi.
org/10.1111/modl.12302
24 J. JEON

Coniam, D. (2014). The linguistic accuracy of chatbots: Usability from an ESL per-
spective. Text & Talk, 34(5), 545–567. https://doi.org/10.1515/text-2014-0018
DeKeyser, R. (2010). Practice for second language learning: Don’t throw out the baby
with the bathwater. International Journal of English Studies, 10(1), 155–165. https://
doi.org/10.6018/ijes/2010/1/114021
Divekar*, R. R., Drozdal*, J., Chabot*, S., Zhou, Y., Su, H., Chen, Y., Zhu, H., Hendler,
J. A., & Braasch, J. (2021). Foreign language acquisition via artificial intelligence and
extended reality: Design and evaluation. Computer Assisted Language Learning.
Advance online publication. https://doi.org/10.1080/09588221.2021.1879162
Divekar, R. R., Drozdal, J., Zhou, Y., Song, Z., Allen, D., Rouhani, R., Zhao, R., Zheng,
S., Balagyozyan, L., & Su, H. (2018 Interaction challenges in AI equipped environments
built to teach foreign languages through dialogue and task-completion [Paper presen-
tation]. DIS 2018 - Proceedings of the 2018 Designing Interactive Systems Conference
(pp. 597–610). https://doi.org/10.1145/3196709.3196717
Dörnyei, Z. (2002). The motivational basis of language learning tasks. In: P. Robinson
(Ed.), Individual differences in second language acquisition (pp. 137–158). John
Benjamins. https://doi.org/10.1075/lllt.2.10dor
Dörnyei, Z. (2014). The psychology of the language learner: Individual differences in
second language acquisition. Routledge. https://doi.org/10.4324/9781410613349
Freiermuth, M. R. (2001). Native speakers or non-native speakers: Who has the floor?
Online and face-to-face interaction in culturally mixed small groups. Computer
Assisted Language Learning, 14(2), 169–199. https://doi.org/10.1076/call.14.2.169.5780
Freiermuth, M. R. (2020). Introduction and overview: The inescapable confluence of
technology, psychology and second language learners and users. In M. R. Freiermuth,
& N. Zarrinabadi (Eds.), Technology and the psychology of second language learners
and users (pp. 3–32). Palgrave Macmillan. https://doi.org/10.1007/978-3-030-34212-8
Freiermuth, M. R., & Huang, H. (2012). Bringing Taiwan and Japan closer: A look at
an intercultural online synchronous chat task and its effect on motivation. Language
Teaching Research, 16(1), 61–88. https://doi.org/10.1177/1362168811423341
Freiermuth, M., & Jarrell, D. (2006). Willingness to communicate: Can online chat
help? International Journal of Applied Linguistics, 16(2), 191–213. https://doi.
org/10.1111/j.1473-4192.2006.00113.x
Fryer, L. K., 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. K., Coniam, D., Carpenter, R., & Lăpușneanu, D. (2020). Bots for language
learning now: Current and future directions. Language Learning & Technology, 24(2),
8–22. https://doi.org/10125/44719
Fryer, L. K., Nakao, K., & Thompson, A. (2019). Chatbot learning partners: Connecting
learning experiences, interest, and competence. Computers in Human Behavior, 93,
279–289. https://doi.org/10.1016/j.chb.2018.12.023
Gibson, J. J. (1986). The ecological approach to visual perception. Lawrence Earlbaum.
Hill, J., Randolph Ford, W., & Farreras, I. G. (2015). Real conversations with artificial
intelligence: A comparison between human-human online conversations and
human-chatbot conversations. Computers in Human Behavior, 49, 245–250. https://
doi.org/10.1016/j.chb.2015.02.026
Huang, W., Hew, K. F., & Fryer, L. K. (2021). Chatbots for language learning—Are they
really useful? A systematic review of chatbot‐supported language learning. Journal of
Computer Assisted Learning. Advance online publication. https://doi.org/10.1111/jcal.12610
Computer Assisted Language Learning 25

Hwang, G. J., & Chang, C. Y. (2021). A review of opportunities and challenges of


chatbots in education. Interactive Learning Environments. Advance online publication.
https://doi.org/10.1080/10494820.2021.1952615
Jeon, J. (2021). Chatbot-assisted dynamic assessment (CA-DA) for L2 vocabulary learn-
ing and diagnosis. Computer Assisted Language Learning. Advance online publication.
https://doi.org/10.1080/09588221.2021.1987272
Jia, J. (2003). The study of the application of a keywords-based chatbot system on the
teaching of foreign languages. 1–11. http://arxiv.org/abs/cs/0310018
Johnson, D. (1992). Approaches to research in second language learning. Longman.
Kessler, G. (2018). Technology and the future of language teaching. Foreign Language
Annals, 51, 205–218. https://doi.org/10.1111/flan.12318
Kim, J.-S., & Jang, E.-S. (2020). Artificial intelligence literacy-based teaching conditions and
model for improving primary English language learners’ global digital citizenship.
Association of Global Studies Education, 12(3), 169–198. https://doi.org/10.19037/agse.12.3.06
Lee, S., Noh, H., Lee, J., Lee, K., Lee, G. G., Sagong, S., & Kim, M. (2011). On the
effectiveness of robot-assisted language learning. ReCALL, 23(1), 25–58. https://doi.
org/10.1017/S0958344010000273
Lee, J. H., Yang, H., Shin, D., & Kim, H. (2020). Chatbots. ELT Journal, 74(3), 338–344.
https://doi.org/10.1093/elt/ccaa035
Plonsky, L., & Ziegler, N. (2016). The CALL-SLA interface: Insights from a second-order
synthesis. Language Learning & Technology, 20(2), 17–37. http://llt.msu.edu/issues/
june2016/plonskyziegler.pdf
Sabharwal, N., & Agrawal, A. (2020). Cognitive virtual assistants using Google Dialogflow.
Apress, https://doi.org/10.1007/978-1-4842-5741-8
Smutny, P., & Schreiberova, P. (2020). Chatbots for learning: A review of educational
chatbots for the Facebook Messenger. Computers & Education, 151, 103862. https://
doi.org/10.1016/j.compedu.2020.103862
Soon, G. Y., Warris, S. N., & Al Marimuthu, R. (2020). Chinese language learners’
intrapersonal and interpersonal perceptions of a pinyin text-to-speech system. In M.
R. Freiermuth, & N. Zarrinabadi (Eds.), Technology and the psychology of second
language learners and users (pp. 381–401). Palgrave Macmillan. https://doi.
org/10.1007/978-3-030-34212-8
Stoffregen, T. A. (2003). Affordances as properties of the animal-environment system.
Ecological Psychology, 15(2), 115–134. https://doi.org/10.4324/9780203726655-2
Sydorenko, T., Smits, T. F. H., Evanini, K., & Ramanarayanan, V. (2019). Simulated speak-
ing environments for language learning: Insights from three cases. Computer Assisted
Language Learning, 32(1–2), 17–48. https://doi.org/10.1080/09588221.2018.1466811
Timpe-Laughlin, V., Sydorenko, T., & Daurio, P. (2020). Using spoken dialogue technol-
ogy for L2 speaking practice: What do teachers think? Computer Assisted Language
Learning. Advance online publication. https://doi.org/10.1080/09588221.2020.1774904
Ur, P. (1990). Discussions that work: Task-centered fluency practice. Cambridge University
Press.
van Doremalen, J., Boves, L., Colpaert, J., Cucchiarini, C., & Strik, H. (2016). Evaluating
automatic speech recognition-based language learning systems: A case study. Computer
Assisted Language Learning, 29(4), 833–851. https://doi.org/10.1080/09588221.2016.1167090
van Lier, L. (2004). The ecology and semiotics of language learning: A sociocultural
perspective. Kluwer Academic Publishers.
Xu, Y., Wang, D., Collins, P., Lee, H., & Warschauer, M. (2021). Same benefits, differ-
ent communication patterns: Comparing Children’s reading with a conversational
26 J. JEON

agent vs. a human partner. Computers & Education, 161, 104059. https://doi.org/10.
1016/j.compedu.2020.104059
Zou, B., Li, H., & Li, J. (2018). Exploring a curriculum app and a social communica-
tion app for EFL learning. Computer Assisted Language Learning, 31(7), 694–713.
https://doi.org/10.1080/09588221.2018.1438474
Zou, B., Liviero, S., Hao, M., & Wei, C. (2020). Artificial intelligence technology for EAP
speaking skills: Student perceptions of opportunities and challenges. In M. R. Freiermuth,
& N. Zarrinabadi (Eds.), Technology and the psychology of second language learners and
users (pp. 433–463). Palgrave Macmillan. https://doi.org/10.1007/978-3-030-34212-8

You might also like