Education and Information Technologies (2023) 28:15873–15892

https://doi.org/10.1007/s10639-023-11834-1

Large language models in education: A focus

on the complementary relationship between human
teachers and ChatGPT

Jaeho Jeon1 · Seongyong Lee2

Received: 21 February 2023 / Accepted: 17 April 2023 / Published online: 2 May 2023
© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature
2023

Abstract
Artificial Intelligence (AI) is developing in a manner that blurs the boundaries
between specific areas of application and expands its capability to be used in a wide
range of applications. The public release of ChatGPT, a generative AI chatbot pow-
ered by a large language model (LLM), represents a significant step forward in this
direction. Accordingly, professionals predict that this technology will affect educa-
tion, including the role of teachers. However, despite some assumptions regarding its
influence on education, how teachers may actually use the technology and the nature
of its relationship with teachers remain under-investigated. Thus, in this study, the
relationship between ChatGPT and teachers was explored with a particular focus on
identifying the complementary roles of each in education. Eleven language teachers
were asked to use ChatGPT for their instruction during a period of two weeks. They
then participated in individual interviews regarding their experiences and provided
interaction logs produced during their use of the technology. Through qualitative
analysis of the data, four ChatGPT roles (interlocutor, content provider, teaching
assistant, and evaluator) and three teacher roles (orchestrating different resources
with quality pedagogical decisions, making students active investigators, and raising
AI ethical awareness) were identified. Based on the findings, an in-depth discussion
of teacher-AI collaboration is presented, highlighting the importance of teachers’
pedagogical expertise when using AI tools. Implications regarding the future use of
LLM-powered chatbots in education are also provided.

Keywords ChatGPT · Large language model · Chatbot · AIEd · Human–computer

interaction · Artificial intelligence · Large language model-powered chatbot

* Seongyong Lee
seongyonglee77@gmail.com
Extended author information available on the last page of the article

13
Vol.:(0123456789)
15874 Education and Information Technologies (2023) 28:15873–15892

1 Introduction

The use of chatbots to enhance students’ language learning experience has gained
much attention in recent years (Huang et al., 2023; Hwang et al., 2020; Jeon,
2021; Jeon, 2022a; Lee & Jeon, 2022). Language researchers have employed dif-
ferent types of chatbots, including those produced commercially and those devel-
oped in research labs, and have identified novel educational opportunities pro-
vided by chatbots (Dizon, 2020; Fryer et al., 2020; Huang et al., 2022). Although
research has consistently shown that chatbots can facilitate language learning,
significant limitations inherent in current chatbot systems have also been revealed
(Bibauw et al., 2019; Jeon, 2022b). These include chatbots’ limited ability to
engage in open-ended conversations with learners and to sustain an extended and
goal-oriented conversation on a specific topic (Kuhail et al., 2022).
The recent public release of ChatGPT has marked an important advancement
in the field of artificial intelligence (AI), specifically in natural language process-
ing (NLP) (Brown et al., 2020). ChatGPT, a generative AI chatbot powered by a
large language model (LLM) that has been trained on vast amounts of internet
text data, is projected to overcome many limitations of previous chatbot technol-
ogy and ultimately affect the way people learn (Heidt, 2023; Kasneci et al, 2023).
In contrast to the majority of existing chatbots, which adhere to predefined dia-
logue paths or feature simplistic question-and-answer dialogue structures (for a
review on existing educational chatbots, see Kuhail et al., 2022), this exponen-
tially advanced chatbot can generate answers according to the context of a given
prompt, and thus, can engage in a conversation that is more akin to human–human
interaction than those of previous chatbots (Susnjak, 2022).
The popularity of ChatGPT, as evidenced by the rapidly growing number of
registered users (Altman, 2022), suggests that this type of technology is poised
to become increasingly integrated into society. Accordingly, educators anticipate
a significant shift in various aspects of education, including the role of teachers
(Stokel-Walker, 2023). As it proves able to provide novel resources that enrich
learning experiences beyond conventional methods and ones provided by cur-
rent chatbot technology (Jeon, 2022b), it may take over some roles traditionally
carried out by teachers. For example, as suggested in initial theoretical explora-
tions on the technology (Kasneci et al, 2023; Zhai, 2022), teachers may utilize the
features of this chatbot to stimulate students’ engagement with textbook content.
Also, teachers may assign the task of initially grading student essays to the tech-
nology and focus their efforts on providing more detailed feedback. Conversely,
some scholars, such as Kasneci et al. (2023), expressed a concern that some
teachers may become excessively reliant on ChatGPT and fail to ensure opportu-
nities to foster students’ creativity, critical thinking, and problem-solving skills.
However, despite such predictions on the impact of ChatGPT, its potential as a
learning tool, how teachers utilize it, and the nature of the relationship between
teachers and this technology have yet to be empirically explored.
In response to this research need, we conducted a qualitative investigation
on the educational potential of ChatGPT in the context of language education.

13
Education and Information Technologies (2023) 28:15873–15892 15875

Specifically, the perspectives and experiences of language teachers who deployed

ChatGPT in their instruction were examined. The findings of this inquiry may
contribute to our understanding of the pedagogical implementation of LLM-
powered chatbots, the future development of the technology, and the professional
development needs of teachers in the utilization of emergent forms of AI tools in
education.

2 Literature background

2.1 Chatbots for language learning: From ELIZA to ChatGPT

Since the development of ELIZA, the initial version of NLP-powered chatbots

(Weizenbaum, 1966), chatbots have undergone several technological advances,
thereby increasing their potential for pedagogical use as a tool for supporting stu-
dents’ language learning. Jeon (2022b) identified three types of chatbots that have
been employed in the context of language learning: (1) general-purpose chatbots
that can engage in daily, simple conversation with users via question-and-answer
dialogue structures (e.g., Dizon, 2020; Fryer et al., 2019), (2) specific-purpose
chatbots with more complex structures designed by commercial developers for lan-
guage learning (e.g., Wang et al., 2023), and (3) customized chatbots created by
researchers or teachers for particular language learning contexts using visual chat-
bot development platforms (e.g., Hew et al., 2023; Lee & Jeon, 2022).
Recent research on these types of chatbots has consistently demonstrated the
efficacy of chatbots in language learning, as evidenced in Bibauw et al.’s (2022)
meta-analysis of previous chatbot studies. Meanwhile, several limitations in chatbot
research have also been identified. First, little is known about the possible roles of
chatbots in language education beyond that of a conversation practice partner that
follows a limited number of pre-determined dialogue paths (Huang et al., 2022; Ji
et al., 2023). Despite the potential for chatbots to assume a variety of more com-
plex roles (Kasneci et al, 2023), such as facilitating the organization of students’
thoughts, providing lesson plan information to teachers, and assisting with the
preparation of instructional materials, these possibilities have not been developed
or remain either at a theoretical or a prototype level (Bibauw et al., 2019). Second,
chatbots previously examined were unable to participate in open-ended conversation
comprising several turns with users, which restricted meaningful exchanges between
chatbots and users to only a few turns (e.g., Jeon, 2021; Jeon, 2022b). As indicated
by Bibauw et al. (2019) and Kuhail et al. (2022), the limitations of current chatbots
can be attributed, at least in part, to the rule-based design structure for conversation
and the limited sources of data to be fed into chatbot systems that were available to
developers at the time of chatbot creation. These limitations resulted in the chatbots’
inability to respond appropriately to a wide variety of user utterances.
The recent release of ChatGPT, a highly advanced version of a generative AI
chatbot powered by an LLM, has successfully addressed some of the limitations of
existing chatbots. An LLM, a type of machine learning model that is trained with a
vast dataset of human language, significantly upgraded the overall performance of

13
15876 Education and Information Technologies (2023) 28:15873–15892

NLP applications (Brown et al., 2020). Thus, it enables chatbots to generate a wider
range of human-like responses than previously possible. Specifically, being built on
LLM technology provides ChatGPT with the following three capabilities (OpenAI,
2023): (1) the ability to remember previous statements made by the user during the
conversation; (2) the ability to comprehend follow-up corrections by the user; (3) the
ability to decline inappropriate requests. These features enable ChatGPT to simu-
late a more human-like conversation and respond to a wide range of utterances if
deemed appropriate. Moreover, the chatbot can maintain potentially limitless con-
versational turns in a goal-oriented conversation on a specific topic. These capabili-
ties, the result of intensive development in NLP, suggest that LLM-powered chat-
bots may be adapted to a wider range of educational roles than previous forms of
chatbots (Kasneci et al, 2023). In sum, the pedagogical use of LLM-powered chat-
bots has now become a viable option for implementation in education, including
language learning.

2.2 Complementary relationships between humans and AI in education

Along with the recent proliferation of AI into classroom contexts, scholars have paid
much attention to human-AI collaboration, positing that more effective learning can be
jointly created by human facilitators and AI than by humans or AI working alone (Hol-
stein et al., 2020; Kim et al., 2022; Xu & Ouyang, 2022). In this line of inquiry, teachers’
agentive roles are emphasized and described as facilitating the positive impact of AI on
education, rather than AI being considered as a potential substitute for teachers (Bower,
2019; Jeon et al., 2022). For example, using frameworks for adaptivity in education,
Holstein et al. (2020) suggested four categories as areas in which human teachers may
enhance the adaptability of AI technology, including goal augmentation, perceptual aug-
mentation, action augmentation, and decision augmentation. In this theoretical framework
for teacher-AI collaboration, they underscored the extensive roles of teachers in maxi-
mizing the effectiveness of AI. Of more relevance to the current research, Ji et al. (2023)
reviewed 24 research studies on chatbots for language learning, focusing on the collabora-
tion between teachers and chatbots. They found that only a limited amount of empirical
evidence existed regarding teacher-chatbot collaboration. Based on the idea of classroom
orchestration (Dillenbourg & Jermann, 2010), they suggested that researchers examine the
potential that teacher-chatbot collaboration may offer, calling for more empirical research
on how the collaboration can be conducted, how it facilitates learning, and how AI can
reduce teachers’ workloads.
In addition to these theoretical explorations, some, albeit limited, empirical
research demonstrated how human teachers can actually complement AI (e.g.,
Holstein et al, 2019). For example, in Holstein and Aleven (2022), teachers
utilized smart glasses that provided real-time data on student learning during
class. They observed, “teachers then made a rich inference about the latent,
underlying cause of the behavior and responded with support and flexibility
that the AI tool could not provide” (p. 43). This observation shows that it was
not the AI tool itself that facilitated learning, but rather the teachers’ pedagogi-
cal expertise that was strengthened by the AI-generated data. On a similar note,

13
Education and Information Technologies (2023) 28:15873–15892 15877

Zhang et al. (2022) offered another illustration on how collaboration between

humans and AI might promote student learning. In a book-reading scenario
involving an AI-powered storytelling app, parents, acting as classroom teach-
ers, were provided with a series of individualized story questions via the app.
Referring to those questions, parents could flexibly choose the most pertinent
question at any given moment while ignoring irrelevant questions; this agentive
action of parents promoted a rich conversation between parents and children.
These cases highlight the importance of human teachers’ pedagogical decisions
when using AI in education with empirical evidence, enriching our understand-
ing of teacher-AI collaboration in education. However, despite their contributions,
there remains an important gap in the literature, particularly concerning the fea-
tures of AI tools used. Previous research has lacked relevant pedagogical insights
regarding the rapidly changing landscape of AI technology, which blurs the
boundaries between specialized areas of application. (Brown et al., 2020; Kasneci
et al., 2023). That is, previous studies have primarily focused on tools that targeted
specific aims during a particular instructional stage, as shown above in the cases
of smart glasses and a storytelling app (Holstein & Aleven, 2022; Zhang et al.,
2022). This emphasis on AI tools tailored to target specific educational aims has
also been found in several review studies conducted in the field of AI in education
(AIEd) (e.g., Celik et al., 2022; Hwang et al., 2020). For example, in a compre-
hensive review of 146 studies, Zawacki-Richter et al. (2019) found 150 AI appli-
cation cases, which they categorized under four main application areas (i.e., profil-
ing and prediction, intelligent tutoring systems, assessment and evaluation, and
adaptive systems and personalization). This suggests that in most of the studies,
investigations of AI tools were confined to specific roles within these four applica-
tion areas. The tendency noted above is also salient in the literature that examined
the role of chatbots in education (e.g., Ji et al., 2023). For example, Kuhail et al.’s
(2022) review of studies on educational chatbots revealed that 37 chatbot applica-
tion cases found in 36 studies were described as performing one or more of four
distinct roles (i.e., teaching agent, peer agent, teachable agent, and motivational
agent), again suggesting the specialized purposes of AI tools. That is, except for
one study, chatbots in the reviewed studies were designed to assume only one of
these four roles.
Consequently, scholarly examination of the rapidly changing field of AI technol-
ogy may have been less exhaustive than necessary. As illustrated above, when teach-
ers utilize AI tools that are specifically designed to achieve particular educational
goals, their responsibilities may be limited to supplementing the tools’ capabilities
(Holstein & Aleven, 2022; Zhang et al., 2022); this may not necessarily reflect the
nature of the emerging relationship between teachers and ChatGPT, which has the
potential to be used in a variety of application areas and contexts (Kasneci et al.,
2023; Zhai, 2022). As demonstrated by ChatGPT, AI is expected to undergo a para-
digm shift in which it will be utilized with increasing versatility. This evolution may
render the teacher-AI relationship in education ever more dynamic and difficult to
predict than before (Kasneci et al., 2023; Zhai, 2022). However, how evolving AI,
such as ChatGPT, can be integrated into education and what roles teachers will play
as a result remains underexplored.

13
15878 Education and Information Technologies (2023) 28:15873–15892

Therefore, this exploratory study aimed to investigate language teachers’ uses

and perceptions of ChatGPT to examine the pedagogical potential of this recently
released LLM-powered chatbot. Specifically, this study attempted to identify the
roles of ChatGPT in language education and how it may influence teachers’ roles.
The following research questions guided the study:

1. What roles can an LLM-powered chatbot play in language education?

2. What roles do teachers perceive as becoming more important with the use of
LLM-powered chatbots?

3 Methodology

We adopted an exploratory qualitative approach to gain insight into the pedagogi-

cal value of an LLM-powered chatbot and how teachers perceived their roles would
change with the use of the technology (Creswell, 2008). To achieve this goal, we
collected data from two sources: individual semi-structured interviews as a primary
source and teachers’ chatbot-use logs as an objective supplement to the interview
data.

3.1 Participants and chatbot

Eleven English language teachers from ten elementary schools in South Korea were
recruited through ads posted on a social online forum for elementary school teach-
ers (Table 1). We selected participants representing a variety of ages, genders, and
teaching experiences from among those who contacted us and expressed their will-
ingness to apply ChatGPT in their teaching. Prior to the study, none of the partici-
pating teachers had used ChatGPT for any purpose, and only one had heard about it
through a news platform.

Table 1  Participant profiles Teacher Age Gender Years of

teaching
experience

1 35 Female 8
2 31 Female 5
3 32 Male 5
4 35 Female 7
5 30 Female 4
6 27 Female 2
7 34 Female 8
8 34 Female 9
9 38 Female 11
10 36 Male 9
11 32 Female 5

13
Education and Information Technologies (2023) 28:15873–15892 15879

The latest version of an LLM-powered chatbot for text generation, ChatGPT, also
called GPT-3.5 in the family of LLMs, was released by OpenAI for public use on
November 30, 2022. It is built on OpenAI’s generative pre-trained transformer-3.5
and is being continually fine-tuned with both supervised and reinforcement learning
techniques (OpenAI, 2023).

3.2 Procedures

Figure 1 illustrates the procedures followed in this research. The recruited teach-
ers adhered to the procedures individually, with their utilization of ChatGPT tak-
ing place throughout the months of January to February 2023. The teachers first
participated in a 60-min seminar provided by one of the researchers, who first
informed the participants of the purpose and procedures of the research, including
the handling of their interview responses and interaction log data in future reports
and publications. They were aware of their rights to withdraw at any stage and
provided their informed consent. For about 20 min, the participants were intro-
duced to ChatGPT and learned about the basic functions of the chatbot, including
how to create an account and access the chatbot and dialogue histories. They were
also shown how the chatbot responded to different types of questions. The rest of
the session, which lasted for around 40 min, involved a Q&A discussion in which
the presiding researcher and a teacher fielded questions from the participants, such
as whether certain educational prompts would be recognized or how they could
craft questions more accurately to obtain specific information. To respond to their
questions, the researcher and teacher collaborated to seek answers by inputting
specific prompts or questions and sharing their opinions about the outcomes. For
the two weeks following the seminar, the participant teachers used ChatGPT how-
ever they chose in their own teaching contexts. It was emphasized that there were
no constraints on the ways of using ChatGPT as long as it was appropriately used
for facilitating teaching and learning. They were also asked to collect up to 10
noteworthy or debatable methods of use in the form of interaction logs with the
chatbot. After the two weeks, the teachers participated in individual semi-struc-
tured interviews with one of the researchers in which they shared their experiences
with and perceptions of the use of ChatGPT. The interviews lasted for an aver-
age of 60 min and were audio-recorded and transcribed, with parts translated into
English.

Fig. 1  Research procedures

13
15880 Education and Information Technologies (2023) 28:15873–15892

3.3 Data collection and analysis

Data comprised the transcripts of the semi-structured in-depth interviews and the
teachers’ interaction logs with the chatbot. To guide the interview process, semi-
structured protocols were developed based on the interview guide used by Timpe-
Laughlin et al. (2022) in a study of teachers’ experiences with and perspectives on
the use of a chatbot system. The interviews were structured around the following
questions:

– Can you tell me how you used the chatbot?

– Did you find using chatbots helpful when teaching English? If so, what functions
of the chatbot were the most useful?
– What parts of your pedagogical work do you think can become easier/more
important with the use of the chatbot?
– Do you think teachers are likely to be replaced by this type of technology? Why
or why not?

Given the exploratory aim of the study, initial questions were designed to elicit
teachers’ general perceptions of and experiences with ChatGPT. Subsequent follow-
up questions focused on the roles assumed by the teachers while using ChatGPT.
The transcribed interview data were subsequently subjected to qualitative anal-
ysis in order to identify recurring patterns of meaning that were pertinent to the
research questions. In accordance with the methodology proposed by Braun and
Clarke (2006), the analysis proceeded through several steps. First, we independently
read through the transcribed data in order to gain a general understanding of the
data. Next, utilizing the interview data from two of the participant teachers, we
each developed an initial coding scheme by reviewing the transcripts for emergent
themes, with a particular focus on the two research questions (i.e., specific appli-
cations of ChatGPT and teacher roles) and identifying subcategories within each
theme. We compared our individual analyses and resolved discrepancies through an
iterative process of repeated rounds of discussion. We then independently applied
the reconciled coding to the remaining interview data. Upon calculating the results
of the coding process, we obtained a Cohen’s Kappa value of 0.89, which demon-
strates a high level of inter-coder reliability as suggested by Lombard et al. (2002).
We then resolved all remaining discrepancies through further discussion. To iden-
tify and examine patterns and thematic trends in the data, we calculated frequency
counts of responses for coding. Representative answers were extracted to capture
response patterns in the words of the teachers.
The teacher-chatbot interaction logs were then analyzed to identify the roles per-
formed by ChatGPT and to supplement the interview data. All of the interaction
transcripts provided by the teachers were thoroughly reviewed. In the interest of
brevity, and as chatbot responses can be directly accessed by readers through the
chatbot using the prompts the teachers provided (visit https://​chat.​openai.​com/​chat
for chatbot use), here we provided only teacher prompts extracted from the inter-
action logs. The selected prompts were presented alongside interview excerpts to
verify the chatbot’s roles.

13
Education and Information Technologies (2023) 28:15873–15892 15881

4 Findings

4.1 Roles of ChatGPT in language education

Based on the teachers’ interview data and interaction transcripts, four roles of Chat-
GPT were identified: interlocutor, content provider, teaching assistant, and evalua-
tor. Following are descriptions of how each role was performed by the chatbot.

4.1.1 Interlocutor

Table 2 provides sample prompts utilized by participant teachers that elicited the
chatbots’ performance of specific functions in the role of interlocutor. Ten teach-
ers utilized the chatbot as an interlocutor, specifically as a role-player, and/or as an
interactive game partner. The teachers indicated that they employed the role-play
function of the chatbot to provide students with language experiences. To initiate
role-plays with chatbots, the teachers provided various prompts such as “Act as a
person…” All 10 teachers indicated that they demonstrated how a conversation with
the chatbot would proceed and allowed students to contribute to the conversation,
while two teachers also had groups of students or individual students practice using
the chatbot independently.

Table 2  ChatGPT as Functions Sample prompts

interlocutor
Role-playing • Act as a waiter at a pizza
restaurant, then I will act as a
customer who wants to order
• Can you role-play with me?
↳The scene is that I call my
friend, Minho, and you answer
the phone. You are Minho in
this role-play. Can you go first?
Interactive language game • Let’s play a color-guessing
game. Give me a hint then I will
guess
• Let’s play a word-guessing
game
↳Choose one word from among
grade, club, sport, baseball,
basketball, and volleyball, and
give me a hint. I will guess
• I want to play a 20-question
game
↳Choose a basic word for 10-year-
old EFL students and give me
a hint
• Let’s play a would-you-rather
game
↳The topic is food

The arrow “↳” indicates a prompt following a chatbot’s response

13
15882 Education and Information Technologies (2023) 28:15873–15892

Additionally, eight teachers reported having the chatbot perform the role of inter-
locutor as an interactive language game partner. The games included specific word-
guessing games, including a color game, an animal guessing game, a 20-question
game, and a would-you-rather game. Teacher 3 noted the benefit of the chatbot par-
ticipating in such games, stating, “When I used language games before, it used to be
just me versus students, but now each student can do it with the chatbot, and they
can have many more opportunities to use English.”

4.1.2 Content provider

As shown in Table 3, the chatbots also performed three different functions in the
role of content provider: production and recommendation of materials, customiza-
tion of materials, and provision of cultural knowledge. All 11 teachers used the chat-
bot to produce materials, including dialogue scripts, short stories, and sample words
or sentences. Regarding this function, Teacher 3 mentioned, “In terms of resource
availability, we now have many more options that we can choose from. We do not
need to stick to the textbook or ready-made online materials.” Additionally, some
teachers leveraged the chatbot’s expertise to recommend existing materials that were

Table 3  ChatGPT as content provider

Functions Sample prompts

Production and recommendation • I want to do a two-member role play. This role-play is for practic-
of materials ing the question-and-answer structure, “What are you doing?” and
“I’m …” Can you make some scripts to do the role-play using only
simple English?
• I can pronounce “good.” Can you search for other words having the
“oo” sound in the word “good”?
• Can you recommend short English cartoon books for ­5th-grade EFL
students?”
• Let me know the names of Disney songs about self-esteem
• Can you teach me how to pronounce “interesting” in a Korean
accent?”
Customization of materials • I’m an elementary student in Korea and I don’t use English so well.
Can you use only simple words if you have the vocabulary list of
Key Stage 1?
↳I just want you to use simple English like the above vocabulary
• Can you write a story using the question “Are you okay?” and the
answer “No, I’m not.”?
↳I want you to use only simple sentences. And please write the story
in 15 sentences
Provision of cultural knowledge • Create simple dialogues about “What grade are you in?”
↳Can you add some cultural misunderstanding in the dialogue above?
• Do you know the book “The Very Hungry Caterpillar”?
↳Can you show me some sample sentences from the book?
↳Can you exchange the words about the food above for Italian food?
• Do people in the US take off their shoes in their homes?
↳What about Singapore?

The arrow ↳ indicates a prompt following a chatbot’s response

13
Education and Information Technologies (2023) 28:15873–15892 15883

tailored to the specific needs of their students, such as English storybooks on spe-
cific topics that were otherwise difficult to find.
All the teachers (n = 11) also mentioned that the chatbot could modify materials
it produced or existing materials. This function was particularly useful for teachers
who desired to provide individualized materials to students. They all expressed that
the chatbot’s ability to generate and revise different levels and types of materials
was a primary reason for them to want to continue to use it. For example, Teacher 5
said, “I do not need to spend a lot of time on developing individual materials as in
the past. I can simply ask the chatbot to make more difficult or easier versions of a
material.”
Last, some teachers (n = 6) stated that they could acquire cultural knowledge from
the chatbot or ask the chatbot to make materials culturally relevant. The teachers
stated that they used to have difficulty understanding the cultures of different coun-
tries. However, they found that the chatbot could not only provide cultural knowl-
edge but also produce or modify lesson materials in manners that reflected cultural
aspects, thereby making their language classes more engaging and culturally rele-
vant at the same time than before.

4.1.3 Teaching assistant

Table 4 displays specific functions in the chatbots’ role as teaching assistant and
sample prompts provided by participant teachers. The teachers used the chatbot
to help students resolve learning difficulties by allowing them to interact with the
chatbot during the class or using it themselves. The chatbot performed the role of
teaching assistant in three different ways. All 11 teachers interviewed reported uti-
lizing the chatbot as a grammar checker or online dictionary. Among them, two
teachers further stated that they also allowed students to use the chatbot during
class to encourage them to find errors in their vocabulary or sentences or search for
word meanings whenever necessary. In addition, a few teachers (n = 2) indicated
that they used the chatbot as a background knowledge activator to give students a
preview of what they were going to learn about a topic.

Table 4  ChatGPT as teaching assistant

Functions Sample prompts

Grammar checker • I wrote the answer “I want go to Jeju-do” to the question

“What do you want to do?” Is there any error in my answer?
• Can you check for any grammar mistakes?
Dictionary • I want to know about specific contextual examples with the
words “happy,” “excited,” and “exhilarated.” What are the
differences among those words?
• What are some phrases that I can use instead of “How are
you”?
• What is the difference between “I am heading” and “I am
headed”?
Background knowledge activator • Can you tell me the origin of “banana”?
• Can you show the words related to the category of “clothes”?

13
15884 Education and Information Technologies (2023) 28:15873–15892

4.1.4 Evaluator

Table 5 shows specific functions the chatbots performed in the role of evalua-
tor and sample prompts provided by teachers. Two teachers used their chatbots
to provide initial grading of students’ writing. Regarding this role, Teacher 10
stated that she elaborated on some feedback in the chatbot’s initial assessment
and ignored some information that she deemed irrelevant. Meanwhile, many of
the teachers (n = 9) stated that they used the chatbot to produce testing materials.
For example, the teachers stated that they asked the chatbot to produce multiple-
choice questions, O/X questions, and gap-filling dialogues on specific topics.
Only two teachers noted that the dialogue records automatically saved on the
chatbot website provided useful information about student performance. Teacher
7, who distributed tablet PCs to allow students to use the chatbot individually,
mentioned:
After a lesson, I brought back tablet PCs that I gave to students and examined
the dialogue history with each student to determine the degree to which the
student performed well or not. In this manner, I could obtain more accurate
assessment information about my students.

4.2 Teachers’ roles when using ChatGPT

In the interviews, all teachers acknowledged the educational potential of ChatGPT

and expressed their willingness to use it in the future. They also agreed that Chat-
GPT would not replace human teachers, but rather support them in maximizing
their professional expertise. That is, although they found ChatGPT to be a power-
ful tool that could automate many aspects of instruction, it was still just perceived
as a tool that could enhance teachers’ pedagogical effectiveness but that could not
substitute for teacher-student interaction. Rather, the teachers believed that their
teaching profession might need to be redefined with the assistance of this versatile
technology in the following three roles of teachers, which emerged as important in
the interviews.

Table 5  ChatGPT as evaluator

Functions Sample prompts

Assessment • This is a sixth-grade elementary EFL student’s daily

diary. Can you suggest some feedback for her?
Evaluation material production • Let me know the full lyrics of the song, “Let It Go.”
↳Make 10 gaps with a few words above
• (After inputting a story from a textbook)
↳Make 10 O/X questions about it
Dialogue history Teachers assess student performance based on dia-
logue histories automatically saved in ChatGPT

The arrow ↳indicates a prompt following a chatbot’s response

13
Education and Information Technologies (2023) 28:15873–15892 15885

4.2.1 Orchestrating different resources while making quality pedagogical decisions

All participants agreed that using ChatGPT would require teachers to become highly
skillful at managing the plethora of teaching resources afforded by the technology
so as to design creative, organized, and engaging lessons. With ChatGPT, teachers
would now have significantly more control over creating and revising lesson materi-
als than in the past, when they had access to fewer teaching resources and had to
manually develop and revise materials.
However, the teachers also noted that the increased availability of helpful
resources would not automatically raise the quality of instruction. They mentioned
that it depended on human teachers’ pedagogical knowledge and judgment in select-
ing materials appropriate to their students’ needs and how effectively they used them
within their instructional contexts. As Teacher 8 stated, “Although we can much
more easily generate and modify lesson materials using ChatGPT, we must carefully
consider how we will use the materials in the classroom,” predicting that “there will
be a significant disparity in the quality of lessons between teachers who possess the
ability to effectively integrate the chatbot into their pedagogy and those who lack
such proficiency.” Similarly, emphasizing the importance of the connection between
teachers and students, Teacher 9 stated:
ChatGPT provides useful materials, but the ultimate acceptance of these
materials by students can only be determined through a process of thoughtful
reflection, taking into account factors such as the rapport between the teacher
and students, individual students’ needs and characteristics, and the collective
dynamic of the classroom.
In conclusion, the teachers made it clear that ChatGPT has the potential to serve
as a valuable instructional tool that can enrich teaching practice, but they empha-
sized that it would need to be employed in a manner that complements and enhances
teachers’ pedagogical expertise based on the nuanced understanding of their stu-
dents and contexts.

4.2.2 Making students active investigators, not passive knowledge recipients

With regard to teachers’ roles in using ChatGPT to promote students’ learning, nine
teachers emphasized the importance of formulating good questions and modeling
effective questioning techniques to obtain high-quality results from the chatbot.
Teacher 1 stated, “The ability to ask a question or in some instances a series of ques-
tions determines whether we can obtain desired information or not.” Also, regarding
this issue, Teacher 8 shared a story about developing good questioning strategies:
I was creating a pronunciation guide for basic vocabulary for my beginning-
level students. I wanted the chatbot to provide me with written pronunciations
of English words in Korean, for example, the English pronunciation of ball
can be written in Korean as 볼 [bol]. I initially attempted prompts such as
“Can you tell me the pronunciation in Korean with this word, interesting?” or

13
15886 Education and Information Technologies (2023) 28:15873–15892

“Can you teach me how to pronounce interesting in a Korean accent?” I was

unsuccessful and thought the chatbot was not able to do it. I gave up. Later, a
coworker suggested trying the following prompt: Can you write the English
pronunciation of interesting in Korean? It worked and I got the information I
wanted.”
In a similar vein, the teachers mentioned the importance of teaching students
how to formulate questions effectively, which they believed influenced the extent to
which students would benefit from their interactions with the chatbot. Teacher 4, for
example, stated,
When I encouraged my students to interact with the chatbot, some of them
actively asked questions about what they wanted to know, experimenting with
various question formats. However, I also discovered that some students did
not even know how to ask questions. The most important thing that we teach-
ers should do is to help them make appropriate questions and craft them to
obtain the best information.
Four teachers said that they prepared worksheets consisting of questions for stu-
dents to ask the chatbot, but Teacher 11 stated, “it was not possible to anticipate
every question that students might ask.” Instead, they all underscored the importance
of the role of teachers that prepare students to construct and craft their own ques-
tions depending on their needs. That is, with the rich language resources of Chat-
GPT, the teachers agreed that it was now even more important not to treat students
as passive knowledge recipients; rather, teachers shared that they would need to cre-
ate a learning environment where students can be active investigators who learn by
skillfully developing a variety of questions about the subject matter on their own.

4.2.3 Raising ethical awareness regarding the use of AI

All teachers interviewed emphasized the importance of ethical considerations in

the use of ChatGPT and guiding students to use the technology responsibly. In this
regard, two aspects of ethical awareness emerged in the interviews: making the use
of the technology transparent and setting rules for ethically interacting with it.
First, all teachers emphasized the need for establishing a shared understanding
of the degree to which students might draw on ChatGPT, while creating an atmos-
phere where students can openly talk about what information they obtained from the
chatbot. Granting that it would not be feasible to monitor all students’ uses of the
technology, some mentioned that they should instruct students to not merely rely
on ChatGPT to obtain what they need to complete tasks; instead, they argued that
teachers should acknowledge students’ efforts to obtain information from the chatbot
while encouraging them to use it to further develop their own understanding. Spe-
cifically, Teacher 2 elaborated on this:
Teachers have to make a classroom atmosphere where using the technol-
ogy itself is not ethically wrong but where the behavior of hiding how they
got information from it and pretending that the knowledge gained from it is

13
Education and Information Technologies (2023) 28:15873–15892 15887

entirely their own is wrong. Students do not need to feel that they should hide
their use of the chatbot. Rather, we have to acknowledge students’ efforts to
use the chatbot but only in a way that they can further their own learning. To
do this, I believe we have to first make it acceptable for students to openly talk
about what they obtained from the chatbot and how they used it to facilitate
their learning experience.
Next, most of the teachers (n = 10) also expressed concerns about inappropriate
or unethical student behavior in interactions with the chatbot and shared anecdotes
regarding such behavior. Specifically, these teachers observed that some students
tried to use inappropriate words with the chatbot or tried to obtain inappropriate
information from it. Specifically, to address this issue, Teacher 7 mentioned, “A
clear guideline outlining appropriate and ethical usage of the chatbot must be estab-
lished before we introduce it to students.” Overall, the consensus among the teachers
was that ChatGPT should be implemented in conjunction with thoughtfully com-
posed ethical guidelines covering a range of considerations from the utilization of
information obtained through the technology to the manner in which students should
interact with it.

5 Discussion

This research explored the educational potential of ChatGPT and language teach-
ers’ perceptions on how their roles would change as a result of its use. By qualita-
tively analyzing the teachers’ chatbot-use logs and conducting in-depth interviews,
we were able to identify the dynamic nature of the relationship between teachers
and AI, with both parties found to complement each other. The findings highlight
that as technology becomes increasingly multifunctional, teachers may need to
assume more critical and professional roles to integrate the technology in a way that
best benefits students. Therefore, this study provides important areas for discussion
regarding both chatbots and teacher-AI collaboration in education.
Previous studies on educational chatbots have consistently identified the numer-
ous limitations of chatbot technology, such as its limited capability to engage in
open-ended conversations with users and to continue an extended, goal-oriented
conversation on a particular topic (Bibauw et al., 2019; Jeon, 2022b). In response
to these limits, the current research introduced an LLM-powered chatbot, as it can
simulate a more human-like conversation and respond to a broader range of user
inquiries. Furthermore, the augmented capabilities enabled the chatbot to perform
various pedagogical roles compared to chatbots previously introduced in the litera-
ture (Ji et al., 2023). For example, Kuhail et al. (2022) discovered that the majority
of educational chatbots introduced in research mainly fulfilled one of four roles (i.e.,
teaching agent, peer agent, teachable agent, and motivational agent). Similarly, Ji
et al. (2023) showed that the primary role of chatbots in language education was the
role of interlocutor. In contrast to these studies, teachers in the current study utilized
ChatGPT in a broader capacity across several instructional phases. In other words,
ChatGPT served not only as a content provider, teaching assistant, and evaluator, but

13
15888 Education and Information Technologies (2023) 28:15873–15892

also assumed the role of a conversation practice partner as an interlocutor, showing

its versatility and potential to play multiple roles across various instructional con-
texts (Celik et al., 2022). Overall, this study presented empirical evidence regarding
the use of ChatGPT in education, enriching our understanding of how LLM technol-
ogy may be incorporated into the educational context.
Regarding the relationship between the teachers and the chatbot, this study’s
findings are consistent with earlier research that supported teacher-AI collabora-
tion (Holstein & Aleven, 2022; Holstein et al., 2020; Kim et al., 2022). Research on
teacher-AI collaboration has shown the importance of teachers’ agentive roles when
using AI. It argues that more effective learning is jointly created by AI and human
facilitators, rather than teachers or AI working alone (Bower, 2019; Holstein et al.,
2020). For example, Holstein and Aleven (2022) showed that teachers could make
better decisions based on students’ past learning data provided by smart glasses.
Zhang et al. (2022) illustrated that parents created a better storytelling environment
by selecting the most pertinent questions at any given moment among different
questions generated by AI. Similarly, we demonstrated that teachers’ pedagogical
expertise could be maximized with the help of the capabilities of the chatbot. It is
of note that this study presented contradictory evidence to the concern that teach-
ers may become overly dependent on LLM technology (Kasneci et al, 2023). Con-
versely, teachers in this study stated that without adequate student training and clear
guidelines, students would not be able to use the chatbot in an effective and ethi-
cally appropriate manner, suggesting that the use of the chatbot may not diminish
the importance of teachers’ roles.
Additionally, by using ChatGPT, which can be utilized across many applica-
tion domains, this study expands on the findings from earlier studies that examined
teachers’ relationship with AI tools built with a particular educational purpose (e.g.,
Holstein & Aleven, 2022; Zawacki-Richter et al., 2019; Zhang et al., 2022), pre-
senting empirical evidence that the increasing versatility of AI technology requires
teachers to make an even wider range of and more professional pedagogical deci-
sions. This was demonstrated by three teacher roles that emerged during teacher
interviews (i.e., orchestrating different resources with quality pedagogical decisions,
making students active investigators, and raising ethical awareness regarding the use
of AI). It is of significant note that these three roles underscore the importance of
teacher-student interaction in understanding the teacher-AI relationship (Choi et al.,
2023a; Luckin et al., 2022; Kim et al., 2022; Xu & Ouyang, 2022), as Teacher 9
stated that teachers’ pedagogical decisions regarding ChatGPT should be made
based on the rapport between the teacher and students, individual students’ needs
and characteristics, and the collective dynamic of the classroom.

6 Limitations

This research has some limitations to be considered in future studies. First, the
participants were selected on the basis of their interest in and willingness to use
ChatGPT. This indicates that they might have been more technology-friendly than
the average among teachers. Also, the sample in this study comprised only a small

13
Education and Information Technologies (2023) 28:15873–15892 15889

number of English teachers working at elementary schools located in one country;

therefore, the findings may not be broadly generalizable. Also, we examined partici-
pant teachers as one group without consideration of their differences. Future inves-
tigations should involve teachers representing a diverse range of subjects, school
levels, countries, and attitudes toward technology for a more comprehensive under-
standing of teachers’ uses of LLM-powered chatbots to support further theorizing.
Next, this study did not address the experiences and perceptions of students. An
important contrast to this exploration will be an examination of the use of Chat-
GPT from the student perspective, such as a qualitative analysis of students’ percep-
tions or a quantitative examination of the chatbot’s effect on students’ performance
or motivation. This would provide an informative and interesting complement to
this study. Finally, given the ubiquity and prevalence of new technologies, it would
be relevant to investigate the impact of teachers’ previous experiences with AI and
their effect on teachers’ acceptance or use of LLM-powered chatbots as educational
tools (Choi et al., 2023b).

7 Implications and conclusion

Important implications for the use of AI in education can be drawn from this study,
in particular, the value of using an LLM-powered chatbot as a multifunctional tool.
First, this study is among the initial attempts to empirically examine the educa-
tional potential of the LLM technology, ChatGPT. To be specific, expanding on
theoretical explorations with empirical evidence, this research found that the chat-
bot serves different instructional roles, including those of interlocutor, content pro-
vider, teaching assistant, and evaluator. Second, this study provided concrete exam-
ples of teacher prompts used to elicit useful responses. Teachers, teacher educators,
and researchers can use this finding to initiate further exploration into the potential
of LLM-powered chatbots and to develop different application methods based on
their own educational contexts. Third, the study offered insight into the dynamicity
of the human-AI relationship that changes in accordance with the rapid advance of
AI technology. In the teacher interviews, it was indicated that the roles of teachers
are predicted to increasingly become both multifaceted and specialized as AI tools
continue to evolve more. Thus, this research provides an initial reference for fur-
ther exploration into how this complementary relationship will unfold in the future,
as technology develops even more and across different local educational contexts.
Last, confirming the idea that teaching with AI requires specific teacher compe-
tencies (e.g., Celik, 2023; Choi et al., 2023a), this study shows a need for teacher
training courses specifically designed for teaching with an LLM-powered chatbot.
After the use of ChatGPT for a period of two weeks, the teachers in this study con-
curred that obtaining quality output from the chatbot depends on the teacher’s abil-
ity to develop quality questions. In this regard, the need for preservice education
and professional development programs to help teachers align the resources made
available by technology with their pedagogical purposes will be another important
issue (Jeon et al., 2022).

13
15890 Education and Information Technologies (2023) 28:15873–15892

To conclude, this research took a crucial step forward to understand the pedagog-
ical potential of LLM technology. Further investigation of this emerging technology,
including different types of LLM technology, is necessary to determine how it can
be incorporated into education as a tool for teachers, as well as students. This may
require an in-depth understanding of the AI-human relationship that comes from a
comprehensive knowledge of teachers’ experiences with and perceptions of the ever-
more sophisticated technologies that will inevitably become integrated into their
profession.

Data availability The datasets generated during and/or analyzed during the current study are available
from the corresponding author on reasonable request.

Declarations
Conflicts of interest None

References
Altman, S. (2022, December 5). Twitter. https://​twitt​er.​com/​sama/​status/​15996​68808​28502​8353?s=​
20&t=​j5ymf​1tUeT​peQuJ​KlWAK​aQ
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.
Bibauw, S., François, T., Van den Noortgate, W., & Desmet, P. (2022). Dialogue systems for language
learning: A meta-analysis. Language Learning & Technology, 26(1), 1–24.
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 Psy-
chology, 3(2), 77–101.
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam,
P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R.,
Ramesh, A., Ziegler, D. M., Wu, Jeffrey, Winter, C. …, & Amodei, D. (2020). Language Models
are Few-Shot Learners. arXiv. https://​arxiv.​org/​abs/​2005.​14165
Celik, I. (2023). Towards Intelligent-TPACK: An empirical study on teachers’ professional knowledge
to ethically integrate artificial intelligence (AI)-based tools into education. Computers in Human
Behavior, 138, 107468. https://​doi.​org/​10.​1016/j.​chb.​2022.​107468
Celik, I., Dindar, M., Muukkonen, H., & Jarvela, S. (2022). The promises and challenges of artificial
intelligence for teachers: A systematic review of research. TechTrends, 66, 616–630. https://​doi.​
org/​10.​1007/​s11528-​022-​00715-y
Choi, S., Jang, Y., & Kim, H. (2023a). Influence of Pedagogical Beliefs and Perceived Trust on Teach-
ers’ Acceptance of Educational Artificial Intelligence Tools. International Journal of Human–
Computer Interaction, 39(4), 910–922. https://​doi.​org/​10.​1080/​10447​318.​2022.​20491​45
Choi, S., Jang, Y., & Kim, H. (2023b). Exploring factors influencing students’ intention to use intelli-
gent personal assistants for learning. Interactive Learning Environments. https://​doi.​org/​10.​1080/​
10494​820.​2023.​21949​27
Creswell, J. W. (2008). Educational research: Planning, conducting, and evaluating quantitative and
qualitative approaches to research (3rd ed.). Merrill/Pearson Education.
Dillenbourg, P., & Jermann, P. (2010). Technology for classroom orchestration. In M. S. Khine & I.
M. Saleh (Eds.), New science of learning (pp. 525–552). Springer Science + Business Media.
Dizon, G. (2020). Evaluating intelligent personal assistants for L2 listening and speaking develop-
ment. Language Learning & Technology, 24(1), 16–26.

13
Education and Information Technologies (2023) 28:15873–15892 15891

Fryer, L., Coniam, D., Carpenter, R., & Lăpușneanu, D. (2020). Bots for language learning now: Cur-
rent and future directions. Language Learning & Technology, 24(2), 8–22.
Fryer, L. K., Nakao, K., & Thompson, A. (2019). Chatbot learning partners: Connecting learning
experiences, interest and competence. Computers in Human Behavior, 93(December 2018), 279–
289. https://​doi.​org/​10.​1016/j.​chb.​2018.​12.​023
Heidt, A. (2023). ‘Arms race with automation’: Professors fret about AI-generated coursework.
Nature. https://​www.​nature.​com/​artic​les/​d41586-​023-​00204-z
Hew, K. F., Huang, W., Du, J., & Jia, C. (2023). Using chatbots to support student goal setting and
social presence in fully online activities: Learner engagement and perceptions. Journal of Com-
puting in Higher Education, 35, 40-68. https://​doi.​org/​10.​1007/​s12528-​022-​09338-x.
Holstein, K., Aleven, V., & Rummel, N. (2020). A Conceptual Framework for Human–AI Hybrid
Adaptivity in Education. In: I. Bittencourt, M., Cukurova, K., Muldner, R., Luckin, & E., Millán
(Eds.), Artificial Intelligence in Education. AIED 2020. Lecture Notes in Computer Science (vol.
12163). Springer, Cham. https://​doi.​org/​10.​1007/​978-3-​030-​52237-7_​20
Holstein, K., & Aleven, V. (2022). Designing for human–AI complementarity in K-12 education. AI
Magazine, 43(2), 239–248. https://​doi.​org/​10.​1002/​aaai.​12058
Holstein, K., McLaren, B. M., & Aleven, V. (2019). Co-designing a real-time classroom orchestration
tool to support teacher–ai complementarity. Journal of Learning Analytics, 6(2), 27–52. https://​
doi.​org/​10.​18608/​jla.​2019.​62.3
Huang, X., Zou D., Cheng, G., Chen, X., & Xie, H. (2023). Trends, research issues and applica-
tions of artificial intelligence in language education. Educational Society & Technology 26(1),
112–131. https://​www.j-​ets.​net/​colle​ction/​forth​coming-​artic​les/​26_1
Huang, W., Hew, K. F., & Fryer, L. K. (2022). Chatbots for language learning—Are they really use-
ful? A systematic review of chatbot-supported language learning. Journal of Computer Assisted
Learning, 38, 237–257. https://​doi.​org/​10.​1111/​jcal.​12610
Hwang, G. J., Xie, H., Wah, B. W., & Gasevic, D. (2020). Vision, challenges, roles, and research
issues of Artificial Intelligence in Education. Computers and Education: Artificial Intelligence,
1, 100001. https://​doi.​org/​10.​1016/j.​caeai.​2020.​100001
Jeon, J. (2021). Chatbot-assisted dynamic assessment (CA-DA) for L2 vocabulary learning and diag-
nosis. Computer Assisted Language Learning. https://​doi.​org/​10.​1080/​09588​221.​2021.​19872​72
Jeon, J. (2022a). Exploring a self-directed interactive app for informal EFL learning: a self-determina-
tion theory perspective. Education and Information Technologies, 27(4), 5767–5787. https://​doi.​
org/​10.​1007/​s10639-​021-​10839-y
Jeon, J. (2022b). Exploring AI chatbot affordances in the EFL classroom: young learners’ experi-
ences and perspectives. Computer Assisted Language Learning. https://​doi.​org/​10.​1080/​09588​
221.​2021.​20212​41
Jeon, J., Lee, S., & Choe, H. (2022). Enhancing EFL pre-service teachers’ affordance noticing and
utilizing with the Synthesis of Qualitative Evidence strategies: An exploratory study of a cus-
tomizable virtual environment platform. Computers & Education, 190, 104620. https://​doi.​org/​
10.​1016/j.​compe​du.​2022.​104620
Ji, H., Han, I., & Ko, Y. (2023). A systematic review of conversational AI in language education:
focusing on the collaboration with human teachers. Journal of Research on Technology in
Education, 55(1), 48–63. https://​doi.​org/​10.​1080/​15391​523.​2022.​21428​73. Advance online
publication.
Kasneci, E., Sessler, K., Kuchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh,
G., Gunnemann, S., Hullermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Jurgen,
P., Poquet, O., Sailer, M., Schmidt, A., Seidel, T. …, & Kasneci, G. (2023). ChatGPT for good?
On opportunities and challenges of large language models for education. Learning and Individ-
ual Differences, 103, 102274. https://​doi.​org/​10.​1016/j.​lindif.​2023.​102274
Kim, J., Lee, H., & Cho, Y. H. (2022). Learning design to support student-AI collaboration: Per-
spectives of leading teachers for AI in education. Education and Information Technologies, 27,
6069–6104. https://​doi.​org/​10.​1007/​s10639-​021-​10831-6
Kuhail, M. A., Alturki, N., Alramlawi, S., & Alhejori, K. (2022). Interacting with educational chat-
bots: A systematic review. Education and Information Technologies. https://​doi.​org/​10.​1007/​
s10639-​022-​11177-3. Advanced online publication.
Lee, S., & Jeon, J. (2022). Visualizing a disembodied agent: Young EFL learners’ perceptions of
voice-controlled conversational agents as language partners. Computer Assisted Language
Learning. https://​doi.​org/​10.​1080/​09588​221.​2022.​20671​82

13
15892 Education and Information Technologies (2023) 28:15873–15892

Lombard, M., Snyder-Duch, J., & Bracken, C. C. (2002). Content analysis in mass communication:
Assessment and reporting of intercoder reliability. Human Communication Research, 28(4),
587–604. https://​doi.​org/​10.​1111/j.​1468-​2958.​2002.​tb008​26.x
Luckin, R., Cukurova, M., Kent, C., & du Boulay, B. (2022). Empowering educators to be AI-ready.
Computers and Education: Artificial Intelligence, 3, 100076. https://​doi.​org/​10.​1016/j.​caeai.​
2022.​100076
OpenAI. (2023). ChatGPT. OpenAI. https://​chat.​openai.​com/​chat
Stokel-Walker, C. (2023). AI Bot ChatGPT writes smart essays--should professors worry?
Nature. https://​doi.​org/​10.​1038/​d41586-​022-​04397-7. https://​www.​nature.​com/​artic​les/​
d41586-​022-​04397-7
Susnjak, T. (2022). ChatGPT: The end of online exam integrity? arXiv. https://​doi.​org/​10.​48550/​arXiv.​
2212.​09292
Timpe-Laughlin, V., Sydorenko, T., & Daurio, P. (2022) Using spoken dialogue technology for L2 speak-
ing practice: what do teachers think? Computer Assisted Language Learning, 35(5–6), 1194–1217.
https://​doi.​org/​10.​1080/​09588​221.​2020.​17749​04
Wang, X., Liu, Q., Pang, H., Tan, S. C., Lei, J., Wallace, M. P., & Li, L. (2023). What matters in AI-sup-
ported learning: A study of human-AI interactions in language learning using cluster analysis and
epistemic network analysis. Computers & Education, 194, 104703. https://​doi.​org/​10.​1016/j.​compe​
du.​2022.​104703
Weizenbaum, J. (1966). ELIZA—A computer program for the study of natural language communication
between man and machine. Communications of the ACM, 9(1), 36–45.
Xu, W., & Ouyang, F. A. (2022). A systematic review of AI role in the educational system based on a
proposed conceptual framework. Education and Information Technologies, 27, 4195–4223. https://​
doi.​org/​10.​1007/​s10639-​021-​10774-y
Zawacki-Richter, O., Marin, V. I., Bond, M., & Gouverneur, F. (2019). Systematic review of research
on artificial intelligence applications in higher education–where are the educators? International
Journal of Educational Technology in Higher Education, 16(1), 39. https://​doi.​org/​10.​1186/​
s41239-​019-​0171-0
Zhai, X. (2022). ChatGPT User Experience: Implications for Education. arXiv. https://​doi.​org/​10.​2139/​
ssrn.​43124​18
Zhang, Z., Xu, Y., Wang, Y., Yao, B., Ritchie, D., Wu, T., Mo, Y.,Wang, D., & Li, T. J. J. (2022). Sto-
ryBuddy: A human-AIcollaborativechatbot for parent–childinteractive storytellingwith flexible
parental involvement. In CHI ’22: CHI conferenceon human factors in computing systems (pp.
1–21).https://​doi.​org/​10.​1145/​34911​02.​35174​79

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under
a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted
manuscript version of this article is solely governed by the terms of such publishing agreement and
applicable law.

Authors and Affiliations

Jaeho Jeon1 · Seongyong Lee2

Jaeho Jeon
jaehojeon21@gmail.com
1
Department of Literacy, Culture, and Language Education, Indiana University, 107 S. Indiana
Avenue, Bloomington, IN 47405‑7000, USA
2
Department of English Education, Hannam University, 70 Hannam‑Ro, Daedeok‑Gu,
Daejeon 34430, Republic of Korea

13