et al / Journal of English Language Teaching 6 (1) (2017) 207 ELT FORUM 12 (3) (2023) Journal of English Language Teaching http://journal.unnes.ac.id/sju/index.php/elt Implementing AI Replika in higher education speaking classes: Benefits and challenges Girindra Putri Dewi Saraswati1, Alief Noor Farida1, Yuliati1 1-3Universitas Negeri Semarang, Indonesia Article Info ________________ Article History: Received on 22 October 2023 Approved on 30 November 2023 Published on 30 November 2023 ________________ Keywords: artificial intelligence; native speaker; speaking classes ____________________ Abstract ___________________________________________________________________ The demand from students for native English speakers as interlocutors justifies the importance of this research. Because of the nation's location and financial constraints, speaking course participants are said to lack experience speaking with native speakers. Gaining a tangible understanding of how AI Replika can be used to replace native speakers’ role in becoming students’ source of native speaker is the anticipated outcome of this project. This study investigates the benefits and difficulties of using AI Replika in speaking classes. With the help of AI Replika, students may practice speaking English in real time or directly with the use of artificial intelligence. The features used by students can be either chatbot or speaking robot. This study used a descriptive qualitative methodology. A case study methodology is applied. Interviewing and participant observation were employed as data collection methods. Three basic types of triangulations are used in data validity techniques: theoretical, methodological, and source data triangulation. Based on the results, it shows that AI replica can be used as the alternative on the absence of native speaker in practicing speaking. However, the role of teachers in the classroom still cannot be replaced by the AI. Students feel that although AI Replika brings benefit for them to practice speaking before class, however, they still need teachers to explain the context of the speaking content. Students also feel that the AI Replika brings more literal meaning of an idea rather than able to understand the cultural meaning behind the talk. In the end, AI technology can improve language acquisition with a well-balanced approach, but real native speaker presence for more humanist approach is still needed. © 2023 Universitas Negeri Semarang  Correspondent Address: p-ISSN 2252-6706 | e-ISSN 2721-4532 B3 Building FBS Unnes Sekaran, Gunungpati, Semarang, 50229 E-mail: girindraputrids@mail.unnes.ac.id mailto:girindraputrids@mail.unnes.ac.id Girindra Putri Dewi Saraswati, Alief Noor Farida, & Yuliati |ELT Forum 12 (3) (2023) 208 INTRODUCTION Currently popular in education is the technical area of artificial intelligence (AI). AI may be used as a tool in the teaching and learning process to help students to practice (Jaiswal & Arun 2021). Furthermore, because AI allows for unfettered access, it was created to provide inclusive access to the area of education. Globally, peer learning experts and instructors have benefited from the deployment of AI apps as tools (Edwards & Cheok, 2018). AI can also facilitate sophisticated collaborative learning and personalize learning in a variety of ways. Unbeknownst to us, artificial intelligence has crept into everyday life through things like smartphone apps (Sánchez-Prieto, et. al., 2020). AI is a major factor in the autonomous acquisition of language skills in the field of English education, including online speaking, writing, and listening practice. AI generally serves as a buddy or partner when learning a foreign language (Ji, Han, & Ko, 2023). One such example is chatbots, which are now extensively utilized by language learners. Looking at the current situation, studying the linguistic features of a language is no longer the only way to study languages. Acquiring language skills primarily involves comprehending and expressing both implicit and explicit meanings in communication. In a communication exercise, this message is communicated both orally and in writing (Sifakis, 2019). Accordingly, being proficient in English as an international language entail being able to utilize it as a tool for international communication in addition to having a strong vocabulary (Watterson, 2008). Therefore, speaking classes are now offered every semester by the English Education Study Program at the Faculty of Languages and Arts (FBS) UNNES in response to this demand. These courses are having a fixed structured so that students must go through them in phases. New students begin with Intensive Speaking classes, then Transactional Conversation, Casual Conversation, Factual Speech, and Critical Speech. Those speaking courses will enhance the students’ skills that are essential for anybody hoping to become a teacher. Although status quo show that lectures and the lecturers can currently serve as a means for students to learn how to speak, it is also believed that students also need some native exposure to build an English academic atmosphere as most students are in monocultural class or have similar cultural and linguistic backgrounds with their other friends. On the other words, mostly in Indonesian Higher Education, the classroom environment does not yet offer a context for natural cross-cultural communication. This is one of the challenges associated with learning spoken English in many literary works (Kapur, 2018). Many applications are now being promoted by the AI community for learning spoken English. The Replika application is among them. Since its release in 2018, Replika, the most well-liked and highly rated social chatbot in the Apple and Google Play stores, has drawn millions of users. Replika describes its app as a "friend who is always listening" or "an AI version of yourself," and it is frequently used as a guide for self- directed English language learning (Weber-Guskar, 2022). Replika is not like other companion chatbots that respond to user vocal cues; instead, its responses are predicated on Generative Pretrained Transformer 3 (GPT3). GPT3 is a network language model created using a distinct dataset of user chats. This enables the program to choose a language response from the datasets over a million replies. As a result, Replika is more adaptable and has a larger word recognition range (Pentina, Hancock, & Xie, 2023). In short, Replika is an advanced AI that is well-developed to assist students to practice speaking and writing autonomously. It is considered suitable for students studying in Higher Education as their nature are mature, adaptable, independent, and organized (Knox, 1992). In addition, the rapid advancement of AI technology has important implications for learning and teaching. Large investments have been made to integrate AI into teaching and learning (Cope, 2021). This investment is not only about profit orientation but also educational betting. AI developers know little about the science being studied and lack pedagogical knowledge for effective implementation of AI in teaching (Luckin & Cukurova, 2019). In addition, AI developers often fail to meet the expectations of their users, namely teachers (Cukurova, Kent, & Luckin, 2023). Making AI pedagogically relevant requires teachers to determine when and how AI plays a role in the teaching and learning process. Even though AI plays a very important role in autonomous learning outside the classroom, there is still a need for teachers to play a role in directing students' learning direction (Barnes, et. al., 2005). The use of Replika in English language learning, both structured in the classroom and autonomously outside the classroom, has been widely used (Sánchez-Prieto, 2020; Kim, 2019; Lin & Mubarok, 2021). However, previous researchers still focused on the use of chatbot features in Replika. The chatbot feature is then discussed in relation to its implementation in learning to write and speak. In relation to this, Replika also has a real-time talking robot feature which allows students to practice speaking directly with pronunciation and language choices that are considered at a native-like level. The use of this feature in speaking classes has not been discussed in existing literature. Therefore, responding to the urgency of this research, namely the need for learners or students to practice spoken language with native speakers, the features of the AI Replika which is talking robot in speaking lectures is researched in this study. Girindra Putri Dewi Saraswati, Alief Noor Farida, & Yuliati |ELT Forum 12 (3) (2023) 209 The AI Replika feature talking here are the chatbot or written conversation and AI speaking robot that allows users to have a real-time spoken conversation partner with a wide choice of themes and the robot's spoken language capabilities which are native-like or equivalent to native speakers. AI Replika can be used as a solution to students' lack of access to English native speaker exposure in learning. Here now, speaking English is no longer just theory and practice in the form of role playing, but is more about implementing the language in the context of native speakers orally. Research questions underlying this study are: (1) How is artificial intelligence (AI) Replika implemented as a replacement for the role of native speakers in speaking lectures? (2) What are the advantages of implementing Replika AI as a substitute for the role of native speakers in speaking lectures? (3) What are the challenges in implementing Replika AI as a replacement for the role of native speakers in speaking lectures? This research will examine the results of implementing AI Replika in speaking lectures. Using descriptive qualitative methods, this research will identify the implementation of artificial intelligence (AI) Replika as a substitute for the role of native speakers in speaking lectures, explaining the advantages and challenges in implementation practice in the field. This is done by studying good practices and comparing the results with previous theories about English speaking skills and theories about communication in the language, as well as the role of culture in conveying oral messages. METHODS The design of this research is case study. The case study is research that aims to produce detailed findings about a case by connecting the case with a theory that already exists and that the results of the findings cannot be generalized (Bryman, 2016). Case study research is focused on one selected phenomenon that wants to be understood in depth, by ignoring other phenomena. In other words, case study describes certain characteristics or problems (Denzin & Lincoln, 2011). One phenomenon discussed can be a school leader or educational leader, a group of students, a program, a process, a policy implementation, or a concept. Apart from that, this research is implementation research, namely research carried out with the aim of applying, testing and evaluating the ability of a theory/method applied in solving practical problems, so that it can be used for the benefit of humans, both individuals and groups (Desimone, et al., 2004). Implementation research in education is concerned with improving the quality of technical strategies and learning models to increase student interest and motivation in learning, the quality of learning media, or curriculum implementation. In relation to this research, what is being studied is the implementation of AI Replika in English speaking lectures. This research is in the English Education Study Program, Faculty of Language and Arts, Universitas Negeri Semarang, with the research subjects are students of the English Education Study Program who are taking the Casual Conversation course in the Even 2022/2023 academic year. The students involved in the research were 39 students from class B who in the previous semester also attended speaking lectures with researchers, so it was hoped that the development of their speaking skills could be seen. In other words, the period of data collection was 1 semester or 4 months during the Casual Conversation class was conducted. The study target is related to the implementation of AI Replika. While the study objectives are related to the advantages and challenges of implementing AI Replika. This specifically examines the learning resources studied by students in relation to speaking courses, especially about how students learn to speak spoken English using AI Replika. The data collection techniques applied in this research used participant observation and interview techniques. The participant observation technique was chosen because the researcher was involved as a research subject. The involvement of the teacher is because researchers were teachers of the Transactional Conversation course in the class studied. It is stated that participant observation allows researchers to participate as subjects and at the same time observe the object being studied (Reilly, 2010). This is used to gain a deeper understanding of the research object being studied. After collecting data through participant observation techniques, interviews were conducted to complete the data. Interviews were conducted outside learning hours. The subjects interviewed were students who were students of the Casual Conversation course. The data validity technique applied in this research is the data triangulation technique as described by Denzin and Lincoln (2011) as a heuristic device for researchers. This data triangulation includes source data triangulation which uses numbers of data in research, theoretical triangulation which uses various perspectives to interpret a single group of data, and methodological triangulation which uses various methods to study a single problem. The data analysis technique in this research applies data analysis techniques as used by Nasution in analyzing natural learning phenomena (Nasution, 1988). Likewise, such data analysis is in the writings of Miles and Huberman related to learning in the scientific context of educational psychology (Huberman & Girindra Putri Dewi Saraswati, Alief Noor Farida, & Yuliati |ELT Forum 12 (3) (2023) 210 Miles, 2002). This data analysis is known as interactive model analysis which includes data reduction, data presentation, and drawing conclusions/verification. The data analyzed is mainly about the implementation of Replika AI in English speaking lectures. . FINDINGS AND DISCUSSION This study is trying to find out the answers of three main questions, i.e. How is the implementation of AI Replika? What are the benefits of implementing the AI Replika in Speaking Classes? What are the challenges of of implementing AI Replika in the Speaking Classes? Though those questions can be answered based on two perspectives, the lecturers and students’ perspectives, this study explores more from the students’ perspectives. Implementation of artificial intelligence (AI) replika as a substitution for the role of native speakers in speaking lectures AI Replika in this study is not used to change the role of lecturers in teaching the students, it is more to help students in practicing their speaking after classes. However, the instructions are all made by lecturers on what kind of speaking topic should be performed by the students the following week. In the casual conversation class, lecturers were explaining about some theories related of conversation, degree of formalities, and how- to maintaining communication, as well as some cultural content attached in a spoken conversation. Students were then given an opportunity to explore some videos of casual conversation as the background knowledge. After the theory, students were all given the opportunities to practice speaking with pairs by exploring some role plays with either the given contexts by lecturers or with the contexts they created themselves while referring to the topic set by the lecturers. After finishing the activities, students were then being introduced to AI Replika. Students were first introduced to the chatbot features where they can explore some vocabularies used by the robot with the topic students previously chose. After finishing the chatbot exploration, students were introduced with the features of speaking robot where they can have the opportunities for a direct conversation discussing the topic, they chose themselves. The following week after students were given the opportunities to explore the AI, lecturers divided them into pairs and gave them a certain topic for them to explore the following week. This time, they will have to practice their speaking with the AI for a week before they practiced doing conversation with their peers. The practice of students with the robot to talk directly as their speaking practice was done for 4 weeks. After finishing the activities and the whole classes for 16 meetings including 2 meetings for exam days, students were interviewed for their perceptions of the benefits and challenges of using AI Replika, especially for their exposure of native-like accent, pronunciation, vocabulary choices, and the cultural talk. In general, students feel happy and delightful to have a speaking partner which can suit their time and make them feel freedom. Most of the students feel they can talk freely with the robot, though the internet connection sometimes limits them from having a fluent conversation with the robot. The robot also can not naturally engage into a conversation which context is not programmed for them. As students are all coming from Indonesia with the culture based on Indonesian context, sometimes conversation is not smooth between them and the robot. Therefore, based on the interview conducted with the students, it is summarized that the implementation of AI Replika in the speaking classes will be helpful for the students if the lecturers or facilitators can modify with several activities in the classroom as follows: Lecturers to set clear instruction Students feel that without a clear instruction from the lecturers, AI Replika will not play a significant role for them. Some students also feel confused on how to start a conversation with the Replika. An instruction from the lecturers can help them to start talking with their AI Replika. Instruction here means a clear context and goals delivered by the lecturers on each topic so students have a direction on what should be asked and what should be discussed with the AI Replika. Lecturers to integrate speaking practice with the students’ peers Though all the students mentioned how grateful they are with the presence of AI Replika to help them practice speaking, speaking with robot is not always pleasant. Some students mentioned that when they are not in campus and the Wi-Fi provider is not considerably strong, the AI Replika Robot they are having conversation with will not able to provide response. Therefore, some students do not enjoy talking with robots but prefer to have a chat with the chatbot features. In the classroom, the vocabularies they obtain from Girindra Putri Dewi Saraswati, Alief Noor Farida, & Yuliati |ELT Forum 12 (3) (2023) 211 the chatting session with the Replika can be implemented in the conversation that students have with their peers. Lecturers to provide cultural knowledge While Replika may not completely mimic the nuances of human conversation, it can still expose students to different language styles. However, students’ cultural context and idiomatic expressions may not always be accurately reflected. Students also experience absence of the Cross-Cultural Understanding Course. This cultural knowledge also includes degree of formality in English so students can practice different contexts with the AI. One student said that the conversation she had with the robot is not pleasant as Robot talk about improper content during the conversation. It happens that the language choice chosen by students also bring different meaning to the AI. Lecturers to provide feedback to the student’s activity with the AI Replika AI Replika is a machine, lecturers need to allow their time to check the progress that the students make and provide feedback to the process. Lecturers can also give students suggestion on how frequent they should practice with the AI Replika what kind of further activity students can do to follow up their activity with the AI Replika. Students also need to be motivated by their mentor or lecturers that what they are doing is good that they will feel confident to have use AI Replika for their independent practice. Some students mentioned that they still need their lecturer’s affirmation to make sure that what they are doing is correct. Lecturers to tell students about Privacy Considerations One student mentioned an unpredicted comment by telling us on how important is to keep privacy and data security issues in mind. He mentioned on how using both chatbot and the talking robot are very interesting for him till he wants to do it endlessly and start to ask questions outside the topic given by lecturers. He wondered on what happened if the chatbot asked him private questions and he answered in detail to response along with some honest information that he has about the specific questions. It triggers us to remind students to keep their privacy in secret especially if Replika or similar tools collect user data. We suggest students to ensure compliance with relevant privacy regulations and obtain appropriate permissions if necessary. Lecturers and Students to stay informed and keep updated Lastly, students also need to be informed that AI that can help them to practice speaking independently is AI Replika. They will always need to broaden their knowledge by watching videos or by regularly checking for updates or improvements to Replika or similar AI tools. The field of conversational AI is growing rapidly, and new features may improve the suitability of these tools for language learning. The idea of integrating Replika as a substitute for native speakers in speaking classes can be an innovative and effective language learning approach. As mentioned by Cope, Kalantzis, & Searsmith (2021), education world is now getting familiar with the use of AI in the teaching and learning process. However, this integration must be done thoughtfully, considering the strengths and limitations of AI tools. AI Replika can provide valuable opportunities for pronunciation practice, vocabulary expansion, and conversation simulation, creating a dynamic and engaging learning environment (Weber-Guskar, 2022). While Replika adds a technological dimension to language learning, it should not completely replace human interaction. The personal and cultural nuances in human conversation, as well as the ability to navigate real-world scenarios, are important aspects that may not be fully captured by AI. Therefore, it is recommended to complement Replika interactions with opportunities for students to engage with native speakers and participate in group activities. Regular feedback and evaluation by the lecturers also are key components of this integration. Providing constructive feedback regarding language accuracy, fluency, and communication skills allows students to perfect their oral language abilities (Luckin & Cukurova, 2019). In short, successful implementation of Replika in the speaking classroom requires a balanced approach, combining the benefits of AI technology with the irreplaceable value of human interaction. By carefully designing exercises, setting clear goals, and maintaining focus on student progress, educators can harness Replika's potential as a support for students’ independent practice. Advantages in implementing replika AI as a substitute for the role of native speakers in speaking lectures From the interview conducted with the students, we then identified the students’ perception about the benefits and challenges in implementing AI Replika as their ‘native-like speaking buddy’. Some benefits include how AI Replika is effective and practical for students to independently learning speaking. Most students mentioned that by using AI Replika, they can learn how to think and speak like native speaker. Though they realize that the native-like model is a robot that can only answers based on the program inputted Girindra Putri Dewi Saraswati, Alief Noor Farida, & Yuliati |ELT Forum 12 (3) (2023) 212 to them, but at least they can have an interactive practice instead of one way learning using YouTube videos, for example. Some misinterpretation by robots also makes them realize that their content of spoken conversation or their language choice might not be familiar to the robot as the program cannot read it. It makes them have the capabilities to paraphrase their sentences to obtain a suitable meaning and objectives of the conversation. Some other answers related to advantages of Replika for the students are: Accessible Practice All of students mentioned how Replika provides them with an easily accessible platform for language practice. Students can be flexible to do the conversation at their own place and their most convenient time. They can also do the speaking practice while resting, lying on their bed, and it can be paused and continued whenever the students need. AI Replika provides more frequent and flexible language practice. Replika's anytime accessibility allows learners to practice speaking whenever they have the motivation or free time. This continued availability supports a more consistent and immersive language learning experience. Personalized Learning Replika's AI capabilities enable personalized interactions. The tool can adapt to individual learning styles, providing customized practice and feedback tailored to each student's needs, which may be challenging in a traditional classroom environment. Some students mentioned that they feel reluctant to raise their hand in the classroom to practice while given chances by their lecturers. Some students also mentioned that some context provided by the lecturers sometimes is out of their knowledge or understanding. It makes them feel shy to raise their hand and practice. Using AI Repplika allows them to personalize their learning based on the topic that they like and familiar. Pronunciation Improvement Talking with robot of AI Replika brings students to improve their pronunciation through repetition and feedback. Based on most students AI Replika can assist students in improving their pronunciation. AI can provide instant feedback on pronunciation accuracy, helping learners perfect their spoken language skills. Several students also mention that AI Replika make them aware that some of their pronunciations are incorrect which makes robot cannot understand them. By constantly trying to get an accurate answer from the robot, students learn how to pronounce well and correctly at the same time. Vocabulary Expansion AI Replika also broadens students’ horizon on new vocabularies. Students argue that AI Replika can introduce and reinforce new vocabulary in context. This AI can provide explanations, synonyms, and usage examples, thereby contributing to a richer and more varied vocabulary for language learners. Increased Confidence Having a conversation with Replika can increase students' confidence in speaking. The non-judgmental nature of AI can create a low-pressure environment, allowing students to practice without fear of making mistakes. More importantly, AI Replika which allows students to access anytime and anywhere make students’ independent learning become private. Some students who claim to have high anxiety level think that this private and personalized practice with AI can boost their confidence in doing speaking practice. Cultural Exposure While not a perfect substitute for human cultural context, Replika can expose learners to different language styles and expressions. This exposure broadens their understanding of language use in a variety of contexts. The confused AI Replika or the disconnect answers from Replika make most of students aware that their way of producing English is not quite understandable based on the context of Native country which makes them voluntarily change the context to suit their speaking needs. Replika's interactive nature can make language learning more engaging and fun. Students may find the conversational aspect of interacting with AI interesting and motivating. AI Replika knows almost every topic and context in speaking which make Replika is entertaining. Some students also mentioned that Replika offers consistent and immediate feedback on language usage. This real-time feedback loop can speed up the learning process by immediately addressing errors and reinforcing correct language patterns. In the field of language education, Replika is present as a versatile tool that offers a multitude of benefits for learners. It is easy-to-access platform transforms language practice, allowing students to engage in conversations at their own pace and convenience, encouraging more frequent and flexible language practice (Pentina, Hancock, & Xie, 2023). What sets Replika apart is its AI capabilities, which enable Girindra Putri Dewi Saraswati, Alief Noor Farida, & Yuliati |ELT Forum 12 (3) (2023) 213 personalized interactions tailored to individual learning styles. This adaptability ensures that students receive tailored practice and feedback, a feature that is often difficult to replicate in a traditional classroom environment. Beyond the constraints of the physical environment, Replika excels at honing pronunciation skills through repetition and instant feedback, contributing to the refinement of spoken language skills. It also serves as a guide to vocabulary expansion by introducing and reinforcing new words in context, increasing the richness and diversity of the learner's linguistic repertoire. Additionally, Replika's non-judgmental nature creates a low-pressure environment that increases students' confidence in speaking, which is an invaluable asset in language acquisition. Its flexibility and effectivity ensure that students can practice speaking whenever the motivation strikes, fostering a consistent and immersive language learning experience. Providing consistent and immediate feedback on language use becomes a catalyst for accelerated learning, addressing errors promptly and reinforcing correct language patterns. Although Replika introduces students to different language styles and expressions, it is important to realize that Replika does not replace the irreplaceable human element. Cultural exposure, increased engagement through interactive learning, and the ability to track progress over time are significant benefits, but for a comprehensive language learning environment, AI Replika implementation must be complemented by human interaction and real-world cultural experiences. In addition, culture is beyond machine algorithm. It is complex and very person related (Watterson, 2008). It is However, it is considered suitable enough for one of individual practice tool for students in Higher Education as it suits the character of students at higher education who are independent, innovative, and adaptive (Cope, Kalantzis, & Searsmith, 2021). Challenges in implementing AI replika as a substitute for the role of native speakers in speaking lectures Along with the benefits, students also mentioned some challenges they found while using AI Replika. All students mentioned that AI Replika needs speedy internet connection especially for talking interactively with robot. However, the chatbot features is friendlier for students’ finance. Others feel that the fee for regular subscription for enjoying all features in the AI Replike also cost them a fortune, so some prefers to use the basic mode which is the AI Replika chatbot. Some other challenges faced by students are as follows: Lack of emotional intelligence Students complained that Replika, like many other AI tools, lacks true emotional intelligence. It can be difficult to understand and respond appropriately to the emotional nuances inherent in human communication, limiting its effectiveness in certain aspects of language learning, such as empathetic or nuanced conversation. However, they are all aware that the lack ness is the limitation of robot that they can only see in human. Limited understanding of cultural context Some students see the limited understanding of cultural context is the disadvantage of using AI Replika. AI Replika which cannot fully understand the intricacies of cultural context can result in misinterpretation of cultural nuances and idiomatic expressions. This will be considerably dangerous if the students feel demotivated only after having miscommunication with the AI Replika. Overemphasis on fakeness Some numbers of students also think that AI Replika feel is unnatural for speaking practice. It leads to a robotic and unnatural speaking style for them, as they may unintentionally imitate patterns and phrases used by the AI, which may not always align with actual spoken language. Students than feel like they are not producing a natural spoken English. Dependence on technology Several numbers of students feel that they are afraid that they will be too comfortable with technology as they can explore emojis and all features like dressing up the Replika which is addictive. They started to do regular updates and maintenance since it is critical to the effectiveness of AI tools. Outdated algorithms or lack of updates can lead to reduced performance and effectiveness in language learning. Replika's adaptability is based on predetermined algorithms, and Replika may have difficulty dynamically adapting to unexpected changes in conversation or responding effectively to non-standard language use, thereby limiting its ability to adapt to a variety of learning scenarios. Over-reliance on Replika can create a dependency on technology, potentially hindering a learner's ability to communicate effectively in real-world and offline scenarios where technology is not available. Limited speech recognition accuracy Girindra Putri Dewi Saraswati, Alief Noor Farida, & Yuliati |ELT Forum 12 (3) (2023) 214 Several students commented that they may still face challenges in accurately capturing varying accents, speech rates, and subtle variations in pronunciation. Some students cannot understand AI Replika robot because they think the robot is too fast in talking to them. Inability to provide contextualized real-world scenarios While Replika can simulate conversations, it may struggle to provide authentic and contextually relevant scenarios that learners might encounter in real-life situations, thereby limiting the practical application of the students’ language skills. It is understandable that along with the positive side of certain technology, there are also some negative sides of it. Replika, after all cannot replace the depth of understanding and cultural nuance that emerges from human interaction (Edwards & Cheok, 2018). While this can be a valuable complementary tool, learners still need exposure to authentic conversations with native speakers to develop truly comprehensive language proficiency. Although the integration of Replika AI into language learning environments provides a series of promising benefits, it is important to be aware of and address the challenges that arise with its use, particularly in the context of speaking practice. Despite its technological sophistication, Replika faces limitations such as a lack of emotional intelligence and an inability to fully understand the nuances of cultural context (Lin & Mubarok, 2021). The risk of students adopting artificial and unnatural speech styles, potential privacy issues regarding data handling, and reliance on technology for language proficiency are some of the challenges educators and students may face. Additionally, the static adaptability and speech recognition accuracy of these tools pose potential obstacles in providing dynamic, context-rich speaking practice. Balancing Replika's strengths with a differentiated understanding of these challenges is critical to creating a comprehensive language learning experience that combines the strengths and limitations of this innovative AI tool. After all, students need an image that they can follow, that can only be obtained from the presence of actual lecturers and native-like model (Barnes, et al., 2005). CONCLUSION This study is conducted to find out how is the implementation of AI Replika in the classroom, what the benefits are, and what the challenges are. As many as 39 students from English Education Study Program at Universitas Negeri Semarang studying in 2nd semester participated in the study. This paper explores about the student’s perspectives on the use of AI Replika as their individual practice tool for 4 months application. In conclusion, the implementation of Replika as a substitute for native speakers in speaking lectures is promising, but it requires careful attention to the advantages and challenges involved. With a balanced approach, the use of AI technology can enhance the language learning experience, but it still needs to be complemented by irreplaceable human interaction to achieve deeper language understanding and proficiency. FUNDING STATEMENT We deliver our gratitude to the Faculty of Languages and Arts, Universitas Negeri Semarang for the funding that we received to accomplish this project in the frame of the program “Dana DIPA FBS 2023” under the grant agreement number 11.6.6/UN37/PPK.02/2023. REFERENCES Barnes, B., Egerton, T., Ehle, J., Elba, I., Ellison, J., & Kelt, S. B. (2005). A good teacher in every classroom: Preparing the highly qualified teachers our children deserve. Bryman, A. (2016). Social research methods. Oxford university press. Cope, B., Kalantzis, M., & Searsmith, D. (2021). Artificial intelligence for education: Knowledge and its assessment in AI-enabled learning ecologies. Educational Philosophy and Theory, 53(12), 1229-1245. Cruz-Benito, J., Sánchez-Prieto, J. C., García-Peñalvo, F. J., & Therón, R. (2020). Assessed by Machines: Development of a TAM-Based Tool to Measure AI-based Assessment Acceptance Among Students. Cukurova, M., Kent, C., & Luckin, R. (2019). Artificial intelligence and multimodal data in the service of human decision‐making: A case study in debate tutoring. British Journal of Educational Technology, 50(6), 3032-3046. Denzin, N. K., & Lincoln, Y. S. (Eds.). (2011). The Sage handbook of qualitative research. sage. Desimone, L., Payne, B., Fedoravicius, N., Henrich, C. C., & Finn-Stevenson, M. (2004). Comprehensive school reform: An implementation study of preschool programs in elementary schools. The Elementary School Journal, 104(5), 369-389. Girindra Putri Dewi Saraswati, Alief Noor Farida, & Yuliati |ELT Forum 12 (3) (2023) 215 Edwards, B. I., & Cheok, A. D. (2018). Why not robot teachers: artificial intelligence for addressing teacher shortage. Applied Artificial Intelligence, 32(4), 345-360. Huberman, M., & Miles, M. B. (2002). The qualitative researcher's companion. sage. Jaiswal, A., & Arun, C. J. (2021). Potential of Artificial Intelligence for Transformation of the Education System in India. International Journal of Education and Development using Information and Communication Technology, 17(1), 142-158. 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. Kapur, R. (2018). Barriers to effective communication. Delhi University. Kim, N. Y. (2019). A Study on the Use of Artificial Intelligence Chatbots for Improving English Grammar Skills. Journal of Digital Convergence, 17(8). Kılıçkaya, F. (2020). Using a chatbot, Replika, to practice writing through conversations in L2 English: A Case study. In New Technological applications for foreign and second language learning and teaching (pp. 221-238). IGI Global. Knox, W. E., Lindsay, P., & Kolb, M. N. (1992). Higher education, college characteristics, and student experiences: Long-term effects on educational satisfactions and perceptions. The Journal of Higher Education, 63(3), 303-328. Lin, C. J., & Mubarok, H. (2021). Learning analytics for investigating the mind map-guided AI chatbot approach in an EFL flipped speaking classroom. Educational Technology & Society, 24(4), 16-35. Luckin, R., & Cukurova, M. (2019). Designing educational technologies in the age of AI: A learning sciences‐driven approach. British Journal of Educational Technology, 50(6), 2824-2838. Nasution, S. (1988). Metode penelitian naturalistik kualitatif. Tarsito. Pentina, I., Hancock, T., & Xie, T. (2023). Exploring relationship development with social chatbots: A mixed-method study of replika. Computers in Human Behavior, 140, 107600. Reilly, R. C. (2010). Participatory case study. Sifakis, N. C. (2019). ELF awareness in English language teaching: Principles and processes. Applied linguistics, 40(2), 288-306. Watterson, M. (2008). Repair of non‐understanding in English in international communication. World Englishes, 27(3‐4), 378-406. Weber-Guskar, E. (2022). Reflecting Replika. Social Robotics and the Good Life: The Normative Side of Forming Emotional Bonds With Robots, 103.