Abstract

This article explores the transformative impact of artificial intelligence (AI) on English language teachers (ELTs). Through two authentic classroom experiences, the author compares the capabilities of XIPU AI, built on OpenAI's GPT model, and Google's Bard, shedding light on the opportunities and challenges presented by AI in the realm of education. The first narrative illustrates the expeditious class preparation facilitated by AI, emphasizing the need for collaborative features. The second narrative details an epistemological journey using AI-generated content and AI chatbots, highlighting the importance of human validation and critical thinking. Practical recommendations are offered for the enhancement of XIPU AI. In conclusion, the author advocates for collaborative engagement among educators to navigate the evolving landscape of AI in language education.

Keywords: English language teaching; Bard VS XIPU AI; collaborative features; AI-assisted research 

Introduction: an attitudinal dichotomy and ongoing debate on AI among the English teaching community  

Since the launch of XIPU AI in September 2023, ELTs at the English Language Center of Xi'an Jiaotong-Liverpool University (XJTLU) have embarked on an exploratory journey of novel possibilities introduced by AI. Built on an OpenAI GPT model, this formidable chatbot holds a compelling allure, as testimonials from educators suggest heightened productivity and the facilitation of self-directed learning initiatives among students.

However, this tempting prospect is met with a sense of trepidation and ambiguity among educators, particularly ELTs, who confront a lack of training and hold reservations about the evolving nature of their professional roles in the era of AI. As indicated in the recent report "Artificial Intelligence and English Language Teaching: Preparing for the Future" by the British Council (hereafter referred to as the Report), sentiments among ETLs towards AI are divergent, with some expressing apprehension about the transformative impact of AI while others exhibiting composure and optimism. Nevertheless, the integration of AI into educational frameworks is an inevitable reality that necessitates proactive engagement.

This article aims to present two narratives that elucidate my navigation through the transformed landscape of language teaching engendered by AI, drawing a comparative analysis between XIPU AI and Google's Bard. Concurrently, practical recommendations will be proffered to the XJTLU Learning Mall developer teams, contributing to the ongoing debate on AI's role in education.

Story One: Class Preparation Expedited with AI

Listening Class Enhancement

According to the Report, the prevalent way ELTs have experimented with AI is in material creation, a theme central to my first narrative. Specifically, it was a listening class where the audio content surpassed the proficiency level of my pre-intermediate students, or A2 level according to the CEFR (for readers outside the ELT community, CEFR is an international standard for describing language ability, with a six-point scale from A1, A2 for beginners, to B1, B2 for independent language users, to C1, C2 for proficient users). To mitigate potential student demotivation during the listening experience, I conceptualized a word-definition-synonym matching activity. Cards featuring A2+ words, accompanied by their definitions and synonyms, were to be laminated and randomly distributed in the class for students to find matching pairs.

Productivity Turbocharged by AI

Traditionally, the compilation of a challenging word list demands painstaking manual effort, involving a meticulous examination of transcripts and dictionaries. With AI, however, the identification of these words was accomplished within seconds of submitting the transcript along with prompts such as “Identify words beyond A2 level” and “Provide definitions and synonyms in a table” Both XIPU AI and Bard responded effectively, as depicted in Picture 1 and Picture 2. Notably, the GPT3.5 version on XIPU AI requires more rigorous human review compared to the more advanced GPT 4.0 version or Bard (See Picture 3).

Picture 1

Picture 2

Picture 3

Bard Takes the Lead and Recommendations for XIPU AI Developers

Despite satisfactory performance, notable differences emerged between XIPU AI and Bard. While both platforms accurately identified words, Bard excelled in enabling users to export responses directly to Google Sheets (Refer to the comparison between Picture 4 and Picture 5, 6), offering a seamless experience for users. The convenience of instantly converting AI responses into shareable Google Docs is not only innovative but also aligns with the collaborative nature increasingly crucial for educators in the AI era. This feature would not only enhance educators' efficiency but also empower students to easily share responses with their peers. This collaborative aspect becomes particularly relevant, given one of the key concerns highlighted by the Report—while AI enables personalized and self-paced learning, it poses a potential threat to students' collaborative abilities.

Therefore, I propose the incorporation of this functionality into XIPU AI, allowing responses to be downloadable and shareable in doc and excel formats, mirroring the capabilities of Bard. Potential platforms for sharing could include BOX, an online cloud space developed by XJTLU, or Google Docs. This enhancement would not only bridge the existing gap but also improve XIPU AI to the same level as its counterparts.

Picture 4

Picture 5

Picture 6

Story Two: Unhindered Epistemological Quest for Academic Excellence

Finding Support for a Persuasive Argument

The perennial question in writing classes, what constitutes a compelling argument, revolves around two primary themes: effectiveness and persuasiveness. Both are contingent on critical elements such as evidence and support. In a writing class centered on the preservation debate of historical buildings, I introduced AI-generated articles from Bard to ignite discussions among students—an illustration of leveraging AI for material creation.

As students engaged in constructing essays on the advantages and disadvantages of preserving historic structures, one student became captivated by a claim (See picture 7) suggesting that the preservation of such buildings could lead to economic disadvantages for urban village residents in Guangzhou, a southern city in China. However, this claim lacked explanation or support, creating a void that required filling for the argument to be effective and persuasive.

Picture 7: source article generated by Bard with language appropriate for A2 students

Bard Wins by a Landslide

In the pre-ChatGPT era, the endeavor to seek evidential support was an arduous and sometimes frustrating process involving formulating precise questions, inputting them into a search bar, sifting through outcomes, and often ending up with uncertain responses. With the advent of AI assistants, the search process has transformed to resemble one-stop shopping. The screenshots in pictures 8 and 9 illustrate the conversation between the student, under my guidance, and AI chatbots following his captivation by a particular argument. While all these platforms began with regret for their inability to provide a definitive answer, Bard's recommendations, as shown in the provided screenshots, gave a more constructive and meaningful approach. This served as a crucial first step in the ongoing quest for accuracy and specificity, which led to epistemological success.

Picture 8

Picture 9

An Epistemological Success that Requires Constant Fine-Tuning of Questions and Meticulous Verification

Bard’s contribution marked a commendable starting point for our inquiry. However, achieving a validated and citable answer was preceded by several failed attempts. Initially, our questions were too broad, such as "Give me research articles where I could understand the disparity between urban villagers and those outside the urban village in Guangzhou" and "Could you find the differences between urban villagers in Guangzhou and Guangzhou residents outside those urban villages in terms of their income levels, employment rates, and housing conditions." These queries resulted either in apologies or articles that did not match further search results, indicating Bard's fabrication of articles, now popularly known as hallucination.

Our approach encountered persistent roadblocks until we refined the question to focus on income disparity, specifically salary. Sequential prompts, including "What is the average salary of people living in urban villages in Guangzhou," "Could you also give me the growth rate of those urban villagers' salary," "Where did you get this information? Could you also provide the source," and "can you find the authors behind those reports," led us to a consequential research article titled “Understanding Urban Villages in China with Three Case Studies in Guangzhou”(Refer Picture 10, 11, 12). The numerical data on salary disparity matched and verified in that article enabled my student to construct an effective and persuasive argument. The power of AI left an indelible impression on the student, sparking motivation to engage in endeavors once perceived as challenging and time-consuming. Concluding our investigative journey, I prompted my student to engage in a reflective dialogue on the impact of AI on our understanding of knowledge, i.e., taking him on board with me for epistemological quests in the age of AI. Despite our shared sense of awe at AI’s remarkable capabilities, this epistemological journey with Bard, however, also underscored the reminder of blind reliance on AI-generated content, emphasizing the need for human validation, critical thinking, and skepticism.

Picture 10

Picture 11

Picture 12

Recommendations for XIPU AI Developers

To elevate the reliability and academic integrity of XIPU AI, it is recommended that the integration of a function allows the AI to access and extract information directly from authentic research papers. This enhancement would empower XIPU AI to provide more substantiated and academically sound responses, aligning with the rigorous standards expected within an academic institution such as XJTLU.

Furthermore, to foster transparency and facilitate thorough verification, it is advisable to incorporate a feature within XIPU AI that provides source links and, when applicable, original documents along with the responses. This inclusion not only empowers users to verify the information provided by the AI but also promotes a culture of academic rigor by encouraging users to engage with the primary materials directly. By implementing these enhancements, XIPU AI can emerge as a more reliable and transparent tool, catering to the research and academic needs of XJTLU students, faculty, and beyond.

Conclusion: Charting the Path Ahead Collaboratively

As English language teachers at Xi'an Jiaotong-Liverpool University and beyond are exploring this transformative journey shaped by AI, we must remain vigilant and informed about the evolving landscape of this technology. Working collaboratively with stakeholders, whether it be school policymakers, fellow teachers, or our students, is crucial. We stand at a crossroads, with potential futures ranging from a balanced integration of AI, and collaborative human-AI endeavors, to scenarios of AI dominance. To navigate this path responsibly, it is paramount to prioritize the best interests of our human stakeholders. Actively sharing our experiences and practices becomes instrumental in fostering a collective dialogue among English language teachers. Through this exchange of insights, we can collectively envision and shape a future where technology enhances, rather than overshadows the landscape of language education.

Reference:

Edmett, A., Ichaporia, N., Crompton, H., & Crichton, R. (2023). Artificial intelligence and English language teaching: Preparing for the future. British Council. https://doi.org/10.57884/78EA-3C69