2025 3rd IEEE International Conference on Industrial Electronics: Developments & Applications (ICIDeA) 979-8-3315-3320-5/25/$31.00 ©2025 IEEE Innovative AI Approaches in Oral Proficiency Testing: Ethical Implications of ASR Systems 1 A Periyasamy Department of English, Rajalakshmi Institute of Technology, Chennai, India Periyasamy.a@ritchennai.edu.in 4 M. Mythili Department of English Nandha Engineering College(Autonomous), Erode, India murugesanmythili@gmail.com 2 Geetha Manoharan, School of Business, SR University, Warangal, Telangana, India. geethamanoharan1988@gmail.com 5 Giftsy Dorcas E Department of English Kristu Jayanti College (Autonomous), Bengaluru, Karnataka, India giftsy@kristujayanti.com 3 Malik Bader Alazzam Faculty of Information Technology, Jadara University, Irbid, Jordan malikbader2@gmail.com 6 Prema S Department of English Panimalar Engineering College, Chennai, India. prema@panimalar.ac.in Abstract—Oral proficiency testing plays a critical role in language assessment; however, classic ASR faces such problems as Americans Bias, which means difficulty for ASR as application to recognize non-American accents, and the evaluation inaccuracy. The major problem here is that classic ASR systems cannot recognize and assess non-American accents as they would reward non-biased, unbiased, and fair scores for speakers with regional or non-native English accents. This bias in particular distorts the fairness of the oral proficiency test. The current approaches are inefficient with regards to the fairness and standardization in natural language scenarios. Unfortunately, in practice the conventional ASR models are not perfect and become a source of errors and biases when trying to assign a score for the given speech; the specific problem with the current approach is that the speakers with regional or non- native accent will be almost impossible to score correctly. To address this bias, this study proposes BERTAdv, incorporating BERT which focuses on contextual understanding into the Adversarial training method for enhancing the ASR’s accuracy in OPI. In an attempt to improve the BERTAdv to analyze both simple and complex linguistic attributes, the model is trained on a large and balanced dataset. The model presents desirable accuracy with a high R-squared of 0.93, MAE of 0.26, and RMSE of 0.46 and is more accurate and fairer than existing methods. Keywords— Oral Proficiency, ASR Systems, Ethical bias mitigation, BERTAdv I. INTRODUCTION Oral language proficiency as a strand in communication skills is central to the arrangements that are in place for readiness for personal, career and civic life [1]. There are professional talents used in views of communication skills, these include talking skills because they fall under the general communication skills that include speaking skills, and this comprises vocal caliber, accent, tone, course, effective speech and the mastery of meaningful dialogues [2]. Fluency, a prerogative of classroom language learning and testing, is an important component of language as it entails listening and comprehending and responding in parallel [3]. Assessment through receptive and productive skills has significant meaning in language acquisition and certification [4]. It appears in educational, professional, and social settings; therefore, it serves to emphasize why testing methods must be accurate, fair and accessible [5]. With the progress of AI technology, the conventional testing method is not immutable but can be improved; accordingly making oral proficiency assessment less subjective and comprehensive to accommodate the learners’ and institutions’ needs to master the language more efficiently [6]. AI and ASR technologies bring unbiased, standardized, and easily accessible solutions as the process of scoring is automated, logistics constrains removed and scalability is achieved [7]. This approach increases fairness, offers formative feedback, and can facilitate ongoing and mass language development assessment of learners from around the globe [8]. This research proposal seeks to explore how these innovations, specifically ASR, can revolutionize the manner in which oral proficiency testing is done, eliminate evaluator bias and make the process more accurate. Its goal is to increase and facilitate test taking and administration through remote testing while reducing reliance on face-to-face exams. For this reason, AI’s capability of providing unique feedback with backing from data enhances the learning process. Ethical issues like privacy, data security concerns and issues depending on the nature of training data fed to AI models are also examined in the study. The key contribution of the study is following, • This study analysing the ethical implications of ASR systems in oral proficiency testing, focusing on bias, fairness, and transparency. • This study Investigating the impact of accent and dialect recognition on ASR accuracy in language assessments • Evaluating the inclusivity of ASR systems for diverse language learners, including those with speech impairments. • BERTAdv mitigates ethical biases related to non-native accents and speech speed, ensuring a more equitable evaluation for diverse linguistic speakers. The research paper is organized in the following way: part III introduces the methodology, while section II provides access to the relevant literature. The fourth section presents the research’s outcomes, while Section V concludes the study. 2025 3rd IEEE International Conference on Industrial Electronics: Developments & Applications (ICIDeA) | 979-8-3315-3320-5/25/$31.00 ©2025 IEEE | DOI: 10.1109/ICIDEA64800.2025.10963053 Authorized licensed use limited to: Panimalar Engineering College - Chennai. Downloaded on September 26,2025 at 07:31:58 UTC from IEEE Xplore. Restrictions apply.