Evaluating accents: when students and AI systems “listen” – A conversation with Rosalba Nodari and Onur Özkaynak

27 February 2026, 15:00 (UTC+01:00)

Please register here: https://events.teams.microsoft.com/event/f85dd9b1-2c4f-4eef-b0ef-5f62512b680e@92513267-03e3-401a-80d4-c58ed6674e3b

Abstract

This webinar features Rosalba Nodari (University of Siena) and Onur Özkaynak (The Ohio State University) discussing issues related to accentism and linguistic legitimacy in language education. In the first talk, Nodari will report on a study exploring Italian secondary school students’ attitudes toward different English varieties (American, African American, British, Chinese, Indian, Italian, Multicultural London, Nigerian, and Ukrainian) within a World Englishes framework. Combining a variety identification task and a verbal-guise technique, 194 students rated speakers on competence, warmth, and speech traits. Results show generally low recognition rates, with Italian English being the most accurately identified, and recognition positively influencing evaluations on all dimensions. Accent effects emerged clearly: American and British English received the highest competence ratings, while Italian English, followed by American and African American English, was rated highest for warmth; Ukrainian English was evaluated least favourably overall. Familiarity with English accents significantly improved evaluations of competence, while appreciation of linguistic diversity positively influenced warmth. Gender effects were also observed, with male voices generally rated more positively, especially for Inner Circle varieties. Overall, the findings highlight the role of recognition, experience, and social framing in shaping evaluations of English varieties, with implications for more inclusive approaches to English language teaching.

In the second talk, Özkaynak will problematize issues of language evaluation from the lens of generative AI, which is increasingly positioned as an “objective” evaluator of language, offering feedback on pronunciation, fluency, and even professional communication. Against this backdrop, Özkaynak will present findings from a Gemini-based accent study that asks an urgent question: when an AI system listens to teachers’ English, does it evaluate intelligibility, or does it reproduce familiar hierarchies of accent and legitimacy? Using a verbal-guise design, a set of short, comparable recordings was prepared in which different speakers deliver brief ELT-style grammar explanation. Gemini was then prompted to evaluate the speakers’ instructional delivery (not the content) using a structured rubric and to provide written justifications for its ratings. Across speakers, Gemini’s evaluations reveal recurring patterns. It tends to treat certain accents as more “clear,” “professional,” and “teacher-like,” while positioning other global or non-dominant accents as less suitable, even when the speech is intelligible. Özkaynak will discuss how these outputs echo long-documented ideologies in TESOL (native-speakerism, standard language ideology, raciolinguistic legitimacy), reflecting on the risks of delegating “listening” and assessment to AI in teacher education, hiring, and classroom feedback practices.

Together, the two contributions illuminate how accent biases—whether held by students or embedded in AI—shape educational experiences and reinforce linguistic hierarchies. Importantly, they also point toward the need for more inclusive and critically informed language education practices.

This webinar will take place in English. Questions on the part of the audience will be welcome in English, Italian and Turkish. The webinar is part of the online series “Diversity Matters” hosted by the Center for Migration and Societal Change of Eurac Research. Taking an interdisciplinary approach, the webinar series explores the impact of migrations, diversities and mobilities on increasingly superdiverse territorial realities. The series is a forum for experts to share their work and expertise with an audience of fellow academics, students, decision-makers and practitioners.