Extraordinary Session

Artificial intelligence (AI) is rapidly transforming both scientific inquiry and its societal applications, reshaping how knowledge is generated, evaluated, and used. This presentation examines the role of AI in assessment as a case study of the broader opportunities and challenges AI presents for education and society. Focusing on the Duolingo English Test, a language proficiency assessment, it illustrates how advances in machine learning, natural language processing, and generative AI enable more adaptive, scalable, and accessible measurement of complex human skills. AI-driven systems support personalized testing pathways, innovative task designs that better reflect real-world language use, and continuous quality monitoring at a scale previously unattainable.

At the same time, the presentation emphasizes that scientific rigor and societal trust depend on maintaining strong theoretical foundations, transparent methodology, and robust validation practices. Human expertise and governance remain essential in defining constructs, interpreting evidence, and ensuring fairness, accountability, and ethical use. Situating AI-powered assessment within the wider landscape of AI in education, the presentation highlights the need for interdisciplinary collaboration among domain scientists, psychometricians, and AI researchers. More broadly, it argues that AI should be understood not as a replacement for human judgment, but as a new type of co-worker,a collaborative technology that, when responsibly designed and governed, can advance scientific progress while promoting equity, access, and social benefit.

Teams link: https://teams.microsoft.com/meet/385793921641525?p=5MWSAZLjczXbS0rSKP

References:

von Davier, A. A. & Yan, D. (Eds.). (2026). Artificial intelligence in educational learning and assessment [Manuscript submitted]. Springer.

Burstein, J., LaFlair, G. T., Yancey, K., von Davier, A. A., & Dotan, R. (2024). Responsible AI for test equity and quality: The Duolingo English Test as a case study. arXiv Preprint arXiv:2409.07476