How far can we trust the results of an AI-marked speaking assessment?

There has been a lot of speculation about the future impact of AI on language teaching and learning, but in this webinar I will talk about a live use of AI. I will discuss the creation of an automated speaking test, looking at the following questions in particular:

• How can a machine assess a learner’s ability to speak English?
• Will its scores be as reliable as those of a human?
• Can it assess communicative competence as well as technical competence in pronunciation, fluency, vocabulary, etc.?

I will talk about the way automated marking works, how it compares with human marking, and how to address issues of validity. I will discuss these questions in relation both to current research and to what was learned during the design and development of ClarityEnglish’s Dynamic Speaking Test, which was launched at IATEFL 2024.

Speaker- Martin Moore