The work Deaf teachers do between an AI answer and an English lesson

One Deaf university instructor in Tehran described teaching English prepositions with AI-generated pictures of a cat and a box. Students compared the positions and wrote corresponding sentences. The pictures gave them a relationship to examine and express in English, moving the activity beyond matching an English word to a Persian translation.[1]

The account appears in a study published in Frontiers in Education on 30 September by Hessameddin Ghanbar, Mohadeseh Asghari and Yasamin Pourhematkhanshir. They interviewed 12 instructors at one specialist university: ten hearing teachers of English as a foreign language and two Deaf academics. These are teachers’ accounts of practice and experimentation; the researchers did not observe classes or independently measure students’ learning.[1]

What the interviews reveal is the work between obtaining an AI output and giving a learner something useful to do with it. Teachers chose visual examples, shortened explanations, checked material with Deaf colleagues and added signed summaries. Those decisions determined how the technology entered a lesson.[1]

The language of the output is part of the task

At the university, English teaching draws on written English, Persian explanations, visual material and interpretation in Esharani, or Iranian Sign Language. Students arrive with different histories of access to those languages. The study describes uneven opportunities to learn both Persian and Esharani before university, so neither a page of Persian nor an English transcript can be assumed to provide the same access for everyone.[1]

The instructor who described the cat-and-box lesson is identified as P11 to protect her identity. She reported stronger engagement and higher quiz performance than in the previous year. That comparison was her assessment, with no independently collected results establishing how much the activity changed learning. The documented contribution is the teaching choice: represent a spatial relationship, invite comparison and ask students to construct a sentence.[1]

The other Deaf academic, P12, described teaching English conditionals through a locally controlled activity supported by Python. It combined Persian explanations with visual timelines showing conditions, causes and consequences. Again, the AI material had a particular job inside a lesson: make the structure of a grammatical relationship available for inspection.[1]

These examples suggest a useful starting question for an educator: what should the learner be able to compare, explain or produce after using this material? A visually polished answer can still leave that question unresolved. The task gives the material a purpose against which a teacher and learner can judge it.

Checking accuracy also means checking access

P11 said she checked generated images and text before using them and, when possible, asked a Deaf teaching assistant to review them too. She also provided short signed video summaries, giving students another route into material that might otherwise depend on written text. Her process joined content checking with a decision about how an explanation would be communicated.[1]

Hearing instructors described related editing work. One reported extracting the useful parts of lengthy grammar feedback and explaining them in simpler language. Another had to show students how to form a prompt step by step. In these accounts, producing more language was easy; choosing language the learner could use remained a teaching task.[1]

A failed activity made the distinction more immediate. P11 recalled a group discussion in which four Deaf students tried speaking to a voice-to-text tool. It failed to produce meaningful text, and she reported that the students felt humiliated. Her account shows why the route into a tool matters alongside the output: requiring spoken input can itself become a barrier.[1]

The Deaf instructors also challenged the idea that visible output necessarily supplies sign-language access. Esharani includes spatial grammar, facial expression and body movement. Captions, generic avatars and pictures have different functions; the teachers did not treat them as interchangeable with communication in that language. They wanted human interpreters to remain part of teaching and hearing colleagues to continue developing their signing skills.[1]

Keep the lesson available when the service is not

P12 supplied a practical reason to retain other ways of teaching: a four-hour internet outage stopped his AI tools, so he returned to a whiteboard. Other instructors described unequal access to devices, reliable home internet and paid services. A resource that works on campus may still be difficult for a student to revisit afterwards.[1]

The participants proposed Deaf-led testing, shared teaching resources and more offline or locally hosted options. The researchers recommend involving Deaf students and educators in judging prospective tools against the actual language and access conditions of their institution.[1]

For someone preparing an AI-supported lesson, the cat-and-box example offers a concrete place to begin. Specify the relationship to teach, check how the material represents it and give the learner a way to express their understanding. The useful finish is a sentence the student can make their own.

Report an error or suggest a correction