Why AI is Still a Distant Promise for Deaf Students

Fewer than one in every 50 Kenyan teachers is trained to work with learners who have special needs.

For deaf students, that gap shapes almost every school day.

The Teachers Service Commission counts more than 439,000 teachers nationwide.

Only 9,340 of them, about 2.1%, are trained to handle learners with special needs, according to

Education News.

No public data shows how many teachers are fluent in Kenyan Sign Language specifically.

That shortage is sharpest in tertiary institutions, where deaf students often sit in classes built for hearing students and lean on a small pool of interpreters just to follow the lesson.

A Dataset Years in the Making

At Maseno University, computational linguist Dr. Lilian Wanzare is trying to close that gap with technology.

Wanzare leads AI4KSL, a project building a digital signing avatar that translates spoken English into Kenyan Sign Language.

Her team pulled in 400 deaf participants and 48 teachers and tutors to build the dataset behind it.

The result is close to 20,000 signed videos and 14,000 English-KSL sentence pairs, one of the largest KSL datasets ever assembled for AI research.

“We have a huge database of recordings of different signs,” Wanzare said during a webinar.

A follow-up publication by the AI4KSL team found real differences in how learners sign.

Girls in the dataset produced more fluent signs than boys, and used facial expressions and other non-manual signals more often.

Wanzare built that finding into the avatar itself, giving it both male and female representations to avoid bias and offer learners a mentor who signs the way they do.

What Teachers Actually Want

Nela Jebet teaches geography and Christian religious education at Ngala Secondary School for the Deaf.

Jebet has noticed the same gender pattern in her own classroom, where girls often sign more easily and boys take longer to respond.

The AI tool she wants most is simple: something that turns speech into Kenyan Sign Language as she talks.

“You just talk to it, it signs,” she said.

A tool like that, she believes, would save teachers real time and energy.

Aaron Mutembei, a deaf teacher at the same school, has heard about the avatar project but has not used it yet.

Mutembei already leans on general AI chatbots like ChatGPT and Copilot to build lesson plans and generate images for his students.

“The biggest sense of learning for a deaf student is their eyes,” he said.

He thinks a speech-to-sign tool is possible eventually, but only with far more KSL examples to train on.

A Technology Barely in the Room

At Rev Muhoro School for the Deaf in Nyeri County, teacher Simon Kaiga says AI sign language tools have barely reached his classroom.

“We haven’t embraced it that much,” Kaiga said.

His closest brush with the technology has been transcription software used occasionally in staff meetings.

Not everyone in the deaf community is convinced AI can handle sign language at all.

Alexis Bosire, a KSL tutor and founder of the Africa Centre for Disability Studies, argues the technology is being asked to do something fundamentally different from other AI tasks.

“Sign language is sign language, either you know the language or you don’t,” Bosire said.

Interpretation, he argues, requires reorganizing entire sentences to match how sign language actually works, not just swapping words for signs.

Even so, Bosire sees promise in animated avatars as a learning aid, especially for people just picking up the language.

Regional Signs Complicate the Picture

Kaiga’s bigger worry is whether any AI system can capture how much sign language varies from place to place.

He wants Kenya to harmonize KSL terms nationally before any large-scale AI training begins.

There is also an unresolved debate inside the deaf community over Signed Exact English, which mirrors spoken word order, versus standard KSL, which has its own grammar.

Mutembei sees the same regional variation play out in his own school.

Signs used in Mombasa can differ sharply from signs used in the Rift Valley, he said, and teachers and students often invent their own signs when no standard one exists.

“It is a big challenge indeed if AI were to be trained for KSL,” Mutembei said.

On the question of AI grading student signing, Mutembei stays open, but only with heavy training behind it.

Kaiga is far more cautious.

“I wouldn’t advocate for AI to be involved, especially at this stage,” he said.

For Kaiga, the bigger obstacle is not the software but the teachers behind it, since many still struggle with basic digital tools.

Where the Data Falls Short

Lincence Boge, business development and partnerships lead at Signvrse, a company building AI-powered sign language translation, says data is the industry’s core problem.

Widely spoken languages have decades of digital text and audio to train on.

Kenyan Sign Language has almost none, forcing developers like Signvrse to collect training data from scratch.

Boge said that work only holds up if deaf communities are involved at every stage, not brought in after the fact.

In a deaf unit at a school in Baba Dogo, Nairobi, interpreter Faith Wahu already uses AI chatbots daily to simplify materials for students at different reading levels.

For Wahu, tools like AI4KSL remain distant from what actually happens in her classroom each morning.

Teachers across the country describe the same gap.

The technology is moving forward in university labs, but in most classrooms with deaf students, it has yet to arrive at all.

This article was produced as part of the Gender+AI Reporting Fellowship, with support from the Africa Women’s Journalism Project in partnership with DW Akademie.