When Handshape Is Blurred, Meaning Is Blurred
- Tim Scannell
- Feb 15
- 3 min read
Updated: Feb 16
Generative AI is now being used to convert written text or spoken language into sign language avatars for websites, videos and public communication.
Innovation in accessibility is positive. However, when working with British Sign Language (BSL), accuracy is essential.
BSL is a complete language. It has its own grammar, structure and visual rules. If these rules are not clear, meaning can change.
HOLME: The Foundation of BSL
Every sign in BSL depends on five key elements. These are often called HOLME:
H – Handshape
O – Orientation
L – Location
M – Movement
E – Expression
If one element changes, the meaning can change.
For example:
A finger slightly bent instead of straight
A palm facing the wrong direction
Movement that is too short or unclear
Facial expression that does not match the grammar
These are not small details. They affect meaning.
In BSL, visual clarity equals linguistic clarity.
Challenges with Early and Current Generative AI
Generative AI systems predict movement based on data patterns. They aim to create smooth and realistic motion.
However, early systems may show:
Blurred fingers during movement
Unclear handshape transitions
Small changes in orientation
Reduced facial expression
Sign order influenced by English structure
These issues may not be obvious to non-signers. But for fluent BSL users, they can reduce clarity and trust.
When handshape is blurred, meaning is blurred.
BSL Is Not English on the Hands
BSL does not follow English word order.
It has its own:
Sentence structure
Grammar
Use of space
Facial grammar
Looking only at English subtitles is not enough to judge BSL quality.
BSL must be evaluated based on BSL standards.
Where you may notice this in everyday life
AI-generated signing and sign avatars are increasingly appearing in places where clarity really matters—often without viewers realising what to look for. Pay attention to BSL quality in:
Social media clips and auto-generated “accessible” versions of videos
Public transport announcements, station screens, and service disruption updates
Healthcare videos, appointment instructions, consent information, and public health messaging
Government and public services guidance, emergency alerts, and community updates
Education and training platforms, onboarding videos, and compliance content
Customer service web widgets, banking/insurance explainer videos, and appointment portals
If you see blurred handshape, weak facial grammar, English-led sign order, or unclear orientation/movement, treat it as a signal to question accuracy—especially when information is safety-critical.
Questions to Ask Before Investing
Many organisations are spending significant budgets on AI accessibility tools.
Before adopting these systems, important questions include:
How is HOLME tested?
Are Deaf native signers involved?
Is BSL grammar reviewed separately from English text?
What quality standards are in place?
Accessibility must include accuracy.
BSL Was Strong Before Generative AI
BSL research, education and community leadership existed long before generative AI.
AI is a tool.
It should support the language, not simplify it.
Technology must respect the standards already built by the Deaf community.
Moving Forward
Generative AI has potential.
But precision must come first.
When handshape is unclear, meaning becomes unclear.
When expression is reduced, grammar is reduced.
BSL deserves accuracy and respect.
Accessibility is more than presence. Its accuracy.
Innovation and linguistic integrity must move forward together—with Deaf expertise at the centre.
Disclaimer: This article discusses general issues in AI-generated sign language and does not refer to any specific organisation or product.
Comments