Enhancing American sign language learning with LLM-assisted feedback: a comparative study with traditional methods
Enhancing American sign language learning with LLM-assisted feedback: a comparative study with traditional methods
We evaluate LLM-assisted learning for American Sign Language acquisition by comparing GPT-4o-powered real-time feedback informed by gesture recognition, against traditional image/text static instruction. In a study with 20 participants, both methods improved performance, but the LLM group showed greater gains in challenging signs and increased engagement with complex content. Effect sizes suggest meaningful advantages for LLM support, despite the limited statistical significance of the findings.
Wang, Jindi
a5af917b-2ac6-4173-8435-c08963dc7ed7
Ivrissimtzis, Ioannis
af7d236c-09e2-4b93-baa5-35fdece8a758
Li, Zhaoxing
65935c45-a640-496c-98b8-43bed39e1850
Shi, Lei
3e73da43-5e7e-4544-a327-9de0046edfa2
9 September 2025
Wang, Jindi
a5af917b-2ac6-4173-8435-c08963dc7ed7
Ivrissimtzis, Ioannis
af7d236c-09e2-4b93-baa5-35fdece8a758
Li, Zhaoxing
65935c45-a640-496c-98b8-43bed39e1850
Shi, Lei
3e73da43-5e7e-4544-a327-9de0046edfa2
Wang, Jindi, Ivrissimtzis, Ioannis, Li, Zhaoxing and Shi, Lei
(2025)
Enhancing American sign language learning with LLM-assisted feedback: a comparative study with traditional methods.
20th IFIP TC13 International Conference on Human-Computer Interaction, Belo Horizonte, Minas Gerais, Brazil.
09 - 12 Sep 2025.
5 pp
.
Record type:
Conference or Workshop Item
(Paper)
Abstract
We evaluate LLM-assisted learning for American Sign Language acquisition by comparing GPT-4o-powered real-time feedback informed by gesture recognition, against traditional image/text static instruction. In a study with 20 participants, both methods improved performance, but the LLM group showed greater gains in challenging signs and increased engagement with complex content. Effect sizes suggest meaningful advantages for LLM support, despite the limited statistical significance of the findings.
Text
_camera_ready__INTERACT_2025
- Accepted Manuscript
Restricted to Repository staff only
Request a copy
More information
Published date: 9 September 2025
Venue - Dates:
20th IFIP TC13 International Conference on Human-Computer Interaction, Belo Horizonte, Minas Gerais, Brazil, 2025-09-09 - 2025-09-12
Identifiers
Local EPrints ID: 506801
URI: http://eprints.soton.ac.uk/id/eprint/506801
PURE UUID: 2c0db719-0b5e-4c4b-9411-b868881ee723
Catalogue record
Date deposited: 18 Nov 2025 17:58
Last modified: 20 Nov 2025 03:07
Export record
Contributors
Author:
Jindi Wang
Author:
Ioannis Ivrissimtzis
Author:
Zhaoxing Li
Author:
Lei Shi
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics