Assessing the Deaf User Perspective on Sign Language Avatars
Michael Kipp, Quan Nguyen, Alexis Heloir, Silke Matthes · 2011 · Proceedings of the 13th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2011) · doi:10.1145/2049536.2049557
Summary
This paper investigates Deaf community attitudes toward sign language avatars — virtual characters that perform sign language — through focus groups and a large-scale online study. Sign language avatars have the potential to make written content accessible to Deaf people whose primary language is sign language and who may have limited reading fluency. Unlike video recordings of human signers, avatars offer advantages including anonymity, cost-effective content production, dynamic content insertion, and customisable appearance. However, avatar research is characterised by a disconnect: most researchers are hearing, and Deaf people typically have little exposure to avatar technology. Commissioned by the German Federal Ministry of Labour and Social Affairs, the researchers conducted two focus groups (N=8) in a pure sign language environment using a dual-moderator approach (a Deaf community member as moderator, a formerly deaf person with cochlear implant as assistant), followed by an accessible online survey (N=317, 85% deaf) with DGS (German Sign Language) video explanations for all questions. Participants were shown 3-6 existing avatar videos spanning different sign languages and animation quality levels, and discussed acceptability, shortcomings, and potential applications.
Key findings
The most critical feedback concerned nonmanual components rather than hand movements — participants ranked facial expression (7 votes), natural movement (5), mouthing (4), and emotions (4) as the most important avatar aspects. Current avatars were described as stiff, emotionless, and robot-like. Mouthing (mouth patterns derived from spoken language) was identified as particularly important for DGS comprehension, and its absence or poor synchronisation with signs was highly disturbing. Participants wanted more natural upper body, shoulder, and torso movement, and noted that signing space was too restricted (mainly horizontal/vertical, lacking sideways depth). The hand-animated DeafWorld avatar received significantly higher ratings than automated avatars, highlighting the large quality gap that remains. For applications, participants envisioned avatars for one-way, simple communication: train station announcements, simple help dialogues, lexicons, news, and fixed texts — but not for complex dialogue or emotional content. A key concern was that avatars should never replace human sign language interpreters (25% of online respondents feared job cuts). Crucially, merely participating in the studies significantly increased positive attitudes toward avatars (p<0.001 for both focus groups and online study), demonstrating that inclusion itself builds acceptance.
Relevance
This research provides essential guidance for anyone developing sign language avatar technology. The finding that nonmanual features (facial expression, mouthing, emotion, body posture) matter more to Deaf users than hand shape accuracy challenges the technical focus of much avatar research. For accessibility practitioners, the study demonstrates that technology acceptance among Deaf people requires genuine community involvement — not just usability testing, but meaningful participation in research design and evaluation. The methodological contributions are also valuable: conducting focus groups in a pure sign language environment with Deaf moderators, using visual materials (icons, images, video) rather than text, and complementing qualitative focus groups with quantitative online surveys to validate findings at scale. The concerns about avatars replacing interpreters and reducing motivation for literacy learning are important ethical considerations that should inform deployment decisions. The study reinforces that Deaf people want avatars as a complement to, not a replacement for, human communication.
Tags: deaf · sign language avatar · German Sign Language · user acceptance · focus groups · participatory design · sign language synthesis · nonmanual features · accessibility