TacChat: Exploring User Generation and Accessible Sharing via Tactile Graphics for the Blind and Visually Impaired
Yuewen Zhang, Danning Li, Kexin Zhou, Pu Jiang, Jingyang Lyu, Kotaro Hara, Baoshui Chen, Shuai Gao, Yang Jiao · 2026 · Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems (CHI EA '26) · doi:10.1145/3772363.3798574
Summary
Zhang and colleagues (Tsinghua University, Singapore Management University, and Beijing School for the Blind) flip the usual framing of tactile-graphics research: rather than asking how blind and low-vision (BLV) people can better consume tactile images created by sighted experts, they ask what happens when BLV people become the authors. They present TacChat, a technology probe that chains an inexpensive 3D printing pen (PCL thermoplastic that leaves a raised, tactually explorable line on paper) with a 60x60 pin-array refreshable tactile display (RTD). The pipeline has three phases. Explore and Author: the BLV author builds a mental model of the target (a route, object, or scene) and sketches it with the 3D pen on canvas. Convert and Modify: a mobile-phone camera photographs the sketch; a three-step image-processing module (crop, skeleton extraction via the Zhang-Suen thinning algorithm, downsampling) renders the drawing as a binary tactile graphic at three user-selectable fidelity levels, applying predefined geometric templates to clean up common sketch flaws (wavy straight lines, imperfect curves, misaligned squares). The author previews and refines on the RTD. Share and Communicate: the graphic plus a verbal narration is pushed over the network to other BLV users' RTD devices. A user study with N=15 BLV participants (mostly late-blind, 20s-30s, all braille readers with prior tactile-graphics experience) used a 5-round Latin-square-counterbalanced design in which participants took turns in three roles — Author (sketches and narrates), Listener (receives narration only), and TacChat User (receives both tactile graphic and narration) — across two tasks: spatial cognition (indoor routes) and object recognition (plush toys).
Key findings
The TacChat condition significantly outperformed verbal-only narration on both communication measures. In Stage 1 (route comprehension) the TacChat group's completion rate was higher (p=0.002) and completion time lower than the listener-only group; in Stage 2 (toy selection) TacChat completion time was also lower (p=0.003), and correctness showed a 20% advantage (though p=0.109). Qualitative thematic analysis of 40-50 minute semi-structured interviews surfaced three insights. First, BLV authors make different design choices than sighted authors: 'sighted people often add too many details... making it difficult to read and understand. BLV authors however typically draw from their own reading experiences, retaining only essential information.' Participants expressed that in some cases BLV authors may produce tactile graphics better suited to BLV readers than sighted authors do. Second, BLV participants strongly preferred self-directed iteration and seamless, integrated authoring tools, citing discomfort with 'having to use it under the supervision of sighted people' as a persistent barrier. Third, the sharing scenario had pronounced social-emotional value — introverted participants opened up through mutual trust, and the authors observed that tactile sharing of personal stories, hobbies, and possessions can help establish social connections among BLV community members, reducing dependence on AI image recognition or sighted helpers to interpret visual-only social media.
Relevance
This paper makes an important point for accessibility practitioners and product teams: accessibility is not only about passive consumption of content but also about enabling creative self-expression. Most commercial assistive products for BLV users (screen readers, OCR apps, AI image describers, even most tactile-graphic production services) treat BLV users as audiences for content produced by sighted authors. TacChat instead treats them as authors, which has direct design implications for social-media accessibility, educational authoring tools, accessible art and photography practices, and peer-to-peer communication within BLV communities. The finding that BLV authors produce graphics with better abstraction for BLV readers is a provocation to design teams that currently rely entirely on sighted experts for tactile content. The technical pipeline — cheap 3D pen for capture, skeleton-based simplification for conversion, RTD for preview and delivery — is practical enough to inspire real-world accessible-authoring tools. Limitations: the 60x60 RTD resolution is coarse, the conversion pipeline assumes simple line drawings, the task set is narrow (routes and plush toys), sample size is small (N=15), participants were mostly late-blind with prior tactile-graphics experience, and no comparison was made with sighted-authored graphics — an obvious next step the authors flag.
Tags: tactile graphics · refreshable tactile display · blind and low vision · accessible authoring · social media accessibility · technology probe · accessibility research · self-expression · non-visual interaction