I-VAMOS: Independent Voting with Accessible Multimodal Offline System for Visually Impaired Users

Paper Title

I-VAMOS: Independent Voting with Accessible Multimodal Offline System for Visually Impaired Users

Publication Info

• Topic area: Accessible voting systems for blind and low-vision users
• Keywords: Blind voters, low-vision voters, accessible voting, multimodal feedback, OCR, tactile aids, independent voting, usability, workload, paper ballots

Background and Problem

• Problem / challenge: Blind and low-vision voters (BLVs) face barriers to independent and confidential voting due to limitations in existing aids like tactile sleeves, braille ballots, and electronic ballot-marking devices (BMDs). These methods often compromise usability, accessibility, privacy, or security.
• Significance: Ensuring independent and secret voting is a constitutional principle critical to democratic participation. Addressing these barriers would empower BLV voters and uphold electoral integrity.
• Motivation and related work: Prior solutions include tactile overlays, braille ballots, and electronic BMDs, but these approaches have limitations such as reliance on braille literacy, high costs, security vulnerabilities, and lack of inclusivity across vision, literacy, and age groups. This paper aims to overcome these gaps by introducing a system that supports independent paper-ballot voting without specialized templates or online connectivity.

Solution

• Proposed approach: I-VAMOS (Independent Voting with Accessible Multimodal Offline System), an offline multimodal voting assistance system integrating OCR-based speech guidance, visual enhancements, and a spring-loaded stamping device to enable independent voting for BLVs.
• Novelty:
  1. A multimodal system combining auditory, visual, and tactile feedback for accessible paper-ballot voting.
  2. Offline operation ensuring cost-effectiveness and security without reliance on online connectivity.
  3. Empirical evidence showing improved usability, accuracy, and reduced workload across diverse BLV subgroups.
  4. Design implications for scalable and inclusive voting systems.
• Procedure and key techniques:
  • Hardware: Ballot slide frame for secure paper alignment, webcam for OCR input, and spring-loaded stamp for precise marking.
  • Software: OCR pipeline for candidate recognition, auditory feedback manager for speech guidance, and visual enhancer for magnified and high-contrast text display.
  • Participatory design: Iterative refinement with BLV users to address practical challenges and ensure usability.

Results

• Concrete findings:
  • Stamping accuracy improved from 75.0% (KEC tools) to 91.7% (I-VAMOS).
  • Usability (SUS) scores increased from 52.8 to 79.1.
  • Workload (NASA-TLX) reduced from 46.2 to 26.1.
  • OCR precision achieved 98.9%.
• Advantage over baselines:
  • Higher accuracy, usability, and reduced workload compared to tactile sleeves and magnifying glasses provided by the Korea Election Commission (KEC).
  • Mitigated disparities across vision status, braille literacy, and age groups.
• Experiments / evaluation:
  • User study with 16 BLV participants (balanced by vision, literacy, and age).
  • Comparison of I-VAMOS with KEC tools using metrics like accuracy, completion time, SUS, and NASA-TLX.
  • Mock ballots modeled on official formats; randomized task order and candidate selection.
• Limitations and future work:
  • Limited generalizability due to single-session study and mock ballots.
  • Longer task completion times compared to existing aids.
  • Future work includes adapting I-VAMOS to diverse ballot formats and collaborating with election authorities for real-world deployment.

Summary

The paper introduces I-VAMOS, a multimodal offline voting system designed to enable blind and low-vision voters to cast paper ballots independently and confidentially. Empirical results from a user study demonstrate significant improvements in accuracy, usability, and workload compared to existing aids, though task completion times were longer. The system effectively mitigates disparities across vision, literacy, and age groups, offering scalable feedback modes tailored to diverse accessibility needs. Future directions include adapting the system to various ballot formats and integrating it into real-world electoral workflows.