Bridging Visual Asymmetry: Exploring AI-Mediated Communication Support for Parents with Visual Impairments and Their Sighted Children in Outdoor Informal Learning

Paper Title

Bridging Visual Asymmetry: Exploring AI-Mediated Communication Support for Parents with Visual Impairments and Their Sighted Children in Outdoor Informal Learning

Publication Info

• Topic area: AI-mediated assistive technology for mixed-ability family learning
• Keywords: visual impairment, AI mediation, parent-child interaction, informal learning, accessibility, assistive technology, joint attention, outdoor learning, mixed-ability families, Symbiotic Learning

Background and Problem

• Problem / challenge: Parents with visual impairments (PVI) face significant barriers in engaging their sighted children during outdoor informal learning due to perceptual asymmetries, knowledge gaps, and interactional challenges. Existing assistive technologies focus on individual compensation rather than reciprocal interaction, leaving a gap in supporting joint attention and co-learning.
• Significance: Addressing these challenges is crucial for fostering equitable parent-child relationships, enhancing children’s exploratory learning, and empowering PVI to actively participate in their children’s education.
• Motivation and related work: Prior research has explored assistive technologies for PVI and informal learning tools for sighted families but has not adequately addressed the unique needs of visually asymmetric families in outdoor learning contexts. This paper builds on these areas by integrating conversational scaffolding with accessible sensing to support real-time educational engagement.

Solution

• Proposed approach: Development of Bond, a distributed Wizard-of-Oz (WoZ) prototype that combines a child-worn camera and parent-facing mobile interface to provide real-time, context-aware prompts for joint attention and dialogue.
• Novelty:
  1. Introduction of Symbiotic Learning as a conceptual framework for AI-mediated relational support in mixed-ability families.
  2. Development of Bond, which bridges perceptual asymmetries through child-led interest detection and parent-directed prompts.
  3. Empirical evidence demonstrating Bond’s impact on enhancing parental responsiveness, dialogue quality, and children’s exploratory motivation.
  4. Design guidelines for AI systems that promote inclusive, co-constructed learning in visually asymmetric contexts.
• Procedure and key techniques:
  • Conducted interviews and observations with 11 families to identify challenges and derive design principles.
  • Held design consultations with PVI to refine interaction modalities and system specifications.
  • Developed Bond with two core functions: interest detection (child-led activation and contextualized descriptions) and prompt generation (learning guidance and knowledge expansion).
  • Evaluated Bond through a WoZ study with 14 families in real-world outdoor settings, analyzing interaction quality and system acceptance.

Results

• Concrete findings:
  • Significant improvements in parental response rates (M=2.03 vs. M=1.58, p<.001) and effective guidance instances (M=12.29 vs. M=5.00, p<.01).
  • Enhanced conversational depth, with more conceptual and strategic talk observed in system-supported conditions.
  • Increased parental self-efficacy across all dimensions, including confidence in guiding exploration (p<.001).
  • High system acceptance (Perceived Usefulness M=4.59, Ease of Use M=4.62, Trust M=4.51).
• Advantage over baselines: Bond enabled more balanced conversational engagement, deeper exploration, and greater parental confidence compared to baseline conditions without system support.
• Experiments / evaluation:
  • Within-subject WoZ study with 14 families in parks, museums, and insect houses.
  • Mixed-method analysis combining thematic coding, learning talk categorization, and quantitative measures (e.g., response rates, turn-taking).
  • Pre- and post-use assessments of parental self-efficacy and technology acceptance.
• Limitations and future work:
  • Short-term study duration without long-term follow-up.
  • Limited personalization of system content and support.
  • Need for validation of educational frameworks in PVI contexts.
  • Ethical concerns regarding privacy, AI dependency, and potential biases.

Summary

This paper addresses the challenges faced by PVI in engaging their sighted children during outdoor informal learning by introducing Bond, an AI-mediated system that bridges perceptual asymmetries and fosters joint exploration. Through a WoZ study with 14 families, Bond demonstrated significant improvements in parental responsiveness, dialogue quality, and children’s exploratory motivation. The proposed Symbiotic Learning framework highlights the relational potential of AI in supporting inclusive, co-constructed learning. Future work should focus on long-term evaluations, adaptive personalization, and broader validation of educational frameworks in accessibility research.