Designing Beyond Language: Sociotechnical Barriers in AI Health Technologies for Limited English Proficiency

Paper Title

Designing Beyond Language: Sociotechnical Barriers in AI Health Technologies for Limited English Proficiency

Publication Info

• Topic area: Sociotechnical barriers and AI design considerations for healthcare technologies targeting Limited English Proficiency (LEP) populations.
• Keywords: AI in healthcare, Limited English Proficiency, sociotechnical barriers, cultural sensitivity, patient navigators, digital literacy, privacy concerns, healthcare equity, language barriers, AI design.

Background and Problem

• Problem / challenge: LEP patients in the U.S. face systemic barriers in healthcare that extend beyond language, including cultural misunderstandings, privacy concerns, and low literacy. Existing AI health technologies often fail to address these challenges, risking the exacerbation of inequities.
• Significance: Addressing these barriers is critical to improving healthcare access and outcomes for over 25 million LEP individuals in the U.S., who are disproportionately affected by misdiagnoses, poor treatment adherence, and communication breakdowns.
• Motivation and related work: While digital health technologies and AI-powered tools (e.g., translation systems, chatbots) have shown potential, they are often designed for digitally literate, English-speaking users, neglecting cultural and linguistic nuances. This paper builds on prior work by exploring how AI can be designed to equitably support LEP populations, focusing on Spanish-speaking individuals.

Solution

• Proposed approach: The study investigates AI design considerations for LEP healthcare through storyboard-driven interviews with patient navigators who support Spanish-speaking LEP individuals.
• Novelty:
  1. Identifies linguistic, cultural, and systemic barriers faced by LEP patients in healthcare.
  2. Explores perceived risks and opportunities for AI to support LEP patient experiences.
  3. Proposes design guidelines for AI tools to mediate culturally sensitive healthcare interactions.
• Procedure and key techniques:
  • Conducted 14 storyboard-centered interviews with patient navigators.
  • Storyboards depicted hypothetical AI use cases in healthcare scenarios.
  • Analyzed findings using thematic analysis to identify barriers, risks, and opportunities for AI.

Results

• Concrete findings:
  • Linguistic barriers include dialectal differences and limited support for indigenous languages.
  • Cultural disconnects, such as reliance on traditional remedies, hinder full disclosure to providers.
  • Low literacy (reading, digital, and health) limits engagement with digital health tools.
  • Privacy concerns, especially among undocumented immigrants, discourage technology use.
• Advantage over baselines: AI could reduce social barriers (e.g., embarrassment in human interactions) and alleviate resource constraints (e.g., interpreter shortages, navigator workload) if designed with cultural and literacy considerations.
• Experiments / evaluation:
  • Participants identified AI opportunities, such as on-demand translation and simplified explanations, and risks, such as loss of human connection and misinformation.
  • Design implications include multimodal interfaces, privacy-preserving models, and culturally sensitive communication styles.
• Limitations and future work:
  • Findings are based on navigator perspectives, not direct patient experiences.
  • Storyboards reflect hypothetical scenarios, not observed behaviors.
  • Future work should include participatory design with LEP patients and providers to validate and extend findings.

Summary

This study highlights the sociotechnical barriers faced by Spanish-speaking LEP patients in accessing healthcare and explores how AI technologies could address these challenges. Through interviews with 14 patient navigators, the authors identify linguistic, cultural, literacy, and privacy barriers, as well as opportunities for AI to enhance communication and care workflows. However, risks such as loss of human connection and misinformation must be carefully managed. The findings inform design guidelines for culturally sensitive, privacy-preserving, and accessible AI tools, with future work needed to directly engage LEP patients in the design process.