Paper Type

ERF

Abstract

Large language models (LLMs) offer transformative potential for personalized patient education, yet their clinical utility is restricted by misinformation and a failure to align with Health Literacy (HL) principles. This paper presents a blueprint for a Design-Constrained LLM (DC-LLM). We translate Nutbeam’s HL constructs - functional, interactive, and critical - into technical constraints, including readability gating, contextual adaptation, and evidence of provenance. We outline a four-phase evaluation plan to validate the system’s comprehension, trust, and clinical adherence. This framework moves generative AI beyond mere text simplification toward a safe, interactive, and evidence-based patient education tool.

Paper Number

1616

Comments

SIGHealth

Share

COinS
 
Aug 15th, 12:00 AM

From Readability to Fidelity: A Design-Constrained LLM for Health-Literacy-Aligned Patient Education

Large language models (LLMs) offer transformative potential for personalized patient education, yet their clinical utility is restricted by misinformation and a failure to align with Health Literacy (HL) principles. This paper presents a blueprint for a Design-Constrained LLM (DC-LLM). We translate Nutbeam’s HL constructs - functional, interactive, and critical - into technical constraints, including readability gating, contextual adaptation, and evidence of provenance. We outline a four-phase evaluation plan to validate the system’s comprehension, trust, and clinical adherence. This framework moves generative AI beyond mere text simplification toward a safe, interactive, and evidence-based patient education tool.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.