From Static Outputs to Living Evidence: AI for Integrated Knowledge Translation in Canadian Health Research

Area of Science:

• Health Services Research
• Information Science
• Knowledge Management

Background:

• Current integrated knowledge translation methods rely on static outputs like reports and presentations.
• These static formats struggle to adapt to the dynamic information needs of decision-makers.
• There is a need for more agile and responsive knowledge-sharing mechanisms.

Purpose of the Study:

• To propose a conceptual design for leveraging retrieval-augmented large language models (LLMs) in integrated knowledge translation.
• To outline the necessary governance structures for implementing such LLM-based systems.
• To enhance the speed and traceability of evidence synthesis and exchange.

Main Methods:

• Conceptual design and framework development for LLM integration.
• Emphasis on governance principles including provenance, privacy, transparency, and equity.
• Discussion of the potential of LLMs to create a conversational layer over curated knowledge assets.

Main Results:

• LLMs can provide a secure, auditable conversational interface for accessing and synthesizing program outputs and research materials.
• This approach facilitates rapid and traceable knowledge synthesis, adaptable to evolving decision-maker questions.
• Governed LLM infrastructure can reduce friction in evidence exchange and shorten the knowledge-to-use lag.

Conclusions:

• Retrieval-augmented LLMs represent a promising infrastructure for modernizing integrated knowledge translation.
• Robust governance is essential for ensuring secure, ethical, and equitable implementation.
• Adoption of these tools can significantly accelerate the translation of knowledge into practice.