Accelerate Knowledge Transformation with Curation and Feedback Systems

This chapter explores the challenges of knowledge management in AI-first settings, focusing on the tension between documented and undocumented information. The scenario of a Network Controller facing a decision under ambiguity highlights the reliance on informal relationships and improvisation in traditional settings. The core issue is not the absence of information but the lack of trusted authority to discern truth from noise. The chapter introduces the Knowledge Transformation Cycle and its Curate phase, which involves synthesizing feedback to improve data quality. The Knowledge Curation Pyramid is proposed as a framework to categorize information by trust and liability, aiming to establish a Canon of trusted truth for AI systems. The chapter underscores the necessity of structured curation to prevent ambiguity from causing unpredictability in automated decision-making processes. A hierarchical model of information quality for AI systems is introduced, categorizing data into four layers: Well-retrieved, Well-sourced, Well-informed, and Canon. The base layer, Well-retrieved, involves raw data retrieval with high variance, suitable for research but not actionable decisions. Well-sourced information is derived from trusted sources, though content-specific validation is pending.