Why Generative AI and relational knowledge remain fundamentally incompatible

The friction between Generative AI and relational knowledge

Generative AI fails to replicate the site-specific, intergenerational wisdom of farming and heritage, exposing a gap between data processing and relational understanding. Large Language Models (LLMs) excel at pattern matching but lack the physical and temporal context required for land-based transmission. This suggests that digital replicas of human expertise remain superficial at best.

Traditional knowledge systems rely on tacit, lived experience passed through generations. Currently, the push for digital efficiency risks replacing these complex human networks with flattened, algorithmic outputs. This study examines how ChatGPT processes inquiries regarding agriculture and cultural heritage, identifying a significant loss in translation when compared to human-led transmission.

Researchers applied Critical Discourse Analysis and autoethnography to interrogate AI-generated responses. The data showed that the model prioritises generic information over the specific, relational links that define cultural and agricultural practices. It suggests that AI often strips the human element from the knowledge it purports to organise, resulting in a sterile representation of active traditions.

The findings indicate that reliance on AI for heritage preservation or agricultural planning could lead to the erosion of localised expertise. Future systems must find ways to honour human-centric practices rather than merely scraping them for data. This study does not provide a technical framework for encoding tacit, land-based data into future neural networks.

• AI simplifies complex, intergenerational wisdom into generic data points.
• Relational knowledge requires physical context that LLMs cannot currently access.
• Digital tools risk eroding localised expertise if used without human oversight.