Unstructured LLM use contributes to epistemic narrowing by privileging convergent outputs and dominant viewpoints¶

Claim · All claims
Evidence · 1 study · theoretical r3 · q1
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q1 i? LLMs optimised for convergence reduce learners' exposure to diverse, ambiguous or contested knowledge and reproduce mainstream, largely Euro-American viewpoints. → Mireia Vendrell 2026

Evidence¶

Mireia Vendrell 2026¶

Mireia Vendrell, Samantha-Kaye Johnston. (2026). Scaffolding critical thinking with generative AI: Design principles for integrating large language models in higher education. Computers and Education: Artificial Intelligence. https://doi.org/10.1016/j.caeai.2026.100572

q1 · i? · theoretical · r3

The article's theoretical argument (type e) that convergent LLM outputs 'prioritise fluency and coherence over exploration and uncertainty', compounded by training data that reproduces dominant cultural and epistemic norms; no empirical test is reported for this claim.

"These systems are typically optimised for convergence, producing singular, self-contained answers that prioritise fluency and coherence over exploration and uncertainty."

Discussion¶

Related Claims¶

• Current LLMs reproduce linguistic hierarchies by being built on dominant standard English varieties, producing a monolithic English that ignores regional variation — related
• Intensive LLM use in educational contexts may lead to an accumulation of cognitive debt that weakens critical thinking and epistemic ownership — related
• AI systems are non-neutral socio-technical artifacts embedding cultural assumptions and power relations — a broader claim this one bears on
• Unstructured LLM use risks cognitive offloading that displaces effortful thought and weakens argument construction — related
• More stylistic information processing makes individuals' viewpoints narrower and more rigid, reducing ability to adopt opposite styles — related