arXiv:2601.14295v4 Announce Type: replace-cross Abstract: Large language models increasingly function as artificial reasoners: they evaluate arguments, assign credibility, and express confidence. Yet their belief-forming behavior is governed by implicit, uninspected epistemic policies. This paper argues for an epistemic constitution for AI: explicit, contestable meta-norms that regulate how systems form and express beliefs. Source attribution bias provides the motivating case: I show that frontier models enforce identity-stance coherence, penalizing arguments attributed to sources whose expect
As large language models become increasingly embedded in critical reasoning tasks, the need to understand and control their inherent biases is immediate and growing.
A strategic reader should care because unchecked AI epistemic biases can lead to systemic misinformation, skewed decision-making, and erosion of trust in AI systems.
The explicit call for 'epistemic constitutionalism' shifts the conversation from merely identifying AI biases to architecting normative frameworks for AI belief formation.
- · AI ethicists
- · AI governance frameworks
- · Organizations prioritizing explainable AI
- · Developers of meta-normative AI systems
- · AI systems with uninspected implicit biases
- · Users relying on unchallenged AI outputs
- · Organizations deploying black-box AI for critical reasoning
Increased focus on designing AI systems with transparent and controllable epistemic policies will emerge.
New standards and regulations may be developed that mandate 'epistemic constitutions' for advanced AI.
The concept of 'truth' and 'credibility' within AI systems could become a battleground for philosophical and political influence.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL