On Epistemic Diversity in Large Language Models

Large language models (LLMs) are increasingly used not only to retrieve information, but to answer questions, explain, teach, and support inquiry. In such settings, evaluation cannot be exhausted by accuracy, fairness, or alignment alone. A system may give a correct answer while still narrowing users' access to knowledge if it repeatedly surfaces only one framing, one explanation, or one reasoning strategy where multiple valid alternatives exist. We call this desideratum epistemic diversity: the range of valid answers, explanations, and reasoning routes that an LLM makes available to users. We argue that epistemic diversity is a distinct evaluative and normative property of LLMs, grounded in ideas from philosophy of science and social epistemology, where plurality is valued as a condition for robust inquiry and understanding. We propose a minimal framework for conceptualizing and measuring epistemic diversity in LLMs, and operationalize it in two domains. We find that frontier LLMs often exhibit epistemic narrowness, repeatedly collapsing large valid answer spaces onto a small canonical subset. These findings suggest that LLM evaluation should move beyond accuracy-oriented paradigms and treat epistemic diversity as an important dimension of model capability.