Shared Nature, Unique Nurture: PRISM for Pluralistic Reasoning via In-context Structure Modeling

TL;DR

This paper introduces PRISM, a system that enhances large language models with dynamic, on-the-fly epistemic graphs to foster pluralistic reasoning, significantly increasing diversity and creativity in outputs, and improving rare-disease diagnosis.

Contribution

It proposes a novel paradigm of pluralistic AI using in-context structure modeling with epistemic graphs, enabling models to generate diverse perspectives and improve real-world problem solving.

Findings

PRISM achieves state-of-the-art novelty on creativity benchmarks.

PRISM uncovers correct long-tail diagnoses missed by standard LLMs.

PRISM significantly expands distributional diversity in model outputs.

Abstract

Large Language Models (LLMs) are converging towards a singular Artificial Hivemind, where shared Nature (pre-training priors) result in a profound collapse of distributional diversity, limiting the distinct perspectives necessary for creative exploration and scientific discovery. To address this, we propose to equip models with inference-time Nurture (individualized epistemic trajectories) using Epistemic Evolution paradigm, progressing through explore, internalize, and express. We instantiate this via PRISM (Pluralistic Reasoning via In-context Structure Modeling), a model-agnostic system that augments LLM with dynamic On-the-fly Epistemic Graphs. On three creativity benchmarks, PRISM achieves state-of-the-art novelty and significantly expands distributional diversity. Moreover, we evaluate the real-world utility via a challenging rare-disease diagnosis benchmark. Results demonstrate that PRISM successfully uncovers correct long-tail diagnoses that standard LLM miss, confirming that its divergence stems from meaningful exploration rather than incoherent noise. Overall, this work establishes a new paradigm for Pluralistic AI, moving beyond monolithic consensus toward a diverse ecosystem of unique cognitive individuals capable of collective, multi-perspective discovery.