The Granularity Axis: A Micro-to-Macro Latent Direction for Social Roles in Language Models

TL;DR

This paper uncovers a dominant latent direction in language models that encodes social role granularity, from individual to institutional levels, and demonstrates its causal influence on model responses.

Contribution

It introduces the Granularity Axis as a principal latent direction in language models that captures social role hierarchy and shows how to manipulate it to control response granularity.

Findings

The Granularity Axis aligns with the principal component of role representations (cosine 0.972).

Projections along the axis increase monotonically across social role levels.

Activation steering along the axis shifts response granularity in models.

Abstract

Large language models (LLMs) are routinely prompted to take on social roles ranging from individuals to institutions, yet it remains unclear whether their internal representations encode the granularity of such roles, from micro-level individual experience to macro-level organizational, institutional, or national reasoning. We show that they do. We define a contrast-based Granularity Axis as the difference between mean macro- and micro-role hidden states. In Qwen3-8B, this axis aligns with the principal axis (PC1) of the role representation space at cosine 0.972 and accounts for 52.6% of its variance, indicating that granularity is the dominant geometric axis organizing prompted social roles. We construct 75 social roles across five granularity levels and collect 91,200 role-conditioned responses over shared questions and prompt variants, then extract role-level hidden states and project them onto the axis. Role projections increase monotonically across all five levels, remain stable across layers, prompt variants, endpoint definitions, held-out splits, and score-filtered subsets, and transfer to Llama-3.1-8B-Instruct. The axis is also causally relevant: activation steering along it shifts response granularity in the predicted direction, with Llama moving from 2.00 to 3.17 on a five-point macro scale under positive steering on prompts that admit local responses. The two models differ in controllability, suggesting that steering depends on each model's default operating regime. Overall, our findings suggest that social role granularity is not merely a stylistic surface feature, but a structured, ordered, and causally manipulable latent direction in role-conditioned language model behavior.