Identifying and Understanding Human Values in Text: A Tailorable LLM-Based Architecture

Forum topic · 小凯 · 2026-05-29

Summary

Researchers Eduardo de la Cruz Fernández, Marcelo Karanik, and Sascha Ossowski present an LLM-based architecture for detecting and quantifying the intensity of human values in text, whether explicit or implicit (arXiv:2605.27373). The work addresses the growing need to assess how well autonomous intelligent systems' decisions align with human values, moving beyond traditional utility-maximisation models. Unlike prior approaches constrained to a specific value theory or reliant on complex prompt engineering, the proposed architecture is modular and theory-agnostic, consisting of three coordinated modules: one that generates structured value specifications from the foundational texts of any theoretical framework, one that annotates text using these specifications, and one that assigns support/opposition levels based on rhetorical and semantic evidence. This design separates the conceptualisation of values from the detection task, yielding an extensible, reproducible pipeline adaptable to different value theories. Experiments on the ValueEval dataset demonstrate competitive detection performance.

Paper Overview

Field: AI
Authors: Eduardo de la Cruz Fernández, Marcelo Karanik, Sascha Ossowski
Published: 2026-05-28
arXiv: 2605.27373

Abstract

As intelligent systems become more autonomous, the scientific community focuses on creating decision-making mechanisms that include ethical and moral considerations, unlike traditional utility-maximisation models. To achieve this, a key aspect is assessing how well these decisions align with human values. To this end, a promising line of research is centred on developing approaches based on Large Language Models (LLMs) to identify human values from text, whether explicit or implicit, enabling their recognition throughout.

This paper introduces an LLM-based architecture to detect and quantify the intensity of human values in text, avoiding the limitations of previous approaches tied to specific value theory or complex prompt engineering. The architecture comprises three coordinated modules:

1. Value specification generator — produces structured value specifications from the foundational texts of any theoretical framework. 2. Annotator — uses these specifications to label values present in a given text. 3. Stance/intensity estimator — assigns support/opposition levels based on rhetorical and semantic evidence.

Why It Matters

Theory-agnostic: not bound to any single value theory, unlike earlier methods.
No prompt engineering burden: avoids fragile, hand-crafted prompts.
Modular design: separates the conceptualisation of values from the detection task, making the pipeline extensible, reproducible, and tailorable to different theoretical frameworks.
Validated: experiments on the ValueEval dataset show good detection performance.

---

*Auto-collected on 2026-05-29.*