# Prompt-Based Value Steering of Large Language Models

- Authors: Giulio Antonio Abbo and Tony Belpaeme
- Date: 2025-10-13
- Type: preprint
- Venue: Value Engineering in Artificial Intelligence (to appear in LNCS) (ECAI '25 (workshop VALE))
- DOI: https://doi.org/10.48550/arXiv.2511.16688
- Code: https://github.com/giubots/value-steering
- Supplementary material: https://doi.org/10.5281/zenodo.17609013
- PDF: https://giubots.net/publications/prompt-based-value/prompt-based-value.pdf
- Research line: [Values and Generative AI](https://giubots.net/research/values-generative-ai/)
- URL: https://giubots.net/publications/prompt-based-value/


## TL;DR

This paper introduces a method to measure how well a candidate prompt can steer large language models’ responses toward specific human values, without fine-tuning the model. This is extremely important in dynamic conversational contexts where values, preferences, and interlocutors can change frequently. We apply this method on a dialogue dataset using Schwartz’s theory, and we show that a simple prompt asking to maximise for a specific value can effectively increase its presence in the generated text.


This paper introduces a method to measure how well a candidate prompt can steer large language models' responses toward specific human values, without fine-tuning the model.
This is extremely important in dynamic conversational contexts where values, preferences, and interlocutors can change frequently.
We apply this method on a dialogue dataset using Schwartz’s theory, and we show that a simple prompt asking to maximise for a specific value can effectively increase its presence in the generated text.

## Abstract

Large language models are increasingly used in applications where alignment with human values is critical. While model fine-tuning is often employed to ensure safe responses, this technique is static and does not lend itself to everyday situations involving dynamic values and preferences. In this paper, we present a practical, reproducible, and model-agnostic procedure to evaluate whether a prompt candidate can effectively steer generated text toward specific human values, formalising a scoring method to quantify the presence and gain of target values in generated responses. We apply our method to a variant of the Wizard-Vicuna language model, using Schwartz’s theory of basic human values and a structured evaluation through a dialogue dataset. With this setup, we compare a baseline prompt to one explicitly conditioned on values, and show that value steering is possible even without altering the model or dynamically optimising prompts.

## How to cite

Giulio Antonio Abbo and Tony Belpaeme. 2025. Prompt-Based Value Steering of Large Language Models. Presented at ECAI '25, Value Engineering in Artificial Intelligence (VALE '25). Preprint.

```bibtex
@misc{abbo2025prompt-based-values,
author="Abbo, Giulio Antonio
and Belpaeme, Tony",
title="Prompt-Based Value Steering of Large Language Models",
booktitle="Value Engineering in Artificial Intelligence",
year="2025",
}
```