ValueDCG: Measuring Comprehensive Human Value Understanding Ability of Language Models

TL;DR

This paper introduces ValueDCG, a new metric to evaluate how well large language models understand human values, focusing on both factual knowledge and underlying reasons, revealing that larger models do not necessarily improve in this understanding.

Contribution

The paper proposes a comprehensive evaluation metric, ValueDCG, for assessing LLMs' understanding of human values, emphasizing the importance of both knowledge and reasoning aspects.

Findings

LLMs' understanding of human values does not scale with model size.

Larger models may generate plausible explanations without true value comprehension.

Evaluation reveals potential risks of LLMs misrepresenting human values.

Abstract

Personal values are a crucial factor behind human decision-making. Considering that Large Language Models (LLMs) have been shown to impact human decisions significantly, it is essential to make sure they accurately understand human values to ensure their safety. However, evaluating their grasp of these values is complex due to the value's intricate and adaptable nature. We argue that truly understanding values in LLMs requires considering both "know what" and "know why". To this end, we present a comprehensive evaluation metric, ValueDCG (Value Discriminator-Critique Gap), to quantitatively assess the two aspects with an engineering implementation. We assess four representative LLMs and provide compelling evidence that the growth rates of LLM's "know what" and "know why" capabilities do not align with increases in parameter numbers, resulting in a decline in the models' capacity to understand human values as larger amounts of parameters. This may further suggest that LLMs might craft plausible explanations based on the provided context without truly understanding their inherent value, indicating potential risks.