MoVa: Towards Generalizable Classification of Human Morals and Values
This provides a generalizable tool for researchers in communication and psychology to analyze morals and values, with potential implications for machine alignment, though it appears incremental in combining existing methods.
The paper tackles the problem of identifying human morals and values in language by introducing MoVa, a suite of resources including 16 labeled datasets and a lightweight LLM prompting strategy that outperforms fine-tuned models across multiple domains and frameworks.
Identifying human morals and values embedded in language is essential to empirical studies of communication. However, researchers often face substantial difficulty navigating the diversity of theoretical frameworks and data available for their analysis. Here, we contribute MoVa, a well-documented suite of resources for generalizable classification of human morals and values, consisting of (1) 16 labeled datasets and benchmarking results from four theoretically-grounded frameworks; (2) a lightweight LLM prompting strategy that outperforms fine-tuned models across multiple domains and frameworks; and (3) a new application that helps evaluate psychological surveys. In practice, we specifically recommend a classification strategy, all@once, that scores all related concepts simultaneously, resembling the well-known multi-label classifier chain. The data and methods in MoVa can facilitate many fine-grained interpretations of human and machine communication, with potential implications for the alignment of machine behavior.