Bottom-up and Top-Down Analysis of Values, Agendas, and Observations in Corpora and LLMs
Abstract
Large language models (LLMs) generate diverse, situated, persuasive texts from a plurality of potential perspectives, influenced heavily by their prompts and training data. As part of LLM adoption, we seek to characterize---and ideally, manage---the socio-cultural values that they express, for reasons of safety, accuracy, inclusion, and cultural fidelity. We present a validated approach to automatically (1) extracting heterogeneous latent value propositions from texts, (2) assessing resonance and conflict of values with texts, and (3) combining these operations to characterize the pluralistic value alignment of human-sourced and LLM-sourced textual data.
Cite
Text
Friedman et al. "Bottom-up and Top-Down Analysis of Values, Agendas, and Observations in Corpora and LLMs." NeurIPS 2024 Workshops: Pluralistic-Alignment, 2024.Markdown
[Friedman et al. "Bottom-up and Top-Down Analysis of Values, Agendas, and Observations in Corpora and LLMs." NeurIPS 2024 Workshops: Pluralistic-Alignment, 2024.](https://mlanthology.org/neuripsw/2024/friedman2024neuripsw-bottomup/)BibTeX
@inproceedings{friedman2024neuripsw-bottomup,
title = {{Bottom-up and Top-Down Analysis of Values, Agendas, and Observations in Corpora and LLMs}},
author = {Friedman, Scott E. and Benkler, Noam and Mosaphir, Drisana and Rye, Jeffrey and Schmer-Galunder, Sonja M. and Goldwater, Micah and McLure, Matthew and Wheelock, Ruta and Gottlieb, Jeremy and Goldman, Robert P. and Miller, Christopher},
booktitle = {NeurIPS 2024 Workshops: Pluralistic-Alignment},
year = {2024},
url = {https://mlanthology.org/neuripsw/2024/friedman2024neuripsw-bottomup/}
}