Hate Personified: Investigating the role of LLMs in content moderation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Masud, Sarah, Singh, Sahajpreet, Hangya, Viktor, Fraser, Alexander, Chakraborty, Tanmoy
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917793726726144
author Masud, Sarah
Singh, Sahajpreet
Hangya, Viktor
Fraser, Alexander
Chakraborty, Tanmoy
author_facet Masud, Sarah
Singh, Sahajpreet
Hangya, Viktor
Fraser, Alexander
Chakraborty, Tanmoy
contents For subjective tasks such as hate detection, where people perceive hate differently, the Large Language Model's (LLM) ability to represent diverse groups is unclear. By including additional context in prompts, we comprehensively analyze LLM's sensitivity to geographical priming, persona attributes, and numerical information to assess how well the needs of various groups are reflected. Our findings on two LLMs, five languages, and six datasets reveal that mimicking persona-based attributes leads to annotation variability. Meanwhile, incorporating geographical signals leads to better regional alignment. We also find that the LLMs are sensitive to numerical anchors, indicating the ability to leverage community-based flagging efforts and exposure to adversaries. Our work provides preliminary guidelines and highlights the nuances of applying LLMs in culturally sensitive cases.
format Preprint
id arxiv_https___arxiv_org_abs_2410_02657
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Hate Personified: Investigating the role of LLMs in content moderation
Masud, Sarah
Singh, Sahajpreet
Hangya, Viktor
Fraser, Alexander
Chakraborty, Tanmoy
Computation and Language
Computers and Society
For subjective tasks such as hate detection, where people perceive hate differently, the Large Language Model's (LLM) ability to represent diverse groups is unclear. By including additional context in prompts, we comprehensively analyze LLM's sensitivity to geographical priming, persona attributes, and numerical information to assess how well the needs of various groups are reflected. Our findings on two LLMs, five languages, and six datasets reveal that mimicking persona-based attributes leads to annotation variability. Meanwhile, incorporating geographical signals leads to better regional alignment. We also find that the LLMs are sensitive to numerical anchors, indicating the ability to leverage community-based flagging efforts and exposure to adversaries. Our work provides preliminary guidelines and highlights the nuances of applying LLMs in culturally sensitive cases.
title Hate Personified: Investigating the role of LLMs in content moderation
topic Computation and Language
Computers and Society
url https://arxiv.org/abs/2410.02657